Fix EXIF viewer breaking on binary MakerNote fields

Pentax and other cameras have binary EXIF fields (MakerNote, etc.) that
contain raw bytes. The previous code used errors="replace" which still
produced strings with replacement characters that broke JSON parsing.

Now properly detect non-printable binary data and display as
"<N bytes binary>" instead.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

.gitignore

@@ -97,3 +97,8 @@ rpi/*.tar.zst.zip
 rpi/*.img
 rpi/*.img.zst
 rpi/*.img.zst.zip
+
+# AUR build artifacts
+aur-upload/
+aur/.SRCINFO
+aur/*.pkg.tar.zst

CLI.md

@@ -164,7 +164,7 @@ stegasoo generate [OPTIONS]
 | `--pin/--no-pin` | | flag | `--pin` | Generate a PIN |
 | `--rsa/--no-rsa` | | flag | `--no-rsa` | Generate an RSA key |
 | `--pin-length` | | 6-9 | 6 | PIN length in digits |
-| `--rsa-bits` | | choice | 2048 | RSA key size (2048, 3072, 4096) |
+| `--rsa-bits` | | choice | 2048 | RSA key size (2048, 3072) |
 | `--words` | | 3-12 | 4 | Words in passphrase |
 | `--output` | `-o` | path | | Save RSA key to file |
 | `--password` | `-p` | string | | Password for RSA key file |
@@ -180,7 +180,7 @@ stegasoo generate
 stegasoo generate --words 6
 
 # Generate with RSA key
-stegasoo generate --rsa --rsa-bits 4096
+stegasoo generate --rsa --rsa-bits 3072
 
 # Save RSA key to encrypted file
 stegasoo generate --rsa -o mykey.pem -p "mysecretpassword"

RELEASE_NOTES.md

@@ -1,49 +1,83 @@
-## Stegasoo v4.1.7
+## Stegasoo v4.2.0
 
-### Mobile UI Polish
+### Performance Optimizations
-- **PIN Entry**: Shrunk digit boxes for 9-digit PIN support on mobile
-- **Mode Selectors**: DCT/LSB buttons now use consistent button-group styling with icons
-- **Navbar**: Left-aligned collapsed menu, shortened channel fingerprint display (`ABCD-••••-3456`)
-- **Text Wrapping**: Fixed button text wrapping issues on narrow screens
 
-### Docker Improvements
+Major performance improvements for Raspberry Pi and resource-constrained deployments.
-- **Reorganized**: Docker files moved to `docker/` directory
-  - `docker/Dockerfile`
-  - `docker/Dockerfile.base`
-  - `docker/docker-compose.yml`
-- **DCT Fix**: Added Reed-Solomon (`reedsolo`) to Docker images - fixes DCT decode failures
-- **Quick Start**: New `docs/DOCKER_QUICKSTART.md` guide
 
-```bash
+#### DCT Vectorization (~14x faster)
-# Build and run
+- Batch DCT processing using `scipy.fft.dctn` with `axes=(1,2)`
-docker build -f docker/Dockerfile.base -t stegasoo-base:latest .
+- Processes 500 blocks at once instead of one-by-one
-docker-compose -f docker/docker-compose.yml up -d
+- Decode time reduced from ~2.6s to ~0.8s on 1MB images
-```
 
-### Raspberry Pi
+#### Memory Optimization (50% reduction)
-- **First-Boot Wizard**: Can now load existing channel key (for joining team deployments)
+- Switched from `float64` to `float32` for all DCT operations
-- **Project Cleanup**: Moved `pishrink.sh` to `rpi/tools/`
+- Peak RAM: 211 MB → 107 MB for encode, 104 MB → 52 MB for decode
+- Critical for Pi 3/4 avoiding swap thrashing
 
-### UI Copy
+#### Progress Callbacks for Decode
-- Changed "Undetectable" to "Covertly Embedded" on encode page (more accurate)
+- `progress_file` parameter added to `decode()` and extraction functions
+- UI can now show decode progress (phases: loading, extracting, decoding, complete)
+- JSON format: `{"current": 80, "total": 100, "percent": 80.0, "phase": "decoding"}`
+
+#### Async API Endpoints
+- Encode/decode operations now run in thread pool via `asyncio.to_thread()`
+- API server can handle concurrent requests without blocking
+- Essential for multi-user Pi deployments
+
+### Compression
+
+#### Zstd Default Compression
+- `zstandard` is now a core dependency (always installed)
+- Better compression ratio than zlib for QR code RSA keys
+- New `STEGASOO-ZS:` prefix for zstd, backward compatible with `STEGASOO-Z:` (zlib)
+
+### QR Code Generation
+
+#### CLI Support
+- `stegasoo generate --rsa --qr key.png` - save RSA key as QR image (PNG/JPG)
+- `stegasoo generate --rsa --qr-ascii` - print ASCII QR to terminal
+
+#### API Support
+- `POST /generate-key-qr` - generate QR from RSA key
+- Supports `png`, `jpg`, and `ascii` output formats
+- Uses zstd compression by default
+
+### Other Changes
+
+- RSA key size capped at 3072 bits (4096 too large for QR codes)
+- File auto-expire increased to 10 minutes
+- Progress bar "candy cane" animation during Argon2 key derivation
+- Optional API service in Pi setup (with security warning)
+
+### Summary
+
+| Metric | v4.1.7 | v4.2.0 | Improvement |
+|--------|--------|--------|-------------|
+| Decode (1MB) | ~2.6s | ~0.8s | **70% faster** |
+| Peak RAM | 211 MB | 107 MB | **50% less** |
+| Concurrent API | No | Yes | ✓ |
+| QR Compression | zlib | zstd | **~15% smaller** |
 
 ### Raspberry Pi Image
-Download `stegasoo-rpi-4.1.7.img.zst.zip` from Releases.
+Download `stegasoo-rpi-4.2.0_final.img.zst` from Releases.
 
 ```bash
 # Flash (auto-detects SD card)
-sudo ./rpi/flash-image.sh stegasoo-rpi-4.1.7.img.zst.zip
+sudo ./rpi/flash-image.sh stegasoo-rpi-4.2.0_final.img.zst
 
 # Or manual
-unzip -p stegasoo-rpi-4.1.7.img.zst.zip | zstdcat | sudo dd of=/dev/sdX bs=4M status=progress
+zstdcat stegasoo-rpi-4.2.0_final.img.zst | sudo dd of=/dev/sdX bs=4M status=progress
 ```
 
 Default login: `admin` / `stegasoo`
 
-First boot runs the setup wizard for WiFi, HTTPS, and channel key configuration.
-
 ### Docker
 ```bash
+# Build and run
+docker build -f docker/Dockerfile.base -t stegasoo-base:latest .
+docker-compose -f docker/docker-compose.yml up -d
+
+# Or individual services
 docker-compose -f docker/docker-compose.yml up -d web  # Web UI on :5000
 docker-compose -f docker/docker-compose.yml up -d api  # REST API on :8000
 ```

WEB_UI.md

@@ -411,7 +411,7 @@ Create a new set of credentials for steganography operations.
 | Use PIN | on/off | on | Generate a numeric PIN |
 | PIN length | 6-9 | 6 | Digits in the PIN |
 | Use RSA Key | on/off | off | Generate an RSA key pair |
-| RSA key size | 2048/3072/4096 | 2048 | Key size in bits |
+| RSA key size | 2048/3072 | 2048 | Key size in bits |
 
 #### Entropy Calculator
 

aur/PKGBUILD

@@ -0,0 +1,116 @@
+# Maintainer: Aaron D. Lee <your-email@example.com>
+pkgname=stegasoo-git
+pkgver=4.2.0.r0.g530e5de
+pkgrel=1
+pkgdesc="Secure steganography with hybrid photo + passphrase + PIN authentication"
+arch=('x86_64')
+url="https://github.com/adlee-was-taken/stegasoo"
+license=('MIT')
+
+# Python 3.11-3.14 supported (uses jpeglib for modern Python compatibility)
+depends=(
+    'python>=3.11'
+    'zbar'  # QR code reading for Web UI
+)
+makedepends=(
+    'git'
+    'python'
+    'python-build'
+    'python-hatchling'
+)
+provides=('stegasoo')
+conflicts=('stegasoo')
+install=stegasoo-git.install
+source=("${pkgname}::git+https://github.com/adlee-was-taken/stegasoo.git#branch=main")
+sha256sums=('SKIP')
+
+pkgver() {
+    cd "$pkgname"
+    git describe --long --tags 2>/dev/null | sed 's/^v//;s/\([^-]*-g\)/r\1/;s/-/./g' || \
+    printf "%s.r%s.g%s" "4.2.0" "$(git rev-list --count HEAD)" "$(git rev-parse --short HEAD)"
+}
+
+build() {
+    cd "$pkgname"
+    python -m build --wheel --no-isolation
+}
+
+package() {
+    cd "$pkgname"
+
+    # Detect Python version for site-packages path
+    local pyver=$(python -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}")')
+
+    # Install to /opt/stegasoo with dedicated venv
+    install -dm755 "$pkgdir/opt/stegasoo"
+
+    # Create fresh venv in package
+    python -m venv "$pkgdir/opt/stegasoo/venv"
+
+    # Install the wheel with all extras
+    local wheel=$(ls dist/*.whl | head -1)
+    "$pkgdir/opt/stegasoo/venv/bin/pip" install --no-cache-dir "${wheel}[all]"
+
+    # Install frontends (not included in wheel)
+    local site_packages="$pkgdir/opt/stegasoo/venv/lib/python${pyver}/site-packages"
+    cp -r frontends "$site_packages/"
+
+    # Create writable directories for stegasoo user
+    install -dm755 "$pkgdir/opt/stegasoo/venv/var/app-instance"
+    install -dm755 "$site_packages/frontends/web/temp_files"
+    install -dm755 "$site_packages/frontends/api/temp_files"
+
+    # Fix shebangs - replace build-time paths with installed paths
+    find "$pkgdir/opt/stegasoo/venv/bin" -type f -exec \
+        sed -i "s|$pkgdir/opt/stegasoo/venv|/opt/stegasoo/venv|g" {} \;
+
+    # Fix pyvenv.cfg
+    sed -i "s|$pkgdir||g" "$pkgdir/opt/stegasoo/venv/pyvenv.cfg"
+
+    # Create symlinks to /usr/bin
+    install -dm755 "$pkgdir/usr/bin"
+    ln -s /opt/stegasoo/venv/bin/stegasoo "$pkgdir/usr/bin/stegasoo"
+
+    # Install license
+    install -Dm644 LICENSE "$pkgdir/usr/share/licenses/$pkgname/LICENSE"
+
+    # Install docs
+    install -Dm644 README.md "$pkgdir/usr/share/doc/$pkgname/README.md"
+
+    # Install systemd service files
+    install -Dm644 /dev/stdin "$pkgdir/usr/lib/systemd/system/stegasoo-web.service" <<EOF
+[Unit]
+Description=Stegasoo Web UI
+After=network.target
+
+[Service]
+Type=simple
+User=stegasoo
+WorkingDirectory=/opt/stegasoo/venv/lib/python${pyver}/site-packages/frontends/web
+Environment="PATH=/opt/stegasoo/venv/bin"
+ExecStart=/opt/stegasoo/venv/bin/gunicorn -b 127.0.0.1:5000 app:app
+Restart=on-failure
+RestartSec=5
+
+[Install]
+WantedBy=multi-user.target
+EOF
+
+    install -Dm644 /dev/stdin "$pkgdir/usr/lib/systemd/system/stegasoo-api.service" <<EOF
+[Unit]
+Description=Stegasoo REST API
+After=network.target
+
+[Service]
+Type=simple
+User=stegasoo
+WorkingDirectory=/opt/stegasoo/venv/lib/python${pyver}/site-packages/frontends/api
+Environment="PATH=/opt/stegasoo/venv/bin"
+ExecStart=/opt/stegasoo/venv/bin/uvicorn main:app --host 127.0.0.1 --port 8000
+Restart=on-failure
+RestartSec=5
+
+[Install]
+WantedBy=multi-user.target
+EOF
+}

aur/README.md

@@ -0,0 +1,79 @@
+# Stegasoo AUR Package
+
+> **Note:** Uses Python 3.12 via `python312` AUR package (jpegio not yet compatible with 3.13)
+
+## Installation
+
+### From AUR (once published)
+```bash
+yay -S stegasoo-git
+# or
+paru -S stegasoo-git
+```
+
+### Manual build
+```bash
+git clone https://aur.archlinux.org/stegasoo-git.git
+cd stegasoo-git
+makepkg -si
+```
+
+## What Gets Installed
+
+- `/opt/stegasoo/venv/` - Self-contained Python 3.12 venv with all dependencies
+- `/usr/bin/stegasoo` - CLI symlink
+- `/usr/lib/systemd/system/stegasoo-web.service` - Web UI service
+- `/usr/lib/systemd/system/stegasoo-api.service` - REST API service
+
+## Optional Dependencies
+
+```bash
+# QR code reading from webcam/images
+sudo pacman -S zbar
+```
+
+All other dependencies are bundled in the venv.
+
+## Usage
+
+### CLI
+```bash
+stegasoo --help
+stegasoo generate --rsa --qr-ascii
+stegasoo encode -i carrier.jpg -r reference.jpg -m "secret" -P passphrase -p 123456
+```
+
+### Web UI (systemd)
+```bash
+# Create service user (first time)
+sudo useradd -r -s /usr/bin/nologin stegasoo
+
+# Start service
+sudo systemctl enable --now stegasoo-web
+
+# Access at http://localhost:5000
+```
+
+### REST API (systemd)
+```bash
+# Start service
+sudo systemctl enable --now stegasoo-api
+
+# Access at http://localhost:8000/docs
+```
+
+### Manual run (without systemd)
+```bash
+# Web UI
+/opt/stegasoo/venv/bin/python -m gunicorn -b 0.0.0.0:5000 \
+    --chdir /opt/stegasoo/venv/lib/python3.12/site-packages/frontends/web app:app
+
+# REST API
+/opt/stegasoo/venv/bin/uvicorn \
+    --app-dir /opt/stegasoo/venv/lib/python3.12/site-packages/frontends/api \
+    main:app --host 0.0.0.0 --port 8000
+```
+
+## Maintainer
+
+Aaron D. Lee

aur/stegasoo-git.install

@@ -0,0 +1,40 @@
+post_install() {
+    # Create stegasoo system user if it doesn't exist
+    if ! getent passwd stegasoo >/dev/null; then
+        useradd -r -s /usr/bin/nologin -d /opt/stegasoo stegasoo
+        echo "Created system user 'stegasoo'"
+    fi
+
+    # Set ownership of instance directory for Flask
+    chown -R stegasoo:stegasoo /opt/stegasoo/venv/var/app-instance 2>/dev/null || true
+
+    echo ""
+    echo "Stegasoo installed successfully!"
+    echo ""
+    echo "CLI usage:"
+    echo "  stegasoo --help"
+    echo ""
+    echo "To start the web UI:"
+    echo "  sudo systemctl start stegasoo-web"
+    echo ""
+    echo "To start the REST API:"
+    echo "  sudo systemctl start stegasoo-api"
+    echo ""
+}
+
+post_upgrade() {
+    post_install
+}
+
+pre_remove() {
+    # Stop services if running
+    systemctl stop stegasoo-web 2>/dev/null || true
+    systemctl stop stegasoo-api 2>/dev/null || true
+}
+
+post_remove() {
+    # Optionally remove the stegasoo user
+    # userdel stegasoo 2>/dev/null || true
+    echo "Stegasoo removed. User 'stegasoo' was not removed."
+    echo "To remove: userdel stegasoo"
+}

aur/test-build.sh

@@ -0,0 +1,22 @@
+#!/bin/bash
+# Test build the AUR package locally
+
+set -e
+
+cd "$(dirname "$0")"
+
+echo "=== Cleaning previous builds ==="
+rm -rf stegasoo-git pkg src *.pkg.tar.zst *.whl 2>/dev/null || true
+
+echo "=== Generating .SRCINFO ==="
+makepkg --printsrcinfo > .SRCINFO
+
+echo "=== Building package ==="
+makepkg -sf
+
+echo "=== Package built ==="
+ls -la *.pkg.tar.zst
+
+echo ""
+echo "To install: sudo pacman -U stegasoo-git-*.pkg.tar.zst"
+echo "To test:    makepkg -si"

docker/Dockerfile.base

@@ -33,7 +33,8 @@ RUN pip install --no-cache-dir \
     argon2-cffi>=23.0.0 \
     pillow>=10.0.0 \
     cryptography>=41.0.0 \
-    reedsolo>=1.7.0
+    reedsolo>=1.7.0 \
+    zstandard>=0.22.0
 
 # Install web/api framework packages (also stable)
 RUN pip install --no-cache-dir \
@@ -48,9 +49,9 @@ RUN pip install --no-cache-dir \
     lz4>=4.0.0
 
 # Verify key packages work
-RUN python -c "import jpegio; import scipy; import numpy; print('jpegio + scipy + numpy OK')"
+RUN python -c "import jpegio; import scipy; import numpy; import zstandard; print('jpegio + scipy + numpy + zstd OK')"
 
 # Label for tracking
 LABEL org.opencontainers.image.title="Stegasoo Base"
 LABEL org.opencontainers.image.description="Pre-compiled dependencies for Stegasoo"
-LABEL org.opencontainers.image.version="4.0.0"
+LABEL org.opencontainers.image.version="4.2.0"

docs/TEMPLATES.md

@@ -126,7 +126,7 @@ Quick reference for all Jinja2 templates in `frontends/web/templates/`.
 - `use_pin` - checkbox
 - `pin_length` - PIN digits (6-9)
 - `use_rsa` - checkbox
-- `rsa_bits` - key size (2048/3072/4096)
+- `rsa_bits` - key size (2048/3072)
 
 **Output panels:**
 - Passphrase display

frontends/api/main.py

@@ -1,10 +1,14 @@
 #!/usr/bin/env python3
 """
-Stegasoo REST API (v4.0.0)
+Stegasoo REST API (v4.2.0)
 
 FastAPI-based REST API for steganography operations.
 Supports both text messages and file embedding.
 
+CHANGES in v4.2.0:
+- Async encode/decode operations (run in thread pool)
+- Server can handle concurrent requests without blocking
+
 CHANGES in v4.0.0:
 - Added channel key support for deployment/group isolation
 - New /channel endpoints for key management
@@ -21,8 +25,10 @@ NEW in v3.0: LSB and DCT embedding modes.
 NEW in v3.0.1: DCT color mode and JPEG output format.
 """
 
+import asyncio
 import base64
 import sys
+from functools import partial
 from pathlib import Path
 from typing import Literal
 
@@ -68,13 +74,20 @@ from stegasoo.constants import (
 try:
     from stegasoo.qr_utils import (
         extract_key_from_qr,
+        generate_qr_ascii,
+        generate_qr_code,
         has_qr_read,
+        has_qr_write,
     )
 
     HAS_QR_READ = has_qr_read()
+    HAS_QR_WRITE = has_qr_write()
 except ImportError:
     HAS_QR_READ = False
+    HAS_QR_WRITE = False
     extract_key_from_qr = None
+    generate_qr_code = None
+    generate_qr_ascii = None
 
 
 # ============================================================================
@@ -357,6 +370,7 @@ class StatusResponse(BaseModel):
     version: str
     has_argon2: bool
     has_qrcode_read: bool
+    has_qrcode_write: bool  # v4.2.0: QR generation capability
     has_dct: bool
     max_payload_kb: int
     available_modes: list[str]
@@ -372,6 +386,32 @@ class QrExtractResponse(BaseModel):
     error: str | None = None
 
 
+class QrGenerateRequest(BaseModel):
+    """Request to generate QR code from RSA key."""
+
+    key_pem: str = Field(..., description="RSA private key in PEM format")
+    output_format: str = Field(
+        default="png",
+        description="Output format: 'png', 'jpg', or 'ascii'",
+    )
+    compress: bool = Field(
+        default=True,
+        description="Compress key data with zstd (recommended for larger keys)",
+    )
+
+
+class QrGenerateResponse(BaseModel):
+    """Response containing generated QR code."""
+
+    success: bool
+    format: str | None = None
+    qr_data: str | None = Field(
+        default=None,
+        description="Base64-encoded image data (for png/jpg) or ASCII string",
+    )
+    error: str | None = None
+
+
 class WillFitRequest(BaseModel):
     """Request to check if payload will fit."""
 
@@ -436,6 +476,27 @@ def _get_channel_info(channel_key: str | None) -> tuple[str, str | None]:
     return info["mode"], info.get("fingerprint")
 
 
+# ============================================================================
+# HELPER: ASYNC EXECUTION
+# ============================================================================
+
+
+async def run_in_thread(func, *args, **kwargs):
+    """
+    Run a CPU-bound function in a thread pool.
+
+    This allows the FastAPI server to handle other requests while
+    encode/decode operations are running. Essential for Pi deployments
+    where operations can take several seconds.
+
+    Usage:
+        result = await run_in_thread(encode, message=msg, carrier_image=carrier, ...)
+    """
+    if kwargs:
+        func = partial(func, **kwargs)
+    return await asyncio.to_thread(func, *args)
+
+
 # ============================================================================
 # ROUTES - STATUS & INFO
 # ============================================================================
@@ -469,6 +530,7 @@ async def root():
         version=__version__,
         has_argon2=has_argon2(),
         has_qrcode_read=HAS_QR_READ,
+        has_qrcode_write=HAS_QR_WRITE,
         has_dct=has_dct_support(),
         max_payload_kb=MAX_FILE_PAYLOAD_SIZE // 1024,
         available_modes=available_modes,
@@ -760,6 +822,51 @@ async def api_extract_key_from_qr(
         return QrExtractResponse(success=False, error=str(e))
 
 
+@app.post("/generate-key-qr", response_model=QrGenerateResponse)
+async def api_generate_key_qr(request: QrGenerateRequest):
+    """
+    Generate QR code from an RSA private key.
+
+    Supports PNG, JPG, and ASCII output formats.
+    Uses zstd compression by default for better QR code density.
+    """
+    if not HAS_QR_WRITE:
+        raise HTTPException(501, "QR code generation not available. Install qrcode library.")
+
+    try:
+        fmt = request.output_format.lower()
+
+        if fmt == "ascii":
+            ascii_qr = generate_qr_ascii(
+                request.key_pem,
+                compress=request.compress,
+                invert=False,
+            )
+            return QrGenerateResponse(success=True, format="ascii", qr_data=ascii_qr)
+
+        elif fmt in ("png", "jpg", "jpeg"):
+            import base64
+
+            qr_bytes = generate_qr_code(
+                request.key_pem,
+                compress=request.compress,
+                output_format=fmt,
+            )
+            qr_b64 = base64.b64encode(qr_bytes).decode("ascii")
+            return QrGenerateResponse(success=True, format=fmt, qr_data=qr_b64)
+
+        else:
+            return QrGenerateResponse(
+                success=False,
+                error=f"Unsupported format: {fmt}. Use 'png', 'jpg', or 'ascii'",
+            )
+
+    except ValueError as e:
+        return QrGenerateResponse(success=False, error=str(e))
+    except Exception as e:
+        return QrGenerateResponse(success=False, error=f"QR generation failed: {e}")
+
+
 # ============================================================================
 # ROUTES - GENERATE
 # ============================================================================
@@ -874,8 +981,9 @@ async def api_encode(request: EncodeRequest):
             request.embed_mode, request.dct_output_format, request.dct_color_mode
         )
 
-        # v4.0.0: Include channel_key
+        # v4.2.0: Run CPU-bound encode in thread pool
-        result = encode(
+        result = await run_in_thread(
+            encode,
             message=request.message,
             reference_photo=ref_photo,
             carrier_image=carrier,
@@ -950,8 +1058,9 @@ async def api_encode_file(request: EncodeFileRequest):
             request.embed_mode, request.dct_output_format, request.dct_color_mode
         )
 
-        # v4.0.0: Include channel_key
+        # v4.2.0: Run CPU-bound encode in thread pool
-        result = encode(
+        result = await run_in_thread(
+            encode,
             message=payload,
             reference_photo=ref_photo,
             carrier_image=carrier,
@@ -1021,8 +1130,9 @@ async def api_decode(request: DecodeRequest):
         ref_photo = base64.b64decode(request.reference_photo_base64)
         rsa_key = base64.b64decode(request.rsa_key_base64) if request.rsa_key_base64 else None
 
-        # v4.0.0: Include channel_key
+        # v4.2.0: Run CPU-bound decode in thread pool
-        result = decode(
+        result = await run_in_thread(
+            decode,
             stego_image=stego,
             reference_photo=ref_photo,
             passphrase=request.passphrase,
@@ -1150,8 +1260,9 @@ async def api_encode_multipart(
         # Get DCT parameters
         dct_params = _get_dct_params(embed_mode, dct_output_format, dct_color_mode)
 
-        # v4.0.0: Include channel_key
+        # v4.2.0: Run CPU-bound encode in thread pool
-        result = encode(
+        result = await run_in_thread(
+            encode,
             message=payload,
             reference_photo=ref_data,
             carrier_image=carrier_data,
@@ -1264,8 +1375,9 @@ async def api_decode_multipart(
         # QR code keys are never password-protected
         effective_password = None if rsa_key_from_qr else (rsa_password if rsa_password else None)
 
-        # v4.0.0: Include channel_key
+        # v4.2.0: Run CPU-bound decode in thread pool
-        result = decode(
+        result = await run_in_thread(
+            decode,
             stego_image=stego_data,
             reference_photo=ref_data,
             passphrase=passphrase,

frontends/cli/main.py

@@ -120,6 +120,7 @@ try:
     from stegasoo.qr_utils import (  # noqa: F401
         can_fit_in_qr,
         extract_key_from_qr_file,
+        generate_qr_ascii,
         generate_qr_code,
         has_qr_read,
         has_qr_write,
@@ -136,6 +137,9 @@ except ImportError:
     def has_qr_write() -> bool:
         return False
 
+    def generate_qr_ascii(*args, **kwargs):
+        raise RuntimeError("QR code generation not available")
+
 
 # ============================================================================
 # CLI SETUP
@@ -236,7 +240,7 @@ def format_channel_status_line(quiet: bool = False) -> str | None:
     help=f"PIN length (6-9, default: {DEFAULT_PIN_LENGTH})",
 )
 @click.option(
-    "--rsa-bits", type=click.Choice(["2048", "3072", "4096"]), default="2048", help="RSA key size"
+    "--rsa-bits", type=click.Choice(["2048", "3072"]), default="2048", help="RSA key size"
 )
 @click.option(
     "--words",
@@ -247,7 +251,13 @@ def format_channel_status_line(quiet: bool = False) -> str | None:
 @click.option("--output", "-o", type=click.Path(), help="Save RSA key to file (requires password)")
 @click.option("--password", "-p", help="Password for RSA key file")
 @click.option("--json", "as_json", is_flag=True, help="Output as JSON")
-def generate(pin, rsa, pin_length, rsa_bits, words, output, password, as_json):
+@click.option(
+    "--qr",
+    type=click.Path(),
+    help="Save RSA key QR code to file (png/jpg, uses zstd compression)",
+)
+@click.option("--qr-ascii", is_flag=True, help="Print RSA key as ASCII QR code to terminal")
+def generate(pin, rsa, pin_length, rsa_bits, words, output, password, as_json, qr, qr_ascii):
     """
     Generate credentials for encoding/decoding.
 
@@ -261,13 +271,18 @@ def generate(pin, rsa, pin_length, rsa_bits, words, output, password, as_json):
     Examples:
         stegasoo generate
         stegasoo generate --words 5
-        stegasoo generate --rsa --rsa-bits 4096
+        stegasoo generate --rsa --rsa-bits 3072
         stegasoo generate --rsa -o mykey.pem -p "secretpassword"
+        stegasoo generate --rsa --qr key.png
+        stegasoo generate --rsa --qr-ascii
         stegasoo generate --no-pin --rsa
     """
     if not pin and not rsa:
         raise click.UsageError("Must enable at least one of --pin or --rsa")
 
+    if (qr or qr_ascii) and not rsa:
+        raise click.UsageError("QR output requires --rsa to generate an RSA key")
+
     if output and not password:
         raise click.UsageError("--password is required when saving RSA key to file")
 
@@ -334,6 +349,33 @@ def generate(pin, rsa, pin_length, rsa_bits, words, output, password, as_json):
                 click.echo(creds.rsa_key_pem)
             click.echo()
 
+            # QR code output (v4.2.0)
+            if qr:
+                if not HAS_QR:
+                    click.secho("    ⚠️  QR code library not available", fg="yellow")
+                else:
+                    # Determine format from extension
+                    qr_path = Path(qr)
+                    ext = qr_path.suffix.lower()
+                    fmt = "jpeg" if ext in (".jpg", ".jpeg") else "png"
+
+                    qr_bytes = generate_qr_code(creds.rsa_key_pem, compress=True, output_format=fmt)
+                    qr_path.write_bytes(qr_bytes)
+                    click.secho(f"─── RSA KEY QR CODE ───", fg="green")
+                    click.secho(f"    Saved to: {qr}", fg="bright_white")
+                    click.secho("    ⚠️  Contains unencrypted private key!", fg="yellow")
+                    click.echo()
+
+            if qr_ascii:
+                if not HAS_QR:
+                    click.secho("    ⚠️  QR code library not available", fg="yellow")
+                else:
+                    click.secho("─── RSA KEY QR CODE (ASCII) ───", fg="green")
+                    click.secho("    ⚠️  Contains unencrypted private key!", fg="yellow")
+                    click.echo()
+                    ascii_qr = generate_qr_ascii(creds.rsa_key_pem, compress=True, invert=True)
+                    click.echo(ascii_qr)
+
         click.secho("─── SECURITY ───", fg="green")
         click.echo(f"    Passphrase entropy: {creds.passphrase_entropy} bits ({words} words)")
         if creds.pin:

frontends/web/app.py

@@ -31,7 +31,7 @@ KEY PATTERNS
 ============
 
 1. SUBPROCESS ISOLATION
-   Stegasoo's DCT mode uses scipy/jpegio which can crash on malformed input.
+   Stegasoo's DCT mode uses scipy/jpeglib which can crash on malformed input.
    We run encode/decode in subprocesses so crashes don't take down the server:
 
        subprocess_stego = SubprocessStego(timeout=180)
@@ -213,7 +213,7 @@ except ImportError:
 #
 # This is a critical reliability pattern. Here's the problem:
 #
-# scipy's DCT and jpegio can crash (segfault) on:
+# scipy's DCT and jpeglib can crash (segfault) on:
 # - Malformed JPEG files
 # - Very large images that exhaust memory
 # - Certain edge cases in coefficient manipulation
@@ -253,6 +253,7 @@ from stegasoo.qr_utils import (
     detect_and_crop_qr,
     extract_key_from_qr,
     generate_qr_code,
+    is_compressed,
 )
 
 # Initialize subprocess wrapper (worker script must be in same directory)
@@ -1116,6 +1117,13 @@ def encode_page():
         # Check if async mode requested
         is_async = request.form.get("async") == "true" or request.headers.get("X-Async") == "true"
 
+        def _error_response(msg):
+            """Return error as JSON (async) or HTML flash (sync)."""
+            if is_async:
+                return jsonify({"error": msg}), 400
+            flash(msg, "error")
+            return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
+
         try:
             # Get files
             ref_photo = request.files.get("reference_photo")
@@ -1124,12 +1132,10 @@ def encode_page():
             payload_file = request.files.get("payload_file")
 
             if not ref_photo or not carrier:
-                flash("Both reference photo and carrier image are required", "error")
+                return _error_response("Both reference photo and carrier image are required")
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             if not allowed_image(ref_photo.filename) or not allowed_image(carrier.filename):
-                flash("Invalid file type. Use PNG, JPG, or BMP", "error")
+                return _error_response("Invalid file type. Use PNG, JPG, or BMP")
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Get form data - v3.2.0: renamed from day_phrase to passphrase
             message = request.form.get("message", "")
@@ -1158,8 +1164,7 @@ def encode_page():
 
             # Check DCT availability
             if embed_mode == "dct" and not has_dct_support():
-                flash("DCT mode requires scipy. Install with: pip install scipy", "error")
+                return _error_response("DCT mode requires scipy. Install with: pip install scipy")
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Determine payload
             if payload_type == "file" and payload_file and payload_file.filename:
@@ -1168,8 +1173,7 @@ def encode_page():
 
                 result = validate_file_payload(file_data, payload_file.filename)
                 if not result.is_valid:
-                    flash(result.error_message, "error")
+                    return _error_response(result.error_message)
-                    return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
                 mime_type, _ = mimetypes.guess_type(payload_file.filename)
                 payload = FilePayload(
@@ -1179,20 +1183,17 @@ def encode_page():
                 # Text message
                 result = validate_message(message)
                 if not result.is_valid:
-                    flash(result.error_message, "error")
+                    return _error_response(result.error_message)
-                    return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
                 payload = message
 
             # v3.2.0: Renamed from day_phrase
             if not passphrase:
-                flash("Passphrase is required", "error")
+                return _error_response("Passphrase is required")
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # v3.2.0: Validate passphrase
             result = validate_passphrase(passphrase)
             if not result.is_valid:
-                flash(result.error_message, "error")
+                return _error_response(result.error_message)
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Show warning if passphrase is short
             if result.warning:
@@ -1209,8 +1210,8 @@ def encode_page():
             rsa_key_from_qr = False
 
             if rsa_key_pem:
-                # Webcam-scanned PEM key (v4.1.5) - may be compressed
+                # Webcam-scanned PEM key (v4.1.5+) - may be compressed (zlib or zstd)
-                if rsa_key_pem.startswith("STEGASOO-Z:"):
+                if is_compressed(rsa_key_pem):
                     rsa_key_pem = decompress_data(rsa_key_pem)
                 rsa_key_data = rsa_key_pem.encode("utf-8")
                 rsa_key_from_qr = True
@@ -1223,21 +1224,18 @@ def encode_page():
                     rsa_key_data = key_pem.encode("utf-8")
                     rsa_key_from_qr = True
                 else:
-                    flash("Could not extract RSA key from QR code image.", "error")
+                    return _error_response("Could not extract RSA key from QR code image.")
-                    return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Validate security factors
             result = validate_security_factors(pin, rsa_key_data)
             if not result.is_valid:
-                flash(result.error_message, "error")
+                return _error_response(result.error_message)
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Validate PIN if provided
             if pin:
                 result = validate_pin(pin)
                 if not result.is_valid:
-                    flash(result.error_message, "error")
+                    return _error_response(result.error_message)
-                    return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Determine key password
             key_password = None if rsa_key_from_qr else (rsa_password if rsa_password else None)
@@ -1246,14 +1244,12 @@ def encode_page():
             if rsa_key_data:
                 result = validate_rsa_key(rsa_key_data, key_password)
                 if not result.is_valid:
-                    flash(result.error_message, "error")
+                    return _error_response(result.error_message)
-                    return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Validate carrier image
             result = validate_image(carrier_data, "Carrier image")
             if not result.is_valid:
-                flash(result.error_message, "error")
+                return _error_response(result.error_message)
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Pre-check payload capacity BEFORE encode (fail fast)
             from stegasoo.steganography import will_fit_by_mode
@@ -1273,8 +1269,7 @@ def encode_page():
                     alt_check = will_fit_by_mode(payload_size, carrier_data, embed_mode="lsb")
                     if alt_check.get("fits"):
                         error_msg += " - Try LSB mode instead."
-                flash(error_msg, "error")
+                return _error_response(error_msg)
-                return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
             # Build encode params for either sync or async
             encode_params = {
@@ -1375,14 +1370,11 @@ def encode_page():
             return redirect(url_for("encode_result", file_id=file_id))
 
         except CapacityError as e:
-            flash(str(e), "error")
+            return _error_response(str(e))
-            return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
         except StegasooError as e:
-            flash(str(e), "error")
+            return _error_response(str(e))
-            return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
         except Exception as e:
-            flash(f"Error: {e}", "error")
+            return _error_response(f"Error: {e}")
-            return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
     return render_template("encode.html", has_qrcode_read=HAS_QRCODE_READ)
 
@@ -1657,8 +1649,8 @@ def decode_page():
             rsa_key_from_qr = False
 
             if rsa_key_pem:
-                # Webcam-scanned PEM key (v4.1.5) - may be compressed
+                # Webcam-scanned PEM key (v4.1.5+) - may be compressed (zlib or zstd)
-                if rsa_key_pem.startswith("STEGASOO-Z:"):
+                if is_compressed(rsa_key_pem):
                     rsa_key_pem = decompress_data(rsa_key_pem)
                 rsa_key_data = rsa_key_pem.encode("utf-8")
                 rsa_key_from_qr = True

frontends/web/static/js/stegasoo.js

@@ -1009,7 +1009,9 @@ const Stegasoo = {
                 const percent = progressData.percent || 0;
                 const phase = progressData.phase || 'processing';
 
-                this.updateProgress(percent, this.formatPhase(phase));
+                // Use indeterminate mode for initializing/starting phases
+                const isIndeterminate = (phase === 'initializing' || phase === 'starting');
+                this.updateProgress(percent, this.formatPhase(phase), isIndeterminate);
 
                 // Continue polling
                 setTimeout(poll, 500);
@@ -1029,7 +1031,7 @@ const Stegasoo = {
     formatPhase(phase) {
         const phases = {
             'starting': 'Starting...',
-            'initializing': 'Initializing...',
+            'initializing': 'Deriving keys (may take a moment)...',
             'embedding': 'Embedding data...',
             'saving': 'Saving image...',
             'finalizing': 'Finalizing...',
@@ -1070,8 +1072,9 @@ const Stegasoo = {
             document.body.appendChild(modal);
         }
 
-        // Reset progress
+        // Reset progress tracking and start with indeterminate state
-        this.updateProgress(0, 'Initializing...');
+        this.resetProgressTracking();
+        this.updateProgress(0, 'Initializing...', true);
 
         // Show modal
         const bsModal = new bootstrap.Modal(modal);
@@ -1090,16 +1093,47 @@ const Stegasoo = {
     },
 
     /**
-     * Update progress bar and text
+     * Track max progress to prevent backwards jumps
      */
-    updateProgress(percent, phase) {
+    _maxProgress: 0,
+
+    /**
+     * Reset progress tracking (call when starting new operation)
+     */
+    resetProgressTracking() {
+        this._maxProgress = 0;
+    },
+
+    /**
+     * Update progress bar and text
+     * Supports indeterminate mode for initializing phase (barber pole at full width)
+     */
+    updateProgress(percent, phase, indeterminate = false) {
         const progressBar = document.getElementById('progressBar');
         const progressText = document.getElementById('progressText');
         const phaseText = document.getElementById('progressPhase');
 
-        if (progressBar) progressBar.style.width = percent + '%';
+        if (indeterminate) {
-        if (progressText) progressText.textContent = Math.round(percent) + '%';
+            // Barber pole animation at full width, no percentage
+            if (progressBar) {
+                progressBar.style.width = '100%';
+                progressBar.classList.add('progress-bar-striped', 'progress-bar-animated');
+            }
+            if (progressText) progressText.textContent = '';
             if (phaseText) phaseText.textContent = phase;
+        } else {
+            // Determinate progress - never go backwards
+            const safePercent = Math.max(percent, this._maxProgress);
+            this._maxProgress = safePercent;
+
+            if (progressBar) {
+                progressBar.style.width = safePercent + '%';
+                // Keep animation but show actual progress
+                progressBar.classList.add('progress-bar-striped', 'progress-bar-animated');
+            }
+            if (progressText) progressText.textContent = Math.round(safePercent) + '%';
+            if (phaseText) phaseText.textContent = phase;
+        }
     },
 
     // ========================================================================
@@ -1187,7 +1221,9 @@ const Stegasoo = {
                 const percent = progressData.percent || 0;
                 const phase = progressData.phase || 'processing';
 
-                this.updateProgress(percent, this.formatDecodePhase(phase));
+                // Use indeterminate mode for initializing/starting/loading phases
+                const isIndeterminate = (phase === 'initializing' || phase === 'starting' || phase === 'loading');
+                this.updateProgress(percent, this.formatDecodePhase(phase), isIndeterminate);
 
                 // Continue polling
                 setTimeout(poll, 500);
@@ -1207,8 +1243,11 @@ const Stegasoo = {
     formatDecodePhase(phase) {
         const phases = {
             'starting': 'Starting...',
+            'initializing': 'Deriving keys (may take a moment)...',
+            'loading': 'Deriving keys (may take a moment)...',
             'reading': 'Reading image...',
             'extracting': 'Extracting data...',
+            'decoding': 'Decoding data...',
             'decrypting': 'Decrypting...',
             'verifying': 'Verifying...',
             'finalizing': 'Finalizing...',

frontends/web/stego_worker.py

@@ -3,7 +3,7 @@
 Stegasoo Subprocess Worker (v4.0.0)
 
 This script runs in a subprocess and handles encode/decode operations.
-If it crashes due to jpegio/scipy issues, the parent Flask process survives.
+If it crashes due to jpeglib/scipy issues, the parent Flask process survives.
 
 CHANGES in v4.0.0:
 - Added channel_key support for encode/decode operations
@@ -171,8 +171,7 @@ def decode_operation(params: dict) -> dict:
     # Resolve channel key (v4.0.0)
     resolved_channel_key = _resolve_channel_key(params.get("channel_key", "auto"))
 
-    _write_decode_progress(progress_file, 25, "extracting")
+    # Library handles progress internally via progress_file parameter
-
     # Call decode with correct parameter names
     result = decode(
         stego_image=stego_data,
@@ -183,9 +182,9 @@ def decode_operation(params: dict) -> dict:
         rsa_password=params.get("rsa_password"),
         embed_mode=params.get("embed_mode", "auto"),
         channel_key=resolved_channel_key,  # v4.0.0
+        progress_file=progress_file,  # v4.2.0: pass through for real-time progress
     )
-
+    # Library writes 100% "complete" - no need for worker to write again
-    _write_decode_progress(progress_file, 90, "finalizing")
 
     if result.is_file:
         return {

frontends/web/subprocess_stego.py

@@ -132,7 +132,7 @@ class SubprocessStego:
     """
     Subprocess-isolated steganography operations.
 
-    All operations run in a separate Python process. If jpegio or scipy
+    All operations run in a separate Python process. If jpeglib or scipy
     crashes, only the subprocess dies - Flask keeps running.
     """
 

frontends/web/templates/about.html

@@ -340,11 +340,13 @@
                 <!-- Current Version - Prominent -->
                 <div class="alert alert-success mb-4">
                     <div class="d-flex align-items-center">
-                        <span class="badge bg-success fs-6 me-3">v4.1.2</span>
+                        <span class="badge bg-success fs-6 me-3">v4.2.0</span>
                         <div>
-                            <strong>Progress bars</strong> for encode operations,
+                            <strong>Performance optimizations:</strong>
-                            <strong>mobile-responsive polish</strong>,
+                            ~70% faster decode (vectorized DCT),
-                            DCT decode bug fix, release validation script
+                            50% less RAM (float32),
+                            async API endpoints,
+                            decode progress callbacks
                         </div>
                     </div>
                 </div>
@@ -362,6 +364,10 @@
                             <div class="accordion-body p-0">
                                 <table class="table table-dark table-sm small mb-0">
                                     <tbody>
+                                        <tr>
+                                            <td width="80"><strong>4.1.7</strong></td>
+                                            <td>Progress bars for encode, mobile polish, release validation</td>
+                                        </tr>
                                         <tr>
                                             <td width="80"><strong>4.1.1</strong></td>
                                             <td>DCT RS format stability, Docker cleanup, first-boot wizard</td>
@@ -559,7 +565,7 @@
                                         </tr>
                                         <tr>
                                             <td><i class="bi bi-clock me-2"></i>File expiry</td>
-                                            <td><strong>5 min</strong></td>
+                                            <td><strong>10 min</strong></td>
                                         </tr>
                                         <tr>
                                             <td><i class="bi bi-key me-2"></i>PIN</td>
@@ -567,7 +573,7 @@
                                         </tr>
                                         <tr>
                                             <td><i class="bi bi-shield me-2"></i>RSA keys</td>
-                                            <td><strong>2048, 3072, 4096 bit</strong></td>
+                                            <td><strong>2048, 3072 bit</strong></td>
                                         </tr>
                                         <tr>
                                             <td><i class="bi bi-chat-quote me-2"></i>Passphrase</td>

frontends/web/templates/decode.html

@@ -158,7 +158,7 @@
 
                 <div class="alert alert-warning small">
                     <i class="bi bi-clock me-1"></i>
-                    <strong>File expires in 5 minutes.</strong> Download now.
+                    <strong>File expires in 10 minutes.</strong> Download now.
                 </div>
 
                 <a href="/decode" class="btn btn-outline-light w-100">

frontends/web/templates/encode_result.html

@@ -128,7 +128,7 @@
                     <i class="bi bi-exclamation-triangle me-1"></i>
                     <strong>Important:</strong>
                     <ul class="mb-0 mt-2">
-                        <li>This file expires in <strong>5 minutes</strong></li>
+                        <li>This file expires in <strong>10 minutes</strong></li>
                         <li>Do <strong>not</strong> resize or recompress the image</li>
                         {% if embed_mode == 'dct' and output_format == 'jpeg' %}
                         <li>JPEG format is lossy - avoid re-saving or editing</li>

frontends/web/templates/generate.html

@@ -65,11 +65,7 @@
                                 <select name="rsa_bits" class="form-select form-select-sm" id="rsaBitsSelect">
                                     <option value="2048" selected>2048 bits (~128 bits entropy)</option>
                                     <option value="3072">3072 bits (~128 bits entropy)</option>
-                                    <option value="4096">4096 bits (~128 bits entropy)</option>
                                 </select>
-                                <div class="form-text text-warning d-none" id="rsaQrWarning">
-                                    <i class="bi bi-exclamation-triangle me-1"></i>QR code unavailable for keys &gt;3072 bits
-                                </div>
                             </div>
                         </div>
                     </div>
@@ -286,12 +282,6 @@
                                     <i class="bi bi-shield-exclamation me-1"></i>
                                     <strong>Security note:</strong> The QR code contains your unencrypted private key.
                                     Only scan in a secure environment. Consider using the password-protected download instead.
-                                    {% if rsa_bits >= 4096 %}
-                                    <br><br>
-                                    <i class="bi bi-exclamation-triangle me-1"></i>
-                                    <strong>4096-bit keys</strong> produce very dense QR codes. If scanning fails, 
-                                    use the PEM text or download options instead.
-                                    {% endif %}
                                 </div>
                             </div>
                         </div>

pyproject.toml

@@ -4,11 +4,11 @@ build-backend = "hatchling.build"
 
 [project]
 name = "stegasoo"
-version = "4.1.5"
+version = "4.2.0"
 description = "Secure steganography with hybrid photo + passphrase + PIN authentication"
 readme = "README.md"
 license = "MIT"
-requires-python = ">=3.10"
+requires-python = ">=3.11"
 authors = [
     { name = "Aaron D. Lee" }
 ]
@@ -29,9 +29,10 @@ classifiers = [
     "License :: OSI Approved :: MIT License",
     "Operating System :: OS Independent",
     "Programming Language :: Python :: 3",
-    "Programming Language :: Python :: 3.10",
     "Programming Language :: Python :: 3.11",
     "Programming Language :: Python :: 3.12",
+    "Programming Language :: Python :: 3.13",
+    "Programming Language :: Python :: 3.14",
     "Topic :: Security :: Cryptography",
     "Topic :: Multimedia :: Graphics",
 ]
@@ -40,6 +41,7 @@ dependencies = [
     "pillow>=10.0.0",
     "cryptography>=41.0.0",
     "argon2-cffi>=23.0.0",
+    "zstandard>=0.22.0",  # v4.2.0: Default compression algorithm
 ]
 
 [project.optional-dependencies]
@@ -47,7 +49,7 @@ dependencies = [
 dct = [
     "numpy>=2.0.0",
     "scipy>=1.10.0",
-    "jpegio>=0.2.0",
+    "jpeglib>=1.0.0",
     "reedsolo>=1.7.0",
 ]
 cli = [
@@ -57,7 +59,7 @@ cli = [
     "rich>=13.0.0",
 ]
 compression = [
-    "lz4>=4.0.0",
+    "lz4>=4.0.0",  # Optional: faster but slightly worse ratio than zstd
 ]
 web = [
     "flask>=3.0.0",
@@ -68,7 +70,7 @@ web = [
     # Include DCT support for web UI
     "numpy>=2.0.0",
     "scipy>=1.10.0",
-    "jpegio>=0.2.0",
+    "jpeglib>=1.0.0",
     "reedsolo>=1.7.0",
 ]
 api = [
@@ -80,7 +82,7 @@ api = [
     # Include DCT support for API
     "numpy>=2.0.0",
     "scipy>=1.10.0",
-    "jpegio>=0.2.0",
+    "jpeglib>=1.0.0",
     "reedsolo>=1.7.0",
 ]
 all = [
@@ -110,7 +112,7 @@ include = [
 ]
 
 [tool.hatch.build.targets.wheel]
-packages = ["src/stegasoo"]
+packages = ["src/stegasoo", "frontends"]
 
 [tool.pytest.ini_options]
 testpaths = ["tests"]
@@ -119,7 +121,7 @@ addopts = "-v --cov=stegasoo --cov-report=term-missing"
 
 [tool.black]
 line-length = 100
-target-version = ["py310", "py311", "py312"]
+target-version = ["py311", "py312", "py313"]
 
 [tool.ruff]
 line-length = 100
@@ -136,7 +138,7 @@ ignore = ["E501"]
 "src/stegasoo/__init__.py" = ["E402"]
 
 [tool.mypy]
-python_version = "3.10"
+python_version = "3.11"
 warn_return_any = true
 warn_unused_configs = true
 ignore_missing_imports = true

rpi/BUILD_IMAGE.md

@@ -26,51 +26,50 @@ ssh admin@stegasoo.local
 ## Step 3: Pre-Setup
 
 ```bash
-# Take ownership of /opt (for pyenv, jpegio builds)
+# Take ownership of /opt
 sudo chown admin:admin /opt
 
-# Install git and zstd (not included in Lite image)
+# Install git (not included in Lite image)
-sudo apt-get update && sudo apt-get install -y git zstd jq
+sudo apt-get update && sudo apt-get install -y git
 ```
 
 ## Step 4: Clone Repo
 
 ```bash
 cd /opt
-git clone -b 4.1 https://github.com/adlee-was-taken/stegasoo.git stegasoo
+git clone -b 4.2 https://github.com/adlee-was-taken/stegasoo.git stegasoo
 ```
 
-## Step 5: Copy Pre-built Tarball (from host)
+## Step 5: Run Setup
-
-> **Dev-only asset:** This tarball is for building Pi images, not for end users.
-> It's available on [Releases](https://github.com/adlee-was-taken/stegasoo/releases) for image builders.
-
-```bash
-# On your host machine:
-scp rpi/stegasoo-rpi-runtime-env-arm64.tar.zst admin@stegasoo.local:/opt/stegasoo/rpi/
-```
-
-This tarball contains:
-- pyenv with Python 3.12 (pre-compiled for ARM64)
-- venv with all dependencies (jpegio, scipy, etc.)
-
-Install time: **~2 minutes** (vs 20+ min from source)
-
-## Step 6: Run Setup
 
 ```bash
 cd /opt/stegasoo
-./rpi/setup.sh  # Detects local tarball, skips download
+./rpi/setup.sh
 ```
 
-### From-Source Build (optional)
+The setup script:
+- Verifies Python 3.11+ (system Python, no pyenv needed)
+- Installs dependencies via apt and pip
+- jpeglib installs cleanly (no ARM patching like jpegio)
+- Creates and enables systemd service
+
+Install time: **5-10 minutes** (from source)
+
+### Pre-built Venv (optional)
+
+For faster installs, you can provide a pre-built venv tarball:
 
-To build without the pre-built tarball:
 ```bash
-./rpi/setup.sh --no-prebuilt  # Takes 15-20 minutes
+# On your host machine:
+scp rpi/stegasoo-rpi-venv-arm64.tar.zst admin@stegasoo.local:/opt/stegasoo/rpi/
+
+# Then on Pi:
+cd /opt/stegasoo && ./rpi/setup.sh  # Detects local tarball, skips pip build
 ```
 
-## Step 7: Test It Works
+Install time with pre-built: **~2 minutes**
+
+## Step 6: Test It Works
 
 ```bash
 sudo systemctl start stegasoo
@@ -78,7 +77,7 @@ curl -k https://localhost:5000
 # Should return HTML
 ```
 
-## Step 8: Sanitize for Distribution
+## Step 7: Sanitize for Distribution
 
 ```bash
 # Full sanitize (for final image - removes WiFi, shuts down)
@@ -98,7 +97,7 @@ This removes:
 
 The script validates all cleanup steps before finishing.
 
-## Step 9: Pull the Image
+## Step 8: Pull the Image
 
 Remove SD card, insert into your Linux machine:
 
@@ -107,12 +106,12 @@ Remove SD card, insert into your Linux machine:
 lsblk
 
 # Pull image (auto-resizes to 16GB, compresses with zstd)
-sudo ./rpi/pull-image.sh /dev/sdX stegasoo-rpi-4.1.5.img.zst
+sudo ./rpi/pull-image.sh /dev/sdX stegasoo-rpi-4.2.0.img.zst
 ```
 
 The script automatically resizes rootfs to 16GB (for smaller download), preserves auto-expand, and compresses.
 
-## Step 10: Distribute
+## Step 9: Distribute
 
 Upload `.img.zst` to GitHub Releases.
 
@@ -130,36 +129,31 @@ zstdcat stegasoo-rpi-*.img.zst | sudo dd of=/dev/sdX bs=4M status=progress
 
 ---
 
-## Creating the Pre-built Tarball
+## Creating the Pre-built Venv Tarball
 
 After a successful from-source build, create the pre-built tarball for future installs:
 
 ```bash
 # On the Pi after successful setup:
-cd ~
+cd /opt/stegasoo
 
-# Strip caches and tests from venv (295MB → 208MB)
+# Strip caches and tests from venv (saves ~100MB)
-find /opt/stegasoo/venv/ -type d -name '__pycache__' -exec rm -rf {} + 2>/dev/null
+find venv/ -type d -name '__pycache__' -exec rm -rf {} + 2>/dev/null
-find /opt/stegasoo/venv/ -type d -name 'tests' -exec rm -rf {} + 2>/dev/null
+find venv/ -type d -name 'tests' -exec rm -rf {} + 2>/dev/null
-find /opt/stegasoo/venv/ -type d -name 'test' -exec rm -rf {} + 2>/dev/null
+find venv/ -type d -name 'test' -exec rm -rf {} + 2>/dev/null
 
 # Create venv tarball
-cd /opt/stegasoo
+tar -cf - venv/ | zstd -19 -T0 > /tmp/stegasoo-rpi-venv-arm64.tar.zst
-tar -cf - venv/ | zstd -19 -T0 > ~/stegasoo-venv.tar.zst
 
-# Create combined tarball (pyenv + venv pointer)
+# Check size (should be ~40-50MB)
-cd ~
+ls -lh /tmp/stegasoo-rpi-venv-arm64.tar.zst
-tar -cf - .pyenv stegasoo-venv.tar.zst | zstd -19 -T0 > /tmp/stegasoo-rpi-runtime-env-arm64.tar.zst
-
-# Check size (should be ~50-60MB)
-ls -lh /tmp/stegasoo-rpi-runtime-env-arm64.tar.zst
 ```
 
 Pull to host and upload to GitHub releases:
 ```bash
 # On host:
-scp admin@stegasoo.local:/tmp/stegasoo-rpi-runtime-env-arm64.tar.zst ./
+scp admin@stegasoo.local:/tmp/stegasoo-rpi-venv-arm64.tar.zst ./rpi/
-# Upload to GitHub releases as stegasoo-rpi-runtime-env-arm64.tar.zst
+# Upload to GitHub releases as stegasoo-rpi-venv-arm64.tar.zst
 ```
 
 ---
@@ -169,18 +163,15 @@ scp admin@stegasoo.local:/tmp/stegasoo-rpi-runtime-env-arm64.tar.zst ./
 ```bash
 # On Pi (after SSH):
 sudo chown admin:admin /opt
-sudo apt-get update && sudo apt-get install -y git zstd jq
+sudo apt-get update && sudo apt-get install -y git
-cd /opt && git clone -b 4.1 https://github.com/adlee-was-taken/stegasoo.git stegasoo
+cd /opt && git clone -b 4.2 https://github.com/adlee-was-taken/stegasoo.git stegasoo
 
-# On host (copy tarball):
+# Run setup:
-scp rpi/stegasoo-rpi-runtime-env-arm64.tar.zst admin@stegasoo.local:/opt/stegasoo/rpi/
-
-# On Pi (run setup):
 cd /opt/stegasoo && ./rpi/setup.sh
 sudo systemctl start stegasoo
 curl -k https://localhost:5000
 sudo /opt/stegasoo/rpi/sanitize-for-image.sh
 
 # On host (pull image - auto-resizes to 16GB):
-sudo ./rpi/pull-image.sh /dev/sdX stegasoo-rpi-4.1.5.img.zst
+sudo ./rpi/pull-image.sh /dev/sdX stegasoo-rpi-4.2.0.img.zst
 ```

rpi/README.md

@@ -4,7 +4,7 @@ Scripts and resources for deploying Stegasoo on Raspberry Pi.
 
 ## Quick Install
 
-On a fresh Raspberry Pi OS Lite (64-bit) installation:
+On a fresh Raspberry Pi OS (64-bit) installation:
 
 ```bash
 # Pre-setup (git not included in Lite image)
@@ -13,16 +13,16 @@ sudo apt-get update && sudo apt-get install -y git
 
 # Clone and run setup
 cd /opt
-git clone -b 4.1 https://github.com/adlee-was-taken/stegasoo.git stegasoo
+git clone -b 4.2 https://github.com/adlee-was-taken/stegasoo.git stegasoo
 cd stegasoo
 ./rpi/setup.sh
 ```
 
 ## What the Setup Script Does
 
-1. **Installs system dependencies** - build tools, libraries
+1. **Verifies Python 3.11+** - uses system Python (no pyenv needed)
-2. **Installs Python 3.12** - via pyenv (Pi OS ships with 3.13 which is incompatible)
+2. **Installs system dependencies** - build tools, libraries
-3. **Builds jpegio for ARM** - patches x86-specific flags
+3. **Installs jpeglib** - DCT steganography (Python 3.11-3.14 compatible)
 4. **Installs Stegasoo** - with web UI and all dependencies
 5. **Creates systemd service** - auto-starts on boot
 6. **Enables the service** - ready to start
@@ -30,11 +30,18 @@ cd stegasoo
 ## Requirements
 
 - Raspberry Pi 4 or 5
-- Raspberry Pi OS Lite (64-bit) - Bookworm or later
+- Raspberry Pi OS (64-bit) - Bookworm (Python 3.11) or Trixie (Python 3.13)
 - 4GB+ RAM recommended (2GB minimum)
 - 16GB+ SD card (pre-built images are 16GB)
 - Internet connection
 
+### Python Compatibility
+
+| Raspberry Pi OS | Python | Supported |
+|-----------------|--------|-----------|
+| Bookworm        | 3.11   | Yes       |
+| Trixie          | 3.13   | Yes       |
+
 ### Performance
 
 On a Pi 4 at 2GHz with USB 3.0 NVMe, expect ~60 seconds to encode/decode a 10MB JPEG with full encryption (passphrase + PIN + reference photo).
@@ -159,7 +166,7 @@ sudo apt-get update && sudo apt-get install -y git
 
 # Clone and run setup
 cd /opt
-git clone -b 4.1 https://github.com/adlee-was-taken/stegasoo.git stegasoo
+git clone -b 4.2 https://github.com/adlee-was-taken/stegasoo.git stegasoo
 cd stegasoo
 ./rpi/setup.sh
 ```
@@ -200,7 +207,7 @@ After Pi shuts down, remove SD card and on another Linux machine:
 lsblk
 
 # Pull image (auto-resizes to 16GB, compresses with zstd)
-sudo ./rpi/pull-image.sh /dev/sdX stegasoo-rpi-4.1.5.img.zst
+sudo ./rpi/pull-image.sh /dev/sdX stegasoo-rpi-4.2.0.img.zst
 ```
 
 The `pull-image.sh` script automatically:

rpi/build-runtime-tarball.sh

@@ -1,10 +1,10 @@
 #!/bin/bash
 #
-# Build Stegasoo Pi Runtime Environment Tarball
+# Build Stegasoo Pi venv Tarball
 # Run this ON THE PI after a successful from-source build
 #
-# Creates: stegasoo-rpi-runtime-env-arm64.tar.zst (~50-60MB)
+# Creates: stegasoo-rpi-venv-arm64.tar.zst (~40-50MB)
-# Contains: pyenv + Python 3.12 + venv with all dependencies
+# Contains: venv with all dependencies (uses system Python 3.11+)
 #
 
 set -e
@@ -16,11 +16,10 @@ YELLOW='\033[1;33m'
 NC='\033[0m'
 
 INSTALL_DIR="${INSTALL_DIR:-/opt/stegasoo}"
-OUTPUT_DIR="${OUTPUT_DIR:-/tmp}"
+OUTPUT_FILE="${1:-$HOME/stegasoo-rpi-venv-arm64.tar.zst}"
-OUTPUT_FILE="$OUTPUT_DIR/stegasoo-rpi-runtime-env-arm64.tar.zst"
 
 echo -e "${GREEN}╔═══════════════════════════════════════════════════════════════╗${NC}"
-echo -e "${GREEN}║         Stegasoo Pi Runtime Tarball Builder                   ║${NC}"
+echo -e "${GREEN}║         Stegasoo Pi venv Tarball Builder                      ║${NC}"
 echo -e "${GREEN}╚═══════════════════════════════════════════════════════════════╝${NC}"
 echo ""
 
@@ -32,13 +31,6 @@ if [[ "$ARCH" != "aarch64" ]]; then
     exit 1
 fi
 
-# Verify pyenv exists
-if [[ ! -d "$HOME/.pyenv" ]]; then
-    echo -e "${RED}Error: pyenv not found at ~/.pyenv${NC}"
-    echo "Run a from-source build first: ./rpi/setup.sh --no-prebuilt"
-    exit 1
-fi
-
 # Verify venv exists
 if [[ ! -d "$INSTALL_DIR/venv" ]]; then
     echo -e "${RED}Error: venv not found at $INSTALL_DIR/venv${NC}"
@@ -47,33 +39,22 @@ if [[ ! -d "$INSTALL_DIR/venv" ]]; then
 fi
 
 # Step 1: Clean caches from venv
-echo -e "${GREEN}[1/4]${NC} Cleaning caches from venv..."
+echo -e "${GREEN}[1/2]${NC} Cleaning caches from venv..."
 VENV_SIZE_BEFORE=$(du -sh "$INSTALL_DIR/venv" | cut -f1)
 find "$INSTALL_DIR/venv/" -type d -name '__pycache__' -exec rm -rf {} + 2>/dev/null || true
 find "$INSTALL_DIR/venv/" -type d -name 'tests' -exec rm -rf {} + 2>/dev/null || true
 find "$INSTALL_DIR/venv/" -type d -name 'test' -exec rm -rf {} + 2>/dev/null || true
 find "$INSTALL_DIR/venv/" -type f -name '*.pyc' -delete 2>/dev/null || true
 VENV_SIZE_AFTER=$(du -sh "$INSTALL_DIR/venv" | cut -f1)
-echo "  venv: $VENV_SIZE_BEFORE → $VENV_SIZE_AFTER"
+echo "  venv: $VENV_SIZE_BEFORE -> $VENV_SIZE_AFTER"
 
-# Step 2: Create venv tarball
+# Step 2: Create tarball
-echo -e "${GREEN}[2/4]${NC} Creating venv tarball..."
+echo -e "${GREEN}[2/2]${NC} Creating tarball..."
 cd "$INSTALL_DIR"
-tar -cf - venv/ | zstd -19 -T0 > "$HOME/stegasoo-venv.tar.zst"
+tar -cf - venv/ | zstd -19 -T0 > "$OUTPUT_FILE"
-VENV_TAR_SIZE=$(ls -lh "$HOME/stegasoo-venv.tar.zst" | awk '{print $5}')
-echo "  Created: ~/stegasoo-venv.tar.zst ($VENV_TAR_SIZE)"
 
-# Step 3: Create combined tarball
+# Summary
-echo -e "${GREEN}[3/4]${NC} Creating combined runtime tarball..."
-cd "$HOME"
-tar -cf - .pyenv stegasoo-venv.tar.zst | zstd -19 -T0 > "$OUTPUT_FILE"
-
-# Cleanup intermediate file
-rm "$HOME/stegasoo-venv.tar.zst"
-
-# Step 4: Summary
 FINAL_SIZE=$(ls -lh "$OUTPUT_FILE" | awk '{print $5}')
-echo -e "${GREEN}[4/4]${NC} Done!"
 echo ""
 echo -e "${GREEN}════════════════════════════════════════════════════════════════${NC}"
 echo -e "  Output: ${YELLOW}$OUTPUT_FILE${NC}"
@@ -83,7 +64,7 @@ echo ""
 echo "To pull to your host machine:"
 echo "  scp $(whoami)@$(hostname).local:$OUTPUT_FILE ./"
 echo ""
-echo "To use in setup.sh, copy to:"
+echo "To use in setup.sh, place at:"
-echo "  rpi/stegasoo-rpi-runtime-env-arm64.tar.zst"
+echo "  rpi/stegasoo-rpi-venv-arm64.tar.zst"
 echo ""
 echo "Or upload to GitHub releases for automatic download."

rpi/flash-image.sh

@@ -80,9 +80,9 @@ if [ -z "$1" ]; then
     echo "Supported formats: .img, .img.zst, .img.xz, .img.gz, .img.zst.zip"
     echo ""
     echo "Examples:"
-    echo "  $0 stegasoo-rpi-4.1.1.img.zst             # auto-detect SD card"
+    echo "  $0 stegasoo-rpi-4.2.0.img.zst             # auto-detect SD card"
-    echo "  $0 stegasoo-rpi-4.1.1.img.zst.zip         # from GitHub release"
+    echo "  $0 stegasoo-rpi-4.2.0.img.zst.zip         # from GitHub release"
-    echo "  $0 stegasoo-rpi-4.1.1.img.zst /dev/sdb    # specify device"
+    echo "  $0 stegasoo-rpi-4.2.0.img.zst /dev/sdb    # specify device"
     exit 1
 fi
 

rpi/patches/README.md

@@ -2,56 +2,39 @@
 
 This directory contains patches for dependencies that need modifications to build on ARM64.
 
+## Current Status (v4.2+)
+
+As of Stegasoo 4.2, we use **jpeglib** instead of jpegio. The jpeglib build process is handled inline in `setup.sh` and includes:
+
+- Cloning from GitHub (PyPI tarball missing headers)
+- Downloading libjpeg headers for each version (6b through 9f)
+- Patching setup.py to skip turbo/mozjpeg (need cmake-generated headers)
+
+See `setup.sh` for the full implementation.
+
+## Legacy: jpegio Patches (v4.1 and earlier)
+
+The `jpegio/` directory contains patches for the old jpegio dependency, which required removing x86-specific `-m64` compiler flags. These are no longer used but kept for reference.
+
+## jpeglib Helper Script
+
+The `jpeglib/install-jpeglib-arm64.sh` script is a standalone version of the jpeglib build process. It's not used by setup.sh (which has the logic inline) but can be useful for manual testing or debugging.
+
 ## Structure
 
 ```
 patches/
-  <package>/
+  jpegio/           # Legacy (v4.1) - not used in v4.2+
-    arm64.patch       # Standard unified diff patch file
+    arm64.patch
-    apply-patch.sh    # Script with fallback strategies
+    apply-patch.sh
+  jpeglib/          # Reference script for manual builds
+    install-jpeglib-arm64.sh
 ```
 
-## How It Works
+## Adding New Patches
 
-The `apply-patch.sh` script tries multiple strategies in order:
+If a new dependency needs ARM64 patches:
-
-1. **Patch file** - Apply the `.patch` file using `patch -p1`
-2. **Sed fallback** - Use sed for simple string replacements
-3. **Python fallback** - Use regex for flexible pattern matching
-
-This layered approach handles:
-- Exact matches (patch file works)
-- Minor upstream changes (sed catches variations)
-- Significant changes (Python regex is most flexible)
-- Already patched files (detected and skipped)
-
-## Adding a New Patch
 
 1. Create a directory: `patches/<package>/`
-2. Create the patch file: `git diff > arm64.patch`
+2. Add patch files or helper scripts
-3. Create `apply-patch.sh` with appropriate fallback logic
+3. Update `setup.sh` to apply the patch during installation
-4. Update `setup.sh` to call the patch script
-
-## jpegio Patch
-
-The jpegio library includes x86-specific `-m64` compiler flags that fail on ARM64.
-The patch removes these flags by replacing:
-
-```python
-cargs.append('-m64')
-```
-
-with:
-
-```python
-pass  # ARM64: removed x86-specific -m64 flag
-```
-
-## Updating Patches
-
-When upstream changes break a patch:
-
-1. Clone the new version
-2. Make the necessary modifications
-3. Generate a new patch: `diff -u original modified > arm64.patch`
-4. Test on a fresh Pi install

rpi/patches/jpeglib/install-jpeglib-arm64.sh

@@ -0,0 +1,57 @@
+#!/bin/bash
+#
+# Install jpeglib on ARM64 Linux (Raspberry Pi)
+# Works around missing headers in the source tarball
+#
+# Usage: ./install-jpeglib-arm64.sh
+#
+
+set -e
+
+echo "Installing jpeglib for ARM64..."
+
+# Create temp directory
+WORKDIR=$(mktemp -d)
+cd "$WORKDIR"
+
+# Download jpeglib source
+echo "  Downloading jpeglib source..."
+pip download jpeglib==1.0.2 --no-binary :all: --no-deps -d . -q
+tar -xzf jpeglib-1.0.2.tar.gz
+cd jpeglib-1.0.2
+
+# Download official libjpeg sources and copy headers
+echo "  Downloading libjpeg headers..."
+CJPEGLIB="src/jpeglib/cjpeglib"
+
+# libjpeg 6b
+curl -sL "https://www.ijg.org/files/jpegsrc.v6b.tar.gz" | tar -xzf -
+cp jpeg-6b/*.h "$CJPEGLIB/6b/"
+
+# libjpeg 7-9f (all use similar headers from 9e)
+curl -sL "https://www.ijg.org/files/jpegsrc.v9f.tar.gz" | tar -xzf -
+for v in 7 8 8a 8b 8c 8d 9 9a 9b 9c 9d 9e 9f; do
+    cp jpeg-9f/*.h "$CJPEGLIB/$v/"
+done
+
+# libjpeg-turbo versions
+curl -sL "https://github.com/libjpeg-turbo/libjpeg-turbo/archive/refs/tags/2.1.0.tar.gz" | tar -xzf -
+for v in turbo120 turbo130 turbo140 turbo150 turbo200 turbo210; do
+    cp libjpeg-turbo-2.1.0/*.h "$CJPEGLIB/$v/" 2>/dev/null || true
+done
+
+# mozjpeg versions
+curl -sL "https://github.com/mozilla/mozjpeg/archive/refs/tags/v4.0.3.tar.gz" | tar -xzf -
+for v in mozjpeg101 mozjpeg201 mozjpeg300 mozjpeg403; do
+    cp mozjpeg-4.0.3/*.h "$CJPEGLIB/$v/" 2>/dev/null || true
+done
+
+# Build and install
+echo "  Building jpeglib..."
+pip install . -q
+
+# Cleanup
+cd /
+rm -rf "$WORKDIR"
+
+echo "  Done! jpeglib installed successfully."

rpi/pull-image.sh

@@ -3,7 +3,7 @@
 # Resizes rootfs to 16GB for consistent image size, then pulls
 #
 # Usage: ./pull-image.sh <device> <output.img.zst>
-# Example: ./pull-image.sh /dev/sdb stegasoo-rpi-4.1.5.img.zst
+# Example: ./pull-image.sh /dev/sdb stegasoo-rpi-4.2.0.img.zst
 
 set -e
 
@@ -15,7 +15,7 @@ NC='\033[0m'
 
 if [ $# -ne 2 ]; then
     echo "Usage: $0 <device> <output.img.zst>"
-    echo "Example: $0 /dev/sdb stegasoo-rpi-4.1.5.img.zst"
+    echo "Example: $0 /dev/sdb stegasoo-rpi-4.2.0.img.zst"
     exit 1
 fi
 

rpi/remote-build-pi.sh

@@ -105,7 +105,7 @@ echo ""
 echo -e "${GREEN}[4/6]${NC} Copying pre-built tarball to Pi..."
 echo ""
 
-TARBALL="$SCRIPT_DIR/stegasoo-rpi-runtime-env-arm64.tar.zst"
+TARBALL="$SCRIPT_DIR/stegasoo-rpi-venv-arm64.tar.zst"
 if [[ -f "$TARBALL" ]]; then
     scp_to_pi "$TARBALL" "/opt/stegasoo/rpi/"
     echo -e "  ${GREEN}✓${NC} Tarball copied"

rpi/sanitize-for-image.sh

@@ -264,49 +264,25 @@ if [ -n "$STEGASOO_DIR" ] && [ -d "$STEGASOO_DIR/venv" ]; then
         echo "  Venv broken or stegasoo not installed, rebuilding..."
         rm -rf "$STEGASOO_DIR/venv"
 
-        # Find Python 3.12 (prefer pyenv, fall back to system)
+        # Find system Python 3.11+ (no pyenv needed)
-        USER_HOME=$(eval echo "~$STEGASOO_USER")
+        PYTHON_BIN=""
-        PYENV_PYTHON="$USER_HOME/.pyenv/versions/3.12*/bin/python"
+        for py in python3.14 python3.13 python3.12 python3.11 python3; do
-        if compgen -G "$PYENV_PYTHON" > /dev/null 2>&1; then
+            if command -v "$py" &>/dev/null; then
-            PYTHON_BIN=$(ls $PYENV_PYTHON 2>/dev/null | head -1)
+                PYTHON_BIN=$(command -v "$py")
-            echo "  Using pyenv Python: $PYTHON_BIN"
+                break
-        elif command -v python3.12 &>/dev/null; then
-            PYTHON_BIN="python3.12"
-            echo "  Using system Python 3.12"
-        else
-            PYTHON_BIN="python3"
-            echo "  Warning: Python 3.12 not found, using $($PYTHON_BIN --version)"
             fi
+        done
 
+        if [ -z "$PYTHON_BIN" ]; then
+            echo "  Error: Python 3.11+ not found"
+            VALIDATION_ERRORS=$((VALIDATION_ERRORS + 1))
+        else
+            echo "  Using: $PYTHON_BIN ($($PYTHON_BIN --version 2>&1))"
             sudo -u "$STEGASOO_USER" "$PYTHON_BIN" -m venv "$STEGASOO_DIR/venv"
             sudo -u "$STEGASOO_USER" "$STEGASOO_DIR/venv/bin/pip" install --quiet --upgrade pip setuptools wheel
-
-        # On ARM64, jpegio needs patching before install
-        ARCH=$(uname -m)
-        if [[ "$ARCH" == "aarch64" || "$ARCH" == "arm64" ]]; then
-            echo "  Building jpegio for ARM64 (this may take a minute)..."
-            # Install build deps
-            sudo -u "$STEGASOO_USER" "$STEGASOO_DIR/venv/bin/pip" install --quiet cython numpy
-            JPEGIO_DIR="/tmp/jpegio-build-$$"
-            rm -rf "$JPEGIO_DIR"
-            if git clone https://github.com/dwgoon/jpegio.git "$JPEGIO_DIR" 2>/dev/null; then
-                # Apply patch to remove -m64 flag
-                if [ -f "$STEGASOO_DIR/rpi/patches/jpegio/apply-patch.sh" ]; then
-                    bash "$STEGASOO_DIR/rpi/patches/jpegio/apply-patch.sh" "$JPEGIO_DIR"
-                else
-                    sed -i "s/cargs.append('-m64')/pass  # ARM64 fix/g" "$JPEGIO_DIR/setup.py"
-                fi
-                # Change ownership so user can build
-                chown -R "$STEGASOO_USER:$STEGASOO_USER" "$JPEGIO_DIR"
-                sudo -u "$STEGASOO_USER" "$STEGASOO_DIR/venv/bin/pip" install "$JPEGIO_DIR"
-                rm -rf "$JPEGIO_DIR"
-            else
-                echo "  Warning: Failed to clone jpegio, DCT mode may not work"
-            fi
-        fi
-
             sudo -u "$STEGASOO_USER" "$STEGASOO_DIR/venv/bin/pip" install --quiet -e "$STEGASOO_DIR[web]"
             echo "  Venv rebuilt and stegasoo installed"
+        fi
     else
         echo "  Venv OK"
     fi

rpi/setup.sh

@@ -4,14 +4,14 @@
 # Tested on: Raspberry Pi 4/5 with Raspberry Pi OS (64-bit)
 #
 # Usage:
-#   curl -sSL https://raw.githubusercontent.com/adlee-was-taken/stegasoo/4.1/rpi/setup.sh | bash
+#   curl -sSL https://raw.githubusercontent.com/adlee-was-taken/stegasoo/4.2/rpi/setup.sh | bash
 #   # or
-#   wget -qO- https://raw.githubusercontent.com/adlee-was-taken/stegasoo/4.1/rpi/setup.sh | bash
+#   wget -qO- https://raw.githubusercontent.com/adlee-was-taken/stegasoo/4.2/rpi/setup.sh | bash
 #
 # What this script does:
 #   1. Installs system dependencies
-#   2. Installs Python 3.12 via pyenv (Pi OS ships with 3.13 which is incompatible)
+#   2. Verifies Python 3.11+ (uses system Python)
-#   3. Patches and builds jpegio for ARM
+#   3. Installs jpeglib for DCT steganography (Python 3.11-3.14 compatible)
 #   4. Installs Stegasoo with web UI
 #   5. Creates systemd service for auto-start
 #   6. Enables the service
@@ -75,9 +75,8 @@ show_help() {
     echo ""
     echo "  Available variables:"
     echo "    INSTALL_DIR       Install location (default: /opt/stegasoo)"
-    echo "    PYTHON_VERSION    Python version (default: 3.12)"
     echo "    STEGASOO_REPO     Git repo URL"
-    echo "    STEGASOO_BRANCH   Git branch (default: 4.1)"
+    echo "    STEGASOO_BRANCH   Git branch (default: 4.2)"
     echo ""
     echo "  Example:"
     echo "    export INSTALL_DIR=\"/home/pi/stegasoo\""
@@ -95,10 +94,8 @@ done
 
 # Default configuration
 INSTALL_DIR="${INSTALL_DIR:-/opt/stegasoo}"
-PYTHON_VERSION="${PYTHON_VERSION:-3.12}"
 STEGASOO_REPO="${STEGASOO_REPO:-https://github.com/adlee-was-taken/stegasoo.git}"
-STEGASOO_BRANCH="${STEGASOO_BRANCH:-4.1}"
+STEGASOO_BRANCH="${STEGASOO_BRANCH:-4.2}"
-JPEGIO_REPO="https://github.com/dwgoon/jpegio.git"
 
 # Load config files (system, then user - user overrides system)
 for config_file in "/etc/stegasoo.conf" "$HOME/.config/stegasoo/stegasoo.conf"; do
@@ -112,7 +109,7 @@ clear
 print_banner "Raspberry Pi Setup"
 echo ""
 echo "  This will install Stegasoo with full DCT support"
-echo "  Estimated time: ~2 minutes (pre-built) or 15-20 min (from source)"
+echo "  Estimated time: ~2 minutes (pre-built) or 5-10 min (from source)"
 echo ""
 
 # Check if running on ARM
@@ -123,6 +120,63 @@ if [[ "$ARCH" != "aarch64" && "$ARCH" != "arm64" ]]; then
     exit 1
 fi
 
+# =============================================================================
+# Python Version Check
+# =============================================================================
+echo -e "${GREEN}Checking Python version...${NC}"
+
+# Find system Python
+SYSTEM_PYTHON=""
+for py in python3.14 python3.13 python3.12 python3.11 python3; do
+    if command -v "$py" &>/dev/null; then
+        SYSTEM_PYTHON=$(command -v "$py")
+        break
+    fi
+done
+
+if [ -z "$SYSTEM_PYTHON" ]; then
+    echo -e "${RED}Error: Python 3 not found.${NC}"
+    echo "Please install Python 3.11 or later:"
+    echo "  sudo apt-get install python3"
+    exit 1
+fi
+
+# Get version numbers
+PY_VERSION=$("$SYSTEM_PYTHON" -c 'import sys; print(f"{sys.version_info.major}.{sys.version_info.minor}")')
+PY_MAJOR=$("$SYSTEM_PYTHON" -c 'import sys; print(sys.version_info.major)')
+PY_MINOR=$("$SYSTEM_PYTHON" -c 'import sys; print(sys.version_info.minor)')
+
+echo "  Found: $SYSTEM_PYTHON (Python $PY_VERSION)"
+
+# Check version range (3.11 <= version <= 3.14)
+if [ "$PY_MAJOR" -ne 3 ]; then
+    echo -e "${RED}Error: Python 3 required, found Python $PY_MAJOR${NC}"
+    exit 1
+fi
+
+if [ "$PY_MINOR" -lt 11 ]; then
+    echo -e "${RED}Error: Python 3.11+ required, found Python $PY_VERSION${NC}"
+    echo ""
+    echo "Raspberry Pi OS Bookworm ships with Python 3.11."
+    echo "Raspberry Pi OS Trixie ships with Python 3.13."
+    echo ""
+    echo "Please upgrade your Raspberry Pi OS or install Python 3.11+."
+    exit 1
+fi
+
+if [ "$PY_MINOR" -gt 14 ]; then
+    echo -e "${YELLOW}Warning: Python $PY_VERSION detected.${NC}"
+    echo "Stegasoo is tested with Python 3.11-3.14."
+    echo "Newer versions may work but are not officially supported."
+    read -p "Continue anyway? [y/N] " -n 1 -r
+    echo
+    if [[ ! $REPLY =~ ^[Yy]$ ]]; then
+        exit 1
+    fi
+fi
+
+echo -e "  ${GREEN}✓${NC} Python $PY_VERSION supported"
+
 # Check available memory
 TOTAL_MEM=$(free -m | awk '/^Mem:/{print $2}')
 if [ "$TOTAL_MEM" -lt 2000 ]; then
@@ -136,8 +190,11 @@ if [ "$TOTAL_MEM" -lt 2000 ]; then
     fi
 fi
 
-# Create /opt/stegasoo with proper permissions
+# =============================================================================
-echo -e "${GREEN}[1/12]${NC} Setting up install directory..."
+# Installation
+# =============================================================================
+
+echo -e "${GREEN}[1/9]${NC} Setting up install directory..."
 if [ ! -d "$INSTALL_DIR" ]; then
     sudo mkdir -p "$INSTALL_DIR"
     sudo chown "$USER:$USER" "$INSTALL_DIR"
@@ -148,7 +205,7 @@ else
     echo "  $INSTALL_DIR exists, updated ownership"
 fi
 
-echo -e "${GREEN}[2/12]${NC} Installing system dependencies..."
+echo -e "${GREEN}[2/9]${NC} Installing system dependencies..."
 sudo apt-get update
 sudo apt-get install -y \
     build-essential \
@@ -170,9 +227,11 @@ sudo apt-get install -y \
     libzbar0 \
     libjpeg-dev \
     python3-dev \
+    python3-venv \
+    python3-pip \
     btop
 
-echo -e "${GREEN}[3/12]${NC} Installing gum (TUI toolkit)..."
+echo -e "${GREEN}[3/9]${NC} Installing gum (TUI toolkit)..."
 # Add Charm repo for gum
 if ! command -v gum &>/dev/null; then
     sudo mkdir -p /etc/apt/keyrings
@@ -198,7 +257,7 @@ else
     echo "  mkcert already installed"
 fi
 
-echo -e "${GREEN}[4/12]${NC} Cloning Stegasoo..."
+echo -e "${GREEN}[4/9]${NC} Cloning Stegasoo..."
 
 # Clone Stegasoo first (needed to check for pre-built tarball)
 if [ -d "$INSTALL_DIR/.git" ]; then
@@ -212,17 +271,16 @@ else
     cd "$INSTALL_DIR"
 fi
 
-# Pre-built environment tarball (skips 20+ min compile time)
+# Pre-built venv tarball (skips pip compile time)
-# Includes both pyenv Python 3.12 AND venv with all dependencies
+PREBUILT_TARBALL="$INSTALL_DIR/rpi/stegasoo-rpi-venv-arm64.tar.zst"
-PREBUILT_TARBALL="$INSTALL_DIR/rpi/stegasoo-rpi-runtime-env-arm64.tar.zst"
+PREBUILT_URL="${PREBUILT_URL:-https://github.com/adlee-was-taken/stegasoo/releases/download/v4.2.0/stegasoo-rpi-venv-arm64.tar.zst}"
-PREBUILT_URL="${PREBUILT_URL:-https://github.com/adlee-was-taken/stegasoo/releases/download/v4.1.5/stegasoo-rpi-runtime-env-arm64.tar.zst}"
 USE_PREBUILT=true
 
 # Use local tarball if present, otherwise will download
 if [ -f "$PREBUILT_TARBALL" ]; then
-    echo -e "${GREEN}Found local pre-built environment - fast install mode${NC}"
+    echo -e "${GREEN}Found local pre-built venv - fast install mode${NC}"
 else
-    echo -e "${GREEN}Will download pre-built environment - fast install mode${NC}"
+    echo -e "${GREEN}Will download pre-built venv - fast install mode${NC}"
 fi
 
 # Allow --no-prebuilt flag to force from-source build
@@ -231,44 +289,30 @@ if [[ " $* " =~ " --no-prebuilt " ]] || [[ " $* " =~ " --from-source " ]]; then
     echo -e "${YELLOW}Building from source (--no-prebuilt specified)${NC}"
 fi
 
-# Fast path: use pre-built environment if available
+echo -e "${GREEN}[5/9]${NC} Setting up Python environment..."
+
 if [ "$USE_PREBUILT" = true ]; then
-    echo -e "${GREEN}[5/8]${NC} Installing pre-built Python environment..."
+    # Fast path: use pre-built venv
 
     # Download if local file doesn't exist
     if [ ! -f "$PREBUILT_TARBALL" ]; then
-        echo "  Downloading pre-built environment (~50MB)..."
+        echo "  Downloading pre-built venv (~50MB)..."
         curl -L -o "$PREBUILT_TARBALL" "$PREBUILT_URL"
     fi
 
-    # Extract pre-built environment (includes pyenv Python + venv)
+    # Extract pre-built venv
-    echo "  Extracting pre-built environment..."
+    echo "  Extracting pre-built venv..."
-    zstd -d "$PREBUILT_TARBALL" --stdout | tar -xf - -C "$HOME"
+    zstd -d "$PREBUILT_TARBALL" --stdout | tar -xf - -C "$INSTALL_DIR"
 
-    # Setup pyenv in current shell
+    # Fix venv Python symlinks to point to system Python
-    export PYENV_ROOT="$HOME/.pyenv"
+    echo "  Updating venv to use system Python..."
-    export PATH="$PYENV_ROOT/bin:$PATH"
+    rm -f "$INSTALL_DIR/venv/bin/python" "$INSTALL_DIR/venv/bin/python3"
-    eval "$(pyenv init -)"
+    ln -s "$SYSTEM_PYTHON" "$INSTALL_DIR/venv/bin/python"
-    pyenv global $PYTHON_VERSION
+    ln -s "$SYSTEM_PYTHON" "$INSTALL_DIR/venv/bin/python3"
 
-    # Add to .bashrc if not already there
+    # Update pip shebang if needed
-    if ! grep -q 'PYENV_ROOT' ~/.bashrc; then
+    if [ -f "$INSTALL_DIR/venv/bin/pip" ]; then
-        echo '' >> ~/.bashrc
+        sed -i "1s|^#!.*|#!$INSTALL_DIR/venv/bin/python|" "$INSTALL_DIR/venv/bin/pip" 2>/dev/null || true
-        echo '# pyenv' >> ~/.bashrc
-        echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bashrc
-        echo '[[ -d $PYENV_ROOT/bin ]] && export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bashrc
-        echo 'eval "$(pyenv init - bash)"' >> ~/.bashrc
-    fi
-
-    # Verify Python
-    INSTALLED_PY=$(python --version 2>&1 | cut -d' ' -f2 | cut -d'.' -f1,2)
-    echo -e "  ${GREEN}✓${NC} Python: $INSTALLED_PY"
-
-    # Extract venv to install dir
-    echo -e "${GREEN}[6/8]${NC} Setting up virtual environment..."
-    if [ -f "$HOME/stegasoo-venv.tar.zst" ]; then
-        zstd -d "$HOME/stegasoo-venv.tar.zst" --stdout | tar -xf - -C "$INSTALL_DIR"
-        rm "$HOME/stegasoo-venv.tar.zst"
     fi
 
     # Activate and verify
@@ -277,105 +321,87 @@ if [ "$USE_PREBUILT" = true ]; then
     echo -e "  ${GREEN}✓${NC} venv Python: $VENV_PY"
 
     # Install stegasoo package in editable mode (quick, no compile)
-    echo -e "${GREEN}[7/8]${NC} Installing Stegasoo package..."
+    echo "  Installing Stegasoo package..."
     pip install -e "." --quiet
-
-    # Adjust step numbers for rest of script
-    STEP_OFFSET=-4
 else
-    echo -e "${GREEN}[5/12]${NC} Installing pyenv and Python $PYTHON_VERSION..."
+    # Build from source
+    echo -e "  ${YELLOW}Building from source (this takes 5-10 minutes)${NC}"
 
-    # Install pyenv if not present
+    # Create venv with system Python
-    if [ ! -d "$HOME/.pyenv" ]; then
+    if [ ! -d "$INSTALL_DIR/venv" ]; then
-        curl https://pyenv.run | bash
+        "$SYSTEM_PYTHON" -m venv "$INSTALL_DIR/venv"
-
-        # Add pyenv to current shell
-        export PYENV_ROOT="$HOME/.pyenv"
-        export PATH="$PYENV_ROOT/bin:$PATH"
-        eval "$(pyenv init -)"
-
-        # Add to .bashrc if not already there
-        if ! grep -q 'PYENV_ROOT' ~/.bashrc; then
-            echo '' >> ~/.bashrc
-            echo '# pyenv' >> ~/.bashrc
-            echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bashrc
-            echo '[[ -d $PYENV_ROOT/bin ]] && export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bashrc
-            echo 'eval "$(pyenv init - bash)"' >> ~/.bashrc
     fi
-    else
+    source "$INSTALL_DIR/venv/bin/activate"
-        echo "  pyenv already installed"
-        export PYENV_ROOT="$HOME/.pyenv"
-        export PATH="$PYENV_ROOT/bin:$PATH"
-        eval "$(pyenv init -)"
-    fi
-
-    # Install Python 3.12 if not present
-    if ! pyenv versions | grep -q "$PYTHON_VERSION"; then
-        echo "  Building Python $PYTHON_VERSION (this takes ~10 minutes)..."
-        pyenv install $PYTHON_VERSION
-    else
-        echo "  Python $PYTHON_VERSION already installed"
-    fi
-    pyenv global $PYTHON_VERSION
-
-    # Verify Python version
-    INSTALLED_PY=$(python --version 2>&1 | cut -d' ' -f2 | cut -d'.' -f1,2)
-    if [ "$INSTALLED_PY" != "$PYTHON_VERSION" ]; then
-        echo -e "${RED}Error: Python $PYTHON_VERSION not active. Got: $INSTALLED_PY${NC}"
-        exit 1
-    fi
-    echo -e "${GREEN}[6/12]${NC} Creating Python virtual environment..."
-    echo -e "  ${YELLOW}Note: No pre-built venv found. Building from source (20+ min)${NC}"
-    echo -e "  ${YELLOW}To speed up future installs, add stegasoo-venv-pi-arm64.tar.gz to rpi/${NC}"
-
-    # Create venv with pyenv Python (not system Python)
-    # Use pyenv which to get actual path (handles 3.12 -> 3.12.12 mapping)
-    PYENV_PYTHON=$(pyenv which python)
-    echo "  Using Python: $PYENV_PYTHON"
-    if [ ! -d "venv" ]; then
-        "$PYENV_PYTHON" -m venv venv
-    fi
-    source venv/bin/activate
 
     # Verify we're using the right Python
     VENV_PY=$(python --version 2>&1 | cut -d' ' -f2 | cut -d'.' -f1,2)
     echo "  venv Python: $VENV_PY"
 
-    echo -e "${GREEN}[7/12]${NC} Building jpegio for ARM..."
+    # Upgrade pip and install build tools
+    pip install --upgrade pip setuptools wheel
 
-    # Clone jpegio
+    # Install jpeglib (no ARM64 wheel, PyPI tarball missing headers - use GitHub)
-    JPEGIO_DIR="/tmp/jpegio-build"
+    echo "  Installing jpeglib for ARM64..."
-    rm -rf "$JPEGIO_DIR"
+    JPEGLIB_WORKDIR=$(mktemp -d)
-    git clone "$JPEGIO_REPO" "$JPEGIO_DIR"
+    cd "$JPEGLIB_WORKDIR"
 
-    # Apply ARM64 patch
+    # Clone from GitHub (PyPI source tarball is missing .h files)
-    if [ -f "$INSTALL_DIR/rpi/patches/jpegio/apply-patch.sh" ]; then
+    echo "    Cloning jpeglib from GitHub..."
-        bash "$INSTALL_DIR/rpi/patches/jpegio/apply-patch.sh" "$JPEGIO_DIR"
+    git clone --depth 1 --branch 1.0.2 https://github.com/martinbenes1996/jpeglib.git
-    else
+    cd jpeglib
-        echo "  Applying inline ARM64 patch..."
+    CJPEGLIB="src/jpeglib/cjpeglib"
-        sed -i "s/cargs.append('-m64')/pass  # ARM64 fix/g" "$JPEGIO_DIR/setup.py"
-    fi
 
-    cd "$JPEGIO_DIR"
+    # Fix broken include paths in setup.py (uses jpeglib/ but files are in src/jpeglib/)
+    ln -s src/jpeglib jpeglib
 
-    # Build jpegio into venv
+    # Download libjpeg headers (not included in repo either)
-    pip install --upgrade pip setuptools wheel cython numpy
+    # Each version needs EXACT matching headers (APIs differ between versions)
+    echo "    Downloading libjpeg headers (all versions)..."
+
+    # Download each version separately (APIs are incompatible between versions)
+    for v in 6b 7 8 8a 8b 8c 8d 9 9a 9b 9c 9d 9e 9f; do
+        echo "      libjpeg $v..."
+        curl -sL "https://www.ijg.org/files/jpegsrc.v${v}.tar.gz" | tar -xzf -
+        cp jpeg-${v}/*.h "$CJPEGLIB/$v/" 2>/dev/null || cp jpeg-${v//.}/*.h "$CJPEGLIB/$v/" 2>/dev/null || true
+    done
+
+    # Skip turbo/mozjpeg - they need cmake-generated headers
+    # Only remove dict entries (lines 49-59), keep if blocks (they're safe when is_turbo=False)
+    echo "    Patching setup.py to skip turbo/mozjpeg (need cmake)..."
+    python3 << 'PYPATCH'
+with open('setup.py', 'r') as f:
+    lines = f.readlines()
+filtered = []
+for line in lines:
+    # Only skip dict entries like "'turbo120': ..." or "'mozjpeg101': ..."
+    stripped = line.strip()
+    if stripped.startswith("'turbo") and ':' in stripped:
+        continue
+    if stripped.startswith("'mozjpeg") and ':' in stripped:
+        continue
+    if stripped.startswith("# 'turbo"):  # commented turbo line
+        continue
+    filtered.append(line)
+with open('setup.py', 'w') as f:
+    f.writelines(filtered)
+PYPATCH
+
+    # Build and install
+    echo "    Building jpeglib (this takes a few minutes)..."
     pip install .
-
     cd "$INSTALL_DIR"
-    rm -rf "$JPEGIO_DIR"
+    rm -rf "$JPEGLIB_WORKDIR"
 
-    echo -e "${GREEN}[8/12]${NC} Installing Stegasoo..."
+    # Install remaining dependencies
-
+    echo "  Installing remaining dependencies..."
-    # Install dependencies (jpegio already in venv, won't re-download)
     pip install -e ".[web]"
-
-    STEP_OFFSET=0
 fi
 
-echo -e "${GREEN}[9/12]${NC} Creating systemd service..."
+echo -e "  ${GREEN}✓${NC} Stegasoo installed"
 
-# Create systemd service file
+echo -e "${GREEN}[6/9]${NC} Creating systemd services..."
+
+# Create systemd service file for Web UI
 sudo tee /etc/systemd/system/stegasoo.service > /dev/null <<EOF
 [Unit]
 Description=Stegasoo Web UI
@@ -397,12 +423,53 @@ RestartSec=5
 WantedBy=multi-user.target
 EOF
 
-echo -e "${GREEN}[10/12]${NC} Enabling service..."
+# Create systemd service file for REST API (optional)
+sudo tee /etc/systemd/system/stegasoo-api.service > /dev/null <<EOF
+[Unit]
+Description=Stegasoo REST API
+After=network.target
+
+[Service]
+Type=simple
+User=$USER
+WorkingDirectory=$INSTALL_DIR/frontends/api
+Environment="PATH=$INSTALL_DIR/venv/bin:/usr/bin"
+Environment="PYTHONPATH=$INSTALL_DIR/src"
+ExecStart=$INSTALL_DIR/venv/bin/uvicorn main:app --host 0.0.0.0 --port 8000
+Restart=on-failure
+RestartSec=5
+
+[Install]
+WantedBy=multi-user.target
+EOF
+
+echo -e "${GREEN}[7/9]${NC} Enabling services..."
 
 sudo systemctl daemon-reload
 sudo systemctl enable stegasoo.service
 
-echo -e "${GREEN}[11/12]${NC} Setting up user environment..."
+# Prompt for REST API service (optional, with security warning)
+echo ""
+echo -e "${CYAN}Would you like to enable the REST API service? (port 8000)${NC}"
+echo ""
+echo -e "  ${RED}⚠ WARNING: The REST API has NO AUTHENTICATION${NC}"
+echo "  Anyone on your network can use it to encode/decode messages."
+echo "  Only enable if you understand the security implications."
+echo ""
+echo "  The Web UI (port 5000) has authentication and works independently."
+echo ""
+read -p "Enable REST API (no auth)? [y/N]: " ENABLE_API
+if [[ "$ENABLE_API" =~ ^[Yy]$ ]]; then
+    sudo systemctl enable stegasoo-api.service
+    STEGASOO_API_ENABLED=true
+    echo -e "  ${YELLOW}⚠${NC} REST API enabled on port 8000 ${RED}(no authentication)${NC}"
+else
+    STEGASOO_API_ENABLED=false
+    echo -e "  ${GREEN}✓${NC} REST API not enabled (recommended)"
+    echo "    Can enable later with: sudo systemctl enable --now stegasoo-api"
+fi
+
+echo -e "${GREEN}[8/9]${NC} Setting up user environment..."
 
 # Add stegasoo venv and rpi scripts to PATH for all users
 sudo tee /etc/profile.d/stegasoo-path.sh > /dev/null <<'PATHEOF'
@@ -436,7 +503,7 @@ if [ -f "$INSTALL_DIR/docs/stegasoo.1" ]; then
     echo "  Installed man page (man stegasoo)"
 fi
 
-echo -e "${GREEN}[12/12]${NC} Setting up login banner..."
+echo -e "${GREEN}[9/9]${NC} Setting up login banner..."
 
 # Create dynamic MOTD script
 sudo tee /etc/profile.d/stegasoo-motd.sh > /dev/null <<'MOTDEOF'
@@ -727,6 +794,14 @@ echo "  Start:   sudo systemctl start stegasoo"
 echo "  Stop:    sudo systemctl stop stegasoo"
 echo "  Status:  sudo systemctl status stegasoo"
 echo "  Logs:    journalctl -u stegasoo -f"
+if [ "$STEGASOO_API_ENABLED" = "true" ]; then
+    echo ""
+    echo -e "${GREEN}REST API Commands:${NC}"
+    echo "  Start:   sudo systemctl start stegasoo-api"
+    echo "  Stop:    sudo systemctl stop stegasoo-api"
+    echo "  Status:  sudo systemctl status stegasoo-api"
+    echo "  Logs:    journalctl -u stegasoo-api -f"
+fi
 echo ""
 
 # Offer to start now
@@ -734,9 +809,12 @@ read -p "Start Stegasoo now? [Y/n] " -n 1 -r
 echo
 if [[ ! $REPLY =~ ^[Nn]$ ]]; then
     sudo systemctl start stegasoo
+    if [ "$STEGASOO_API_ENABLED" = "true" ]; then
+        sudo systemctl start stegasoo-api
+    fi
     sleep 2
     if systemctl is-active --quiet stegasoo; then
-        echo -e "${GREEN}✓ Stegasoo is running!${NC}"
+        echo -e "${GREEN}✓ Stegasoo Web UI is running!${NC}"
         if [ "$ENABLE_HTTPS" = "true" ]; then
             if [ "$USE_PORT_443" = "true" ]; then
                 echo -e "  Create admin: ${YELLOW}https://$PI_HOST.local/setup${NC} or ${YELLOW}https://$PI_IP/setup${NC}"
@@ -746,6 +824,13 @@ if [[ ! $REPLY =~ ^[Nn]$ ]]; then
         else
             echo -e "  Create admin: ${YELLOW}http://$PI_HOST.local:5000/setup${NC} or ${YELLOW}http://$PI_IP:5000/setup${NC}"
         fi
+        if [ "$STEGASOO_API_ENABLED" = "true" ]; then
+            if systemctl is-active --quiet stegasoo-api; then
+                echo -e "${GREEN}✓ Stegasoo REST API is running on port 8000${NC}"
+            else
+                echo -e "${YELLOW}⚠ REST API failed to start. Check logs:${NC} journalctl -u stegasoo-api -f"
+            fi
+        fi
     else
         echo -e "${RED}✗ Failed to start. Check logs:${NC} journalctl -u stegasoo -f"
     fi

scripts/build.sh

@@ -68,6 +68,15 @@ case "${1:-fast}" in
         echo -e "${GREEN}Cleaned!${NC}"
         ;;
 
+    rebuild)
+        echo -e "${YELLOW}Full rebuild from scratch (no cache)...${NC}"
+        $SUDO $COMPOSE_CMD -f "$COMPOSE_FILE" down --rmi local -v 2>/dev/null || true
+        $SUDO docker rmi stegasoo-base:latest 2>/dev/null || true
+        $SUDO docker build --no-cache -f "$DOCKER_DIR/Dockerfile.base" -t stegasoo-base:latest .
+        $SUDO $COMPOSE_CMD -f "$COMPOSE_FILE" build --no-cache
+        echo -e "${GREEN}Done! Start with: $COMPOSE_CMD -f docker/docker-compose.yml up -d${NC}"
+        ;;
+
     *)
         echo -e "${CYAN}Stegasoo Build Script${NC}"
         echo ""
@@ -76,12 +85,14 @@ case "${1:-fast}" in
         echo "Commands:"
         echo "  base    Build the base image (one-time, 5-10 min)"
         echo "  fast    Fast build using base image (default, ~10 sec)"
-        echo "  full   Full rebuild from scratch (slow, no base needed)"
+        echo "  full    Rebuild services without cache (uses existing base)"
+        echo "  rebuild Complete rebuild with no cache (base + services)"
         echo "  clean   Remove all images and volumes"
         echo ""
         echo "Typical workflow:"
         echo "  1. First time:   $0 base"
         echo "  2. Daily dev:    $0 fast"
         echo "  3. Deps change:  $0 base"
+        echo "  4. Nuclear:      $0 rebuild"
         ;;
 esac

scripts/screenshots.sh

@@ -33,8 +33,8 @@ for cmd in chromium magick curl; do
     fi
 done
 
-# Check if server is running
+# Check if server is running (-k for self-signed certs)
-if ! curl -s "$BASE_URL" > /dev/null 2>&1; then
+if ! curl -sk "$BASE_URL" > /dev/null 2>&1; then
     echo "Error: Server not responding at $BASE_URL"
     echo "Start with: STEGASOO_AUTH_ENABLED=false python frontends/web/app.py"
     exit 1
@@ -49,7 +49,7 @@ capture() {
     printf "  %-20s <- %s\n" "$name" "$route"
     chromium --headless --screenshot="$OUTPUT_DIR/$name.png" \
         --window-size="$WINDOW_SIZE" --hide-scrollbars \
-        --disable-gpu --no-sandbox \
+        --disable-gpu --no-sandbox --ignore-certificate-errors \
         "$url" 2>/dev/null
 }
 

src/stegasoo/__init__.py

@@ -7,7 +7,7 @@ Changes in v4.0.0:
 - encode() and decode() now accept channel_key parameter
 """
 
-__version__ = "4.1.7"
+__version__ = "4.2.0"
 
 # Core functionality
 # Channel key management (v4.0.0)

src/stegasoo/compression.py

@@ -17,6 +17,14 @@ try:
 except ImportError:
     HAS_LZ4 = False
 
+# Optional ZSTD support (best ratio, fast)
+try:
+    import zstandard as zstd
+
+    HAS_ZSTD = True
+except ImportError:
+    HAS_ZSTD = False
+
 
 class CompressionAlgorithm(IntEnum):
     """Supported compression algorithms."""
@@ -24,6 +32,7 @@ class CompressionAlgorithm(IntEnum):
     NONE = 0
     ZLIB = 1
     LZ4 = 2
+    ZSTD = 3  # v4.2.0: Best ratio, fast compression
 
 
 # Magic bytes for compressed payloads
@@ -72,6 +81,15 @@ def compress(data: bytes, algorithm: CompressionAlgorithm = CompressionAlgorithm
             algorithm = CompressionAlgorithm.ZLIB
         else:
             compressed = lz4.frame.compress(data)
+
+    elif algorithm == CompressionAlgorithm.ZSTD:
+        if not HAS_ZSTD:
+            # Fall back to zlib if ZSTD not available
+            compressed = zlib.compress(data, level=ZLIB_LEVEL)
+            algorithm = CompressionAlgorithm.ZLIB
+        else:
+            cctx = zstd.ZstdCompressor(level=19)  # High compression level
+            compressed = cctx.compress(data)
     else:
         raise CompressionError(f"Unknown compression algorithm: {algorithm}")
 
@@ -123,6 +141,15 @@ def decompress(data: bytes) -> bytes:
             result = lz4.frame.decompress(compressed_data)
         except Exception as e:
             raise CompressionError(f"LZ4 decompression failed: {e}")
+
+    elif algorithm == CompressionAlgorithm.ZSTD:
+        if not HAS_ZSTD:
+            raise CompressionError("ZSTD compression used but zstandard package not installed")
+        try:
+            dctx = zstd.ZstdDecompressor()
+            result = dctx.decompress(compressed_data)
+        except Exception as e:
+            raise CompressionError(f"ZSTD decompression failed: {e}")
     else:
         raise CompressionError(f"Unknown compression algorithm: {algorithm}")
 
@@ -181,6 +208,9 @@ def estimate_compressed_size(
         compressed_sample = zlib.compress(sample, level=ZLIB_LEVEL)
     elif algorithm == CompressionAlgorithm.LZ4 and HAS_LZ4:
         compressed_sample = lz4.frame.compress(sample)
+    elif algorithm == CompressionAlgorithm.ZSTD and HAS_ZSTD:
+        cctx = zstd.ZstdCompressor(level=19)
+        compressed_sample = cctx.compress(sample)
     else:
         compressed_sample = zlib.compress(sample, level=ZLIB_LEVEL)
 
@@ -195,14 +225,24 @@ def get_available_algorithms() -> list[CompressionAlgorithm]:
     algorithms = [CompressionAlgorithm.NONE, CompressionAlgorithm.ZLIB]
     if HAS_LZ4:
         algorithms.append(CompressionAlgorithm.LZ4)
+    if HAS_ZSTD:
+        algorithms.append(CompressionAlgorithm.ZSTD)
     return algorithms
 
 
+def get_best_algorithm() -> CompressionAlgorithm:
+    """Get the best available compression algorithm (prefer ZSTD > ZLIB > LZ4)."""
+    if HAS_ZSTD:
+        return CompressionAlgorithm.ZSTD
+    return CompressionAlgorithm.ZLIB
+
+
 def algorithm_name(algo: CompressionAlgorithm) -> str:
     """Get human-readable algorithm name."""
     names = {
         CompressionAlgorithm.NONE: "None",
         CompressionAlgorithm.ZLIB: "Zlib (deflate)",
         CompressionAlgorithm.LZ4: "LZ4 (fast)",
+        CompressionAlgorithm.ZSTD: "Zstd (best)",
     }
     return names.get(algo, "Unknown")

src/stegasoo/constants.py

@@ -1,9 +1,15 @@
 """
-Stegasoo Constants and Configuration (v4.0.2 - Web UI Authentication)
+Stegasoo Constants and Configuration (v4.2.0 - Performance & Compression)
 
 Central location for all magic numbers, limits, and crypto parameters.
 All version numbers, limits, and configuration values should be defined here.
 
+CHANGES in v4.2.0:
+- Added zstd compression for QR codes (better ratio than zlib)
+- RSA key size capped at 3072 bits (4096 too large for QR codes)
+- Progress bar improvements for encode/decode operations
+- File auto-expire increased to 10 minutes
+
 CHANGES in v4.0.2:
 - Added Web UI authentication with SQLite3 user storage
 - Added optional HTTPS with auto-generated self-signed certificates
@@ -25,7 +31,7 @@ from pathlib import Path
 # VERSION
 # ============================================================================
 
-__version__ = "4.1.5"
+__version__ = "4.2.0"
 
 # ============================================================================
 # FILE FORMAT
@@ -98,7 +104,7 @@ DEFAULT_PHRASE_WORDS = DEFAULT_PASSPHRASE_WORDS
 
 # RSA configuration
 MIN_RSA_BITS = 2048
-VALID_RSA_SIZES = (2048, 3072, 4096)
+VALID_RSA_SIZES = (2048, 3072)  # 4096 removed - too large for QR codes
 DEFAULT_RSA_BITS = 2048
 
 MIN_KEY_PASSWORD_LENGTH = 8
@@ -108,8 +114,8 @@ MIN_KEY_PASSWORD_LENGTH = 8
 # ============================================================================
 
 # Temporary file storage
-TEMP_FILE_EXPIRY = 300  # 5 minutes in seconds
+TEMP_FILE_EXPIRY = 600  # 10 minutes in seconds
-TEMP_FILE_EXPIRY_MINUTES = 5
+TEMP_FILE_EXPIRY_MINUTES = 10
 
 # Thumbnail settings
 THUMBNAIL_SIZE = (250, 250)  # Maximum dimensions for thumbnails

src/stegasoo/dct_steganography.py

@@ -12,7 +12,7 @@ Why is this cool?
 
 Two approaches depending on what you want:
 1. PNG output: We do our own DCT math via scipy (works on any image)
-2. JPEG output: We use jpegio to directly tweak the coefficients (chef's kiss)
+2. JPEG output: We use jpeglib to directly tweak the coefficients (chef's kiss)
 
 v4.1.0 - The "please stop corrupting my data" release:
 - Reed-Solomon error correction (can fix up to 16 byte errors per chunk)
@@ -24,7 +24,7 @@ v3.2.0-patch2 - The "scipy why are you like this" release:
 - Process blocks one at a time with fresh arrays
 - Yes, it's slower. No, I don't care. Correctness > speed.
 
-Requires: scipy (PNG mode), optionally jpegio (JPEG mode), reedsolo (error correction)
+Requires: scipy (PNG mode), optionally jpeglib (JPEG mode), reedsolo (error correction)
 """
 
 import gc
@@ -40,27 +40,30 @@ from PIL import Image
 # Check for scipy availability (for PNG/DCT mode)
 # Prefer scipy.fft (newer, more stable) over scipy.fftpack
 try:
-    from scipy.fft import dct, idct
+    from scipy.fft import dct, idct, dctn, idctn
 
     HAS_SCIPY = True
 except ImportError:
     try:
-        from scipy.fftpack import dct, idct
+        from scipy.fftpack import dct, idct, dctn, idctn
 
         HAS_SCIPY = True
     except ImportError:
         HAS_SCIPY = False
         dct = None
         idct = None
+        dctn = None
+        idctn = None
 
-# Check for jpegio availability (for proper JPEG mode)
+# Check for jpeglib availability (for proper JPEG mode)
+# jpeglib replaces jpegio for Python 3.13+ compatibility
 try:
-    import jpegio as jio
+    import jpeglib
 
-    HAS_JPEGIO = True
+    HAS_JPEGIO = True  # Keep variable name for compatibility
 except ImportError:
     HAS_JPEGIO = False
-    jio = None
+    jpeglib = None
 
 # Import custom exceptions
 from .exceptions import InvalidMagicBytesError
@@ -406,28 +409,30 @@ def _safe_idct2(block: np.ndarray) -> np.ndarray:
 def _to_grayscale(image_data: bytes) -> np.ndarray:
     img = Image.open(io.BytesIO(image_data))
     gray = img.convert("L")
-    return np.array(gray, dtype=np.float64, copy=True, order="C")
+    return np.array(gray, dtype=np.float32, copy=True, order="C")
 
 
 def _extract_y_channel(image_data: bytes) -> np.ndarray:
+    """Extract Y (luminance) channel - float32 for memory efficiency."""
     img = Image.open(io.BytesIO(image_data))
     if img.mode != "RGB":
         img = img.convert("RGB")
 
-    rgb = np.array(img, dtype=np.float64, copy=True, order="C")
+    rgb = np.array(img, dtype=np.float32, copy=True, order="C")
     Y = 0.299 * rgb[:, :, 0] + 0.587 * rgb[:, :, 1] + 0.114 * rgb[:, :, 2]
-    return np.array(Y, dtype=np.float64, copy=True, order="C")
+    return np.array(Y, dtype=np.float32, copy=True, order="C")
 
 
 def _pad_to_blocks(image: np.ndarray) -> tuple[np.ndarray, tuple[int, int]]:
+    """Pad image to block boundaries - uses float32 for memory efficiency."""
     h, w = image.shape
     new_h = ((h + BLOCK_SIZE - 1) // BLOCK_SIZE) * BLOCK_SIZE
     new_w = ((w + BLOCK_SIZE - 1) // BLOCK_SIZE) * BLOCK_SIZE
 
     if new_h == h and new_w == w:
-        return np.array(image, dtype=np.float64, copy=True, order="C"), (h, w)
+        return np.array(image, dtype=np.float32, copy=True, order="C"), (h, w)
 
-    padded = np.zeros((new_h, new_w), dtype=np.float64, order="C")
+    padded = np.zeros((new_h, new_w), dtype=np.float32, order="C")
     padded[:h, :w] = image
 
     # Simple edge replication for padding
@@ -444,8 +449,9 @@ def _pad_to_blocks(image: np.ndarray) -> tuple[np.ndarray, tuple[int, int]]:
 
 
 def _unpad_image(image: np.ndarray, original_size: tuple[int, int]) -> np.ndarray:
+    """Remove padding - uses float32 for memory efficiency."""
     h, w = original_size
-    return np.array(image[:h, :w], dtype=np.float64, copy=True, order="C")
+    return np.array(image[:h, :w], dtype=np.float32, copy=True, order="C")
 
 
 def _embed_bit_in_coeff(coef: float, bit: int, quant_step: int = QUANT_STEP) -> float:
@@ -543,20 +549,23 @@ def _rgb_to_ycbcr(rgb: np.ndarray) -> tuple[np.ndarray, np.ndarray, np.ndarray]:
     - Cb/Cr are often subsampled (4:2:0) so Y has more capacity anyway
 
     The coefficients here are from ITU-R BT.601 - the standard for video.
+
+    Uses float32 to reduce memory usage (~50% savings vs float64).
     """
-    R = rgb[:, :, 0].astype(np.float64)
+    # Use float32 - sufficient precision for 8-bit images, halves memory
-    G = rgb[:, :, 1].astype(np.float64)
+    R = rgb[:, :, 0].astype(np.float32)
-    B = rgb[:, :, 2].astype(np.float64)
+    G = rgb[:, :, 1].astype(np.float32)
+    B = rgb[:, :, 2].astype(np.float32)
 
     # Y = luminance (brightness). Green contributes most because eyes are most sensitive to it.
-    Y = np.array(0.299 * R + 0.587 * G + 0.114 * B, dtype=np.float64, copy=True, order="C")
+    Y = np.array(0.299 * R + 0.587 * G + 0.114 * B, dtype=np.float32, copy=True, order="C")
     # Cb = blue-difference chroma (centered at 128)
     Cb = np.array(
-        128 - 0.168736 * R - 0.331264 * G + 0.5 * B, dtype=np.float64, copy=True, order="C"
+        128 - 0.168736 * R - 0.331264 * G + 0.5 * B, dtype=np.float32, copy=True, order="C"
     )
     # Cr = red-difference chroma (centered at 128)
     Cr = np.array(
-        128 + 0.5 * R - 0.418688 * G - 0.081312 * B, dtype=np.float64, copy=True, order="C"
+        128 + 0.5 * R - 0.418688 * G - 0.081312 * B, dtype=np.float32, copy=True, order="C"
     )
 
     return Y, Cb, Cr
@@ -569,11 +578,12 @@ def _ycbcr_to_rgb(Y: np.ndarray, Cb: np.ndarray, Cr: np.ndarray) -> np.ndarray:
     After embedding in the Y channel, we need to reconstruct RGB for display.
     The Cb/Cr channels are unchanged - we only touched luminance.
     """
+    # Use float32 for memory efficiency
     R = Y + 1.402 * (Cr - 128)
     G = Y - 0.344136 * (Cb - 128) - 0.714136 * (Cr - 128)
     B = Y + 1.772 * (Cb - 128)
 
-    rgb = np.zeros((Y.shape[0], Y.shape[1], 3), dtype=np.float64, order="C")
+    rgb = np.zeros((Y.shape[0], Y.shape[1], 3), dtype=np.float32, order="C")
     rgb[:, :, 0] = R
     rgb[:, :, 1] = G
     rgb[:, :, 2] = B
@@ -733,7 +743,7 @@ def estimate_capacity_comparison(image_data: bytes) -> dict:
         },
         "jpeg_native": {
             "available": HAS_JPEGIO,
-            "note": "Uses jpegio for proper JPEG coefficient embedding",
+            "note": "Uses jpeglib for proper JPEG coefficient embedding",
         },
     }
 
@@ -818,8 +828,8 @@ def _embed_scipy_dct_safe(
         if img.mode == "RGBA":
             img = img.convert("RGB")
 
-        # Process color image
+        # Process color image (float32 for memory efficiency)
-        rgb = np.array(img, dtype=np.float64, copy=True, order="C")
+        rgb = np.array(img, dtype=np.float32, copy=True, order="C")
         img.close()
 
         Y, Cb, Cr = _rgb_to_ycbcr(rgb)
@@ -891,61 +901,105 @@ def _embed_in_channel_safe(
     progress_file: str | None = None,
 ) -> np.ndarray:
     """
-    Embed bits in channel using safe DCT operations.
+    Embed bits in channel using vectorized DCT operations.
 
-    Processes one block at a time with fresh array allocations.
+    Processes blocks in batches for ~10x speedup over sequential processing.
     """
     h, w = channel.shape
 
-    # Create result with explicit new memory
+    # Create result with explicit new memory (float32 for memory efficiency)
-    result = np.array(channel, dtype=np.float64, copy=True, order="C")
+    result = np.array(channel, dtype=np.float32, copy=True, order="C")
 
+    # Pre-compute embed positions as numpy indices
+    embed_rows = np.array([pos[0] for pos in DEFAULT_EMBED_POSITIONS])
+    embed_cols = np.array([pos[1] for pos in DEFAULT_EMBED_POSITIONS])
+    bits_per_block = len(DEFAULT_EMBED_POSITIONS)
+
+    # Calculate how many blocks we need
+    total_bits = len(bits)
+    blocks_needed = (total_bits + bits_per_block - 1) // bits_per_block
+    blocks_to_process = min(blocks_needed, len(block_order))
+
+    # Initial progress write - signals Argon2/prep is done, embedding starting
+    if progress_file:
+        _write_progress(progress_file, 5, 100, "embedding")
+
+    # Vectorized embedding: process blocks in batches
+    BATCH_SIZE = 500
     bit_idx = 0
-    total_blocks = len(block_order)
+    block_idx = 0
 
-    for block_idx, block_num in enumerate(block_order):
+    while block_idx < blocks_to_process and bit_idx < total_bits:
-        if bit_idx >= len(bits):
+        # Determine batch size
-            break
+        batch_end = min(block_idx + BATCH_SIZE, blocks_to_process)
+        batch_order = block_order[block_idx:batch_end]
+        batch_count = len(batch_order)
 
+        # Extract blocks into 3D array (float32 for memory efficiency)
+        blocks = np.zeros((batch_count, BLOCK_SIZE, BLOCK_SIZE), dtype=np.float32)
+        block_positions = []
+        for i, block_num in enumerate(batch_order):
             by = (block_num // blocks_x) * BLOCK_SIZE
             bx = (block_num % blocks_x) * BLOCK_SIZE
+            blocks[i] = result[by : by + BLOCK_SIZE, bx : bx + BLOCK_SIZE]
+            block_positions.append((by, bx))
 
-        # Extract block - create brand new array
+        # Vectorized 2D DCT on all blocks at once
-        block = np.array(
+        dct_blocks = dctn(blocks, axes=(1, 2), norm="ortho")
-            result[by : by + BLOCK_SIZE, bx : bx + BLOCK_SIZE],
-            dtype=np.float64,
-            copy=True,
-            order="C",
-        )
 
-        # Apply safe DCT (row-by-row)
+        # Embed bits in each block (vectorized where possible)
-        dct_block = _safe_dct2(block)
+        for i in range(batch_count):
-
+            if bit_idx >= total_bits:
-        # Embed bits
-        for pos in DEFAULT_EMBED_POSITIONS:
-            if bit_idx >= len(bits):
                 break
-            dct_block[pos[0], pos[1]] = _embed_bit_in_coeff(
+
-                float(dct_block[pos[0], pos[1]]), bits[bit_idx]
+            # Get bits for this block
+            block_bits = bits[bit_idx : bit_idx + bits_per_block]
+            num_bits = len(block_bits)
+
+            if num_bits == bits_per_block:
+                # Full block - vectorized embedding
+                coeffs = dct_blocks[i, embed_rows, embed_cols]
+                bit_array = np.array(block_bits)
+                # QIM embedding: round to grid, adjust for bit
+                quantized = np.round(coeffs / QUANT_STEP).astype(int)
+                # If quantized % 2 != bit, nudge coefficient
+                needs_adjust = (quantized % 2) != bit_array
+                # Determine direction to nudge
+                dct_blocks[i, embed_rows[needs_adjust], embed_cols[needs_adjust]] = (
+                    (quantized[needs_adjust] + (1 - 2 * (quantized[needs_adjust] % 2 == 1))) * QUANT_STEP
+                ).astype(np.float64)
+                # For bits that already match, just quantize
+                dct_blocks[i, embed_rows[~needs_adjust], embed_cols[~needs_adjust]] = (
+                    quantized[~needs_adjust] * QUANT_STEP
+                ).astype(np.float64)
+            else:
+                # Partial block - process remaining bits individually
+                for j, bit in enumerate(block_bits):
+                    row, col = embed_rows[j], embed_cols[j]
+                    dct_blocks[i, row, col] = _embed_bit_in_coeff(
+                        float(dct_blocks[i, row, col]), bit
                     )
-            bit_idx += 1
 
-        # Apply safe inverse DCT
+            bit_idx += num_bits
-        modified_block = _safe_idct2(dct_block)
 
-        # Copy back
+        # Vectorized inverse DCT
-        result[by : by + BLOCK_SIZE, bx : bx + BLOCK_SIZE] = modified_block
+        modified_blocks = idctn(dct_blocks, axes=(1, 2), norm="ortho")
 
-        # Clean up this iteration
+        # Copy modified blocks back to result
-        del block, dct_block, modified_block
+        for i, (by, bx) in enumerate(block_positions):
+            result[by : by + BLOCK_SIZE, bx : bx + BLOCK_SIZE] = modified_blocks[i]
+
+        # Cleanup
+        del blocks, dct_blocks, modified_blocks
+        block_idx = batch_end
 
         # Report progress periodically
         if progress_file and block_idx % PROGRESS_INTERVAL == 0:
-            _write_progress(progress_file, block_idx, total_blocks, "embedding")
+            _write_progress(progress_file, block_idx, blocks_to_process, "embedding")
 
     # Final progress update
     if progress_file:
-        _write_progress(progress_file, total_blocks, total_blocks, "finalizing")
+        _write_progress(progress_file, blocks_to_process, blocks_to_process, "finalizing")
 
     # Force garbage collection
     gc.collect()
@@ -1029,7 +1083,7 @@ def _embed_jpegio(
     flags = FLAG_COLOR_MODE if color_mode == "color" else 0
 
     try:
-        jpeg = jio.read(input_path)
+        jpeg = jpeglib.to_jpegio(jpeglib.read_dct(input_path))
         coef_array = jpeg.coef_arrays[JPEGIO_EMBED_CHANNEL]
 
         all_positions = _jpegio_get_usable_positions(coef_array)
@@ -1064,6 +1118,10 @@ def _embed_jpegio(
         total_bits = len(bits)
         progress_interval = max(total_bits // 20, 100)  # Report ~20 times or every 100 bits
 
+        # Initial progress write - signals prep is done, embedding starting
+        if progress_file:
+            _write_progress(progress_file, 5, 100, "embedding")
+
         for bit_idx, pos_idx in enumerate(order):
             if bit_idx >= len(bits):
                 break
@@ -1087,7 +1145,7 @@ def _embed_jpegio(
         if progress_file:
             _write_progress(progress_file, total_bits, total_bits, "saving")
 
-        jio.write(jpeg, output_path)
+        jpeg.write(output_path)
 
         with open(output_path, "rb") as f:
             stego_bytes = f.read()
@@ -1115,7 +1173,11 @@ def _embed_jpegio(
                 pass
 
 
-def extract_from_dct(stego_image: bytes, seed: bytes) -> bytes:
+def extract_from_dct(
+    stego_image: bytes,
+    seed: bytes,
+    progress_file: str | None = None,
+) -> bytes:
     """Extract data from DCT stego image."""
     img = Image.open(io.BytesIO(stego_image))
     fmt = img.format
@@ -1123,16 +1185,22 @@ def extract_from_dct(stego_image: bytes, seed: bytes) -> bytes:
 
     if fmt == "JPEG" and HAS_JPEGIO:
         try:
-            return _extract_jpegio(stego_image, seed)
+            return _extract_jpegio(stego_image, seed, progress_file)
         except ValueError:
             pass
 
     _check_scipy()
-    return _extract_scipy_dct_safe(stego_image, seed)
+    return _extract_scipy_dct_safe(stego_image, seed, progress_file)
 
 
-def _extract_scipy_dct_safe(stego_image: bytes, seed: bytes) -> bytes:
+def _extract_scipy_dct_safe(
-    """Extract using safe DCT operations."""
+    stego_image: bytes,
+    seed: bytes,
+    progress_file: str | None = None,
+) -> bytes:
+    """Extract using safe DCT operations with vectorized processing."""
+    # Progress starts at 25% (decode.py writes 20% for Argon2, 25% before extraction)
+
     img = Image.open(io.BytesIO(stego_image))
     width, height = img.size
     mode = img.mode
@@ -1156,26 +1224,54 @@ def _extract_scipy_dct_safe(stego_image: bytes, seed: bytes) -> bytes:
 
     block_order = _generate_block_order(num_blocks, seed)
 
+    # Vectorized extraction: process blocks in batches for ~10x speedup
+    # Batch size balances memory usage vs. parallelization benefit
+    BATCH_SIZE = 500
     all_bits = []
 
-    for block_num in block_order:
+    # Pre-compute embed positions as numpy indices for vectorized access
+    embed_rows = np.array([pos[0] for pos in DEFAULT_EMBED_POSITIONS])
+    embed_cols = np.array([pos[1] for pos in DEFAULT_EMBED_POSITIONS])
+
+    # Progress reporting interval - report frequently for responsive UI
+    PROGRESS_INTERVAL = 500  # Report every N blocks (matches BATCH_SIZE)
+
+    block_idx = 0
+    while block_idx < len(block_order):
+        # Determine batch size (may be smaller at end)
+        batch_end = min(block_idx + BATCH_SIZE, len(block_order))
+        batch_order = block_order[block_idx:batch_end]
+        batch_count = len(batch_order)
+
+        # Extract blocks into 3D array (batch_count, 8, 8) - float32 for memory efficiency
+        blocks = np.zeros((batch_count, BLOCK_SIZE, BLOCK_SIZE), dtype=np.float32)
+        for i, block_num in enumerate(batch_order):
             by = (block_num // blocks_x) * BLOCK_SIZE
             bx = (block_num % blocks_x) * BLOCK_SIZE
+            blocks[i] = padded[by : by + BLOCK_SIZE, bx : bx + BLOCK_SIZE]
 
-        block = np.array(
+        # Vectorized 2D DCT on all blocks at once (~10-15x faster than sequential)
-            padded[by : by + BLOCK_SIZE, bx : bx + BLOCK_SIZE],
+        dct_blocks = dctn(blocks, axes=(1, 2), norm="ortho")
-            dtype=np.float64,
-            copy=True,
-            order="C",
-        )
-        dct_block = _safe_dct2(block)
 
-        for pos in DEFAULT_EMBED_POSITIONS:
+        # Extract bits from embed positions (vectorized)
-            bit = _extract_bit_from_coeff(float(dct_block[pos[0], pos[1]]))
+        # Shape: (batch_count, num_positions)
-            all_bits.append(bit)
+        coeffs = dct_blocks[:, embed_rows, embed_cols]
 
-        del block, dct_block
+        # Quantize and extract bits (vectorized)
+        quantized = np.round(coeffs / QUANT_STEP).astype(int)
+        bits = (quantized % 2).flatten().tolist()
+        all_bits.extend(bits)
 
+        del blocks, dct_blocks, coeffs, quantized
+        block_idx = batch_end
+
+        # Report progress (scale to 25-70% range, RS decode gets 70-100%)
+        # Starts at 25% because decode.py writes 25% before calling extraction
+        if progress_file and block_idx % PROGRESS_INTERVAL < BATCH_SIZE:
+            extract_pct = 25 + int(45 * block_idx / num_blocks)
+            _write_progress(progress_file, extract_pct, 100, "extracting")
+
+        # Check if we have enough bits (early exit)
         if len(all_bits) >= HEADER_SIZE * 8:
             try:
                 _, flags, data_length = _parse_header(all_bits[: HEADER_SIZE * 8])
@@ -1188,6 +1284,9 @@ def _extract_scipy_dct_safe(stego_image: bytes, seed: bytes) -> bytes:
     del padded
     gc.collect()
 
+    # Extraction done, RS decode starts at 70%
+    _write_progress(progress_file, 70, 100, "decoding")
+
     # Try RS-protected format first (has 24-byte length prefix: 3 copies of 8-byte header)
     if HAS_REEDSOLO and len(all_bits) >= RS_LENGTH_PREFIX_SIZE * 8:
         # Extract length prefix (24 bytes: 3 copies of 8-byte header for majority voting)
@@ -1240,10 +1339,16 @@ def _extract_scipy_dct_safe(stego_image: bytes, seed: bytes) -> bytes:
                     ]
                 )
 
+                # 75% - bits converted, starting RS decode (slow part)
+                _write_progress(progress_file, 75, 100, "decoding")
+
                 try:
                     # RS decode to get header + data
                     raw_payload = _rs_decode(rs_encoded)
 
+                    # 95% - RS decode done
+                    _write_progress(progress_file, 95, 100, "decoding")
+
                     # Parse header from decoded payload
                     _, flags, data_length = _parse_header(
                         [((raw_payload[i // 8] >> (7 - i % 8)) & 1) for i in range(HEADER_SIZE * 8)]
@@ -1251,6 +1356,7 @@ def _extract_scipy_dct_safe(stego_image: bytes, seed: bytes) -> bytes:
 
                     # Extract data
                     data = raw_payload[HEADER_SIZE : HEADER_SIZE + data_length]
+                    _write_progress(progress_file, 100, 100, "complete")
                     return data
                 except (ValueError, struct.error):
                     pass  # Fall through to legacy format
@@ -1266,13 +1372,20 @@ def _extract_scipy_dct_safe(stego_image: bytes, seed: bytes) -> bytes:
         ]
     )
 
+    _write_progress(progress_file, 100, 100, "complete")
     return data
 
 
-def _extract_jpegio(stego_image: bytes, seed: bytes) -> bytes:
+def _extract_jpegio(
+    stego_image: bytes,
+    seed: bytes,
+    progress_file: str | None = None,
+) -> bytes:
     """Extract using jpegio for JPEG images."""
     import os
 
+    # Progress starts at 25% (decode.py writes 20% for Argon2, 25% before extraction)
+
     # Normalize JPEG to avoid crashes with quality=100 images
     # (shouldn't happen with stego images, but be defensive)
     stego_image = _normalize_jpeg_for_jpegio(stego_image)
@@ -1280,12 +1393,14 @@ def _extract_jpegio(stego_image: bytes, seed: bytes) -> bytes:
     temp_path = _jpegio_bytes_to_file(stego_image, suffix=".jpg")
 
     try:
-        jpeg = jio.read(temp_path)
+        jpeg = jpeglib.to_jpegio(jpeglib.read_dct(temp_path))
         coef_array = jpeg.coef_arrays[JPEGIO_EMBED_CHANNEL]
 
         all_positions = _jpegio_get_usable_positions(coef_array)
         order = _jpegio_generate_order(len(all_positions), seed)
 
+        _write_progress(progress_file, 30, 100, "extracting")
+
         # Try RS-protected format first (has 24-byte length prefix: 3 copies for majority voting)
         if HAS_REEDSOLO and len(all_positions) >= RS_LENGTH_PREFIX_SIZE * 8:
             # Extract length prefix (24 bytes: 3 copies of 8-byte header)
@@ -1349,9 +1464,12 @@ def _extract_jpegio(stego_image: bytes, seed: bytes) -> bytes:
                     )
 
                     try:
+                        _write_progress(progress_file, 75, 100, "decoding")
                         raw_payload = _rs_decode(rs_encoded)
+                        _write_progress(progress_file, 95, 100, "decoding")
                         _, flags, data_length = _jpegio_parse_header(raw_payload[:HEADER_SIZE])
                         data = raw_payload[HEADER_SIZE : HEADER_SIZE + data_length]
+                        _write_progress(progress_file, 100, 100, "complete")
                         return data
                     except (ValueError, struct.error):
                         pass  # Fall through to legacy format
@@ -1389,6 +1507,7 @@ def _extract_jpegio(stego_image: bytes, seed: bytes) -> bytes:
             ]
         )
 
+        _write_progress(progress_file, 100, 100, "complete")
         return data
 
     finally:

src/stegasoo/decode.py

@@ -8,6 +8,7 @@ Changes in v4.0.0:
 - Improved error messages for channel key mismatches
 """
 
+import json
 from pathlib import Path
 
 from .constants import EMBED_MODE_AUTO
@@ -24,6 +25,22 @@ from .validation import (
 )
 
 
+def _write_progress(progress_file: str | None, current: int, total: int, phase: str) -> None:
+    """Write progress to file for UI polling."""
+    if progress_file is None:
+        return
+    try:
+        with open(progress_file, "w") as f:
+            json.dump({
+                "current": current,
+                "total": total,
+                "percent": (current / total * 100) if total > 0 else 0,
+                "phase": phase,
+            }, f)
+    except OSError:
+        pass
+
+
 def decode(
     stego_image: bytes,
     reference_photo: bytes,
@@ -33,6 +50,7 @@ def decode(
     rsa_password: str | None = None,
     embed_mode: str = EMBED_MODE_AUTO,
     channel_key: str | bool | None = None,
+    progress_file: str | None = None,
 ) -> DecodeResult:
     """
     Decode a message or file from a stego image.
@@ -45,6 +63,7 @@ def decode(
         rsa_key_data: Optional RSA key bytes (if used during encoding)
         rsa_password: Optional RSA key password
         embed_mode: 'auto' (default), 'lsb', or 'dct'
+        progress_file: Optional path to write progress JSON for UI polling
         channel_key: Channel key for deployment/group isolation:
             - None or "auto": Use server's configured key
             - str: Use this specific channel key
@@ -91,16 +110,23 @@ def decode(
     if rsa_key_data:
         require_valid_rsa_key(rsa_key_data, rsa_password)
 
+    # Progress: starting key derivation (Argon2 - slow on Pi)
+    _write_progress(progress_file, 20, 100, "initializing")
+
     # Derive pixel/coefficient selection key (with channel key)
     from .crypto import derive_pixel_key
 
     pixel_key = derive_pixel_key(reference_photo, passphrase, pin, rsa_key_data, channel_key)
 
+    # Progress: key derivation done, starting extraction
+    _write_progress(progress_file, 25, 100, "extracting")
+
     # Extract encrypted data
     encrypted = extract_from_image(
         stego_image,
         pixel_key,
         embed_mode=embed_mode,
+        progress_file=progress_file,
     )
 
     if not encrypted:
@@ -126,6 +152,7 @@ def decode_file(
     rsa_password: str | None = None,
     embed_mode: str = EMBED_MODE_AUTO,
     channel_key: str | bool | None = None,
+    progress_file: str | None = None,
 ) -> Path:
     """
     Decode a file from a stego image and save it.
@@ -140,6 +167,7 @@ def decode_file(
         rsa_password: Optional RSA key password
         embed_mode: 'auto', 'lsb', or 'dct'
         channel_key: Channel key parameter (see decode())
+        progress_file: Optional path to write progress JSON for UI polling
 
     Returns:
         Path where file was saved
@@ -156,6 +184,7 @@ def decode_file(
         rsa_password,
         embed_mode,
         channel_key,
+        progress_file,
     )
 
     if not result.is_file:
@@ -184,6 +213,7 @@ def decode_text(
     rsa_password: str | None = None,
     embed_mode: str = EMBED_MODE_AUTO,
     channel_key: str | bool | None = None,
+    progress_file: str | None = None,
 ) -> str:
     """
     Decode a text message from a stego image.
@@ -199,6 +229,7 @@ def decode_text(
         rsa_password: Optional RSA key password
         embed_mode: 'auto', 'lsb', or 'dct'
         channel_key: Channel key parameter (see decode())
+        progress_file: Optional path to write progress JSON for UI polling
 
     Returns:
         Decoded message string
@@ -215,6 +246,7 @@ def decode_text(
         rsa_password,
         embed_mode,
         channel_key,
+        progress_file,
     )
 
     if result.is_file:

src/stegasoo/generate.py

@@ -82,7 +82,7 @@ def generate_rsa_key(bits: int = DEFAULT_RSA_BITS, password: str | None = None)
     Generate an RSA private key in PEM format.
 
     Args:
-        bits: Key size (2048, 3072, or 4096, default 2048)
+        bits: Key size (2048 or 3072, default 2048)
         password: Optional password to encrypt the key
 
     Returns:

src/stegasoo/keygen.py

@@ -136,7 +136,7 @@ def generate_rsa_key(bits: int = DEFAULT_RSA_BITS) -> rsa.RSAPrivateKey:
     Generate an RSA private key.
 
     Args:
-        bits: Key size (2048, 3072, or 4096)
+        bits: Key size (2048 or 3072)
 
     Returns:
         RSA private key object

src/stegasoo/qr_utils.py

@@ -8,6 +8,7 @@ IMPROVEMENTS IN THIS VERSION:
 - Much more robust PEM normalization
 - Better handling of QR code extraction edge cases
 - Improved error messages
+- v4.2.0: Added zstd compression (better ratio than zlib)
 """
 
 import base64
@@ -16,6 +17,14 @@ import zlib
 
 from PIL import Image
 
+# Optional ZSTD support (better compression ratio)
+try:
+    import zstandard as zstd
+
+    HAS_ZSTD = True
+except ImportError:
+    HAS_ZSTD = False
+
 # QR code generation
 try:
     import qrcode
@@ -42,30 +51,46 @@ from .constants import (
 )
 
 # Constants
-COMPRESSION_PREFIX = "STEGASOO-Z:"
+COMPRESSION_PREFIX_ZLIB = "STEGASOO-Z:"  # Legacy zlib compression
+COMPRESSION_PREFIX_ZSTD = "STEGASOO-ZS:"  # v4.2.0: New zstd compression (better ratio)
+COMPRESSION_PREFIX = COMPRESSION_PREFIX_ZSTD if HAS_ZSTD else COMPRESSION_PREFIX_ZLIB
 
 
 def compress_data(data: str) -> str:
     """
     Compress string data for QR code storage.
 
+    Uses zstd if available (better ratio), falls back to zlib.
+
     Args:
         data: String to compress
 
     Returns:
-        Compressed string with STEGASOO-Z: prefix
+        Compressed string with STEGASOO-ZS: (zstd) or STEGASOO-Z: (zlib) prefix
     """
-    compressed = zlib.compress(data.encode("utf-8"), level=9)
+    data_bytes = data.encode("utf-8")
+
+    if HAS_ZSTD:
+        # Use zstd (better compression ratio)
+        cctx = zstd.ZstdCompressor(level=19)
+        compressed = cctx.compress(data_bytes)
         encoded = base64.b64encode(compressed).decode("ascii")
-    return COMPRESSION_PREFIX + encoded
+        return COMPRESSION_PREFIX_ZSTD + encoded
+    else:
+        # Fall back to zlib
+        compressed = zlib.compress(data_bytes, level=9)
+        encoded = base64.b64encode(compressed).decode("ascii")
+        return COMPRESSION_PREFIX_ZLIB + encoded
 
 
 def decompress_data(data: str) -> str:
     """
     Decompress data from QR code.
 
+    Supports both zstd (STEGASOO-ZS:) and zlib (STEGASOO-Z:) formats.
+
     Args:
-        data: Compressed string with STEGASOO-Z: prefix
+        data: Compressed string with STEGASOO-ZS: or STEGASOO-Z: prefix
 
     Returns:
         Original uncompressed string
@@ -73,13 +98,27 @@ def decompress_data(data: str) -> str:
     Raises:
         ValueError: If data is not valid compressed format
     """
-    if not data.startswith(COMPRESSION_PREFIX):
+    if data.startswith(COMPRESSION_PREFIX_ZSTD):
-        raise ValueError("Data is not in compressed format")
+        # v4.2.0: ZSTD compression
+        if not HAS_ZSTD:
+            raise ValueError(
+                "Data compressed with zstd but zstandard package not installed. "
+                "Run: pip install zstandard"
+            )
+        encoded = data[len(COMPRESSION_PREFIX_ZSTD):]
+        compressed = base64.b64decode(encoded)
+        dctx = zstd.ZstdDecompressor()
+        return dctx.decompress(compressed).decode("utf-8")
 
-    encoded = data[len(COMPRESSION_PREFIX) :]
+    elif data.startswith(COMPRESSION_PREFIX_ZLIB):
+        # Legacy zlib compression
+        encoded = data[len(COMPRESSION_PREFIX_ZLIB):]
         compressed = base64.b64decode(encoded)
         return zlib.decompress(compressed).decode("utf-8")
 
+    else:
+        raise ValueError("Data is not in compressed format")
+
 
 def normalize_pem(pem_data: str) -> str:
     """
@@ -166,8 +205,8 @@ def normalize_pem(pem_data: str) -> str:
 
 
 def is_compressed(data: str) -> bool:
-    """Check if data has compression prefix."""
+    """Check if data has compression prefix (zstd or zlib)."""
-    return data.startswith(COMPRESSION_PREFIX)
+    return data.startswith(COMPRESSION_PREFIX_ZSTD) or data.startswith(COMPRESSION_PREFIX_ZLIB)
 
 
 def auto_decompress(data: str) -> str:
@@ -213,17 +252,23 @@ def needs_compression(data: str) -> bool:
     return not can_fit_in_qr(data, compress=False) and can_fit_in_qr(data, compress=True)
 
 
-def generate_qr_code(data: str, compress: bool = False, error_correction=None) -> bytes:
+def generate_qr_code(
+    data: str,
+    compress: bool = False,
+    error_correction=None,
+    output_format: str = "png",
+) -> bytes:
     """
-    Generate a QR code PNG from string data.
+    Generate a QR code image from string data.
 
     Args:
         data: String data to encode
         compress: Whether to compress data first
         error_correction: QR error correction level (default: auto)
+        output_format: Image format - 'png' or 'jpg'/'jpeg'
 
     Returns:
-        PNG image bytes
+        Image bytes in requested format
 
     Raises:
         RuntimeError: If qrcode library not available
@@ -260,11 +305,79 @@ def generate_qr_code(data: str, compress: bool = False, error_correction=None) -
     img = qr.make_image(fill_color="black", back_color="white")
 
     buf = io.BytesIO()
+    fmt = output_format.lower()
+    if fmt in ("jpg", "jpeg"):
+        # Convert to RGB for JPEG (no alpha channel)
+        img = img.convert("RGB")
+        img.save(buf, format="JPEG", quality=95)
+    else:
         img.save(buf, format="PNG")
     buf.seek(0)
     return buf.getvalue()
 
 
+def generate_qr_ascii(
+    data: str,
+    compress: bool = False,
+    invert: bool = False,
+) -> str:
+    """
+    Generate an ASCII representation of a QR code.
+
+    Uses Unicode block characters for compact display.
+
+    Args:
+        data: String data to encode
+        compress: Whether to compress data first
+        invert: Invert colors (white on black for dark terminals)
+
+    Returns:
+        ASCII string representation of QR code
+
+    Raises:
+        RuntimeError: If qrcode library not available
+        ValueError: If data too large for QR code
+    """
+    if not HAS_QRCODE_WRITE:
+        raise RuntimeError("qrcode library not installed. Run: pip install qrcode[pil]")
+
+    qr_data = data
+
+    # Compress if requested
+    if compress:
+        qr_data = compress_data(data)
+
+    # Check size
+    if len(qr_data.encode("utf-8")) > QR_MAX_BINARY:
+        raise ValueError(
+            f"Data too large for QR code ({len(qr_data)} bytes). " f"Maximum: {QR_MAX_BINARY} bytes"
+        )
+
+    qr = qrcode.QRCode(
+        version=None,
+        error_correction=ERROR_CORRECT_L,
+        box_size=1,
+        border=2,
+    )
+    qr.add_data(qr_data)
+    qr.make(fit=True)
+
+    # Get the QR matrix
+    # Use print_ascii to a StringIO to capture output
+    import sys
+    from io import StringIO
+
+    old_stdout = sys.stdout
+    sys.stdout = StringIO()
+    try:
+        qr.print_ascii(invert=invert)
+        ascii_qr = sys.stdout.getvalue()
+    finally:
+        sys.stdout = old_stdout
+
+    return ascii_qr
+
+
 def read_qr_code(image_data: bytes) -> str | None:
     """
     Read QR code from image data.

src/stegasoo/steganography.py

@@ -156,7 +156,7 @@ def has_dct_support() -> bool:
         dct_mod = _get_dct_module()
         return dct_mod.has_dct_support()
     except (ImportError, ValueError):
-        # ValueError: numpy binary incompatibility (e.g., jpegio built against numpy 2.x)
+        # ValueError: numpy binary incompatibility (e.g., jpeglib built against numpy 2.x)
         return False
 
 
@@ -746,6 +746,10 @@ def _embed_lsb(
         modified_pixels = 0
         total_pixels_to_process = len(selected_indices)
 
+        # Initial progress write - signals prep is done, embedding starting
+        if progress_file:
+            _write_progress(progress_file, 5, 100, "embedding")
+
         for progress_idx, pixel_idx in enumerate(selected_indices):
             if bit_idx >= len(binary_data):
                 break
@@ -839,6 +843,7 @@ def extract_from_image(
     pixel_key: bytes,
     bits_per_channel: int = 1,
     embed_mode: str = EMBED_MODE_AUTO,
+    progress_file: str | None = None,
 ) -> bytes | None:
     """
     Extract hidden data from a stego image.
@@ -848,6 +853,7 @@ def extract_from_image(
         pixel_key: Key for pixel/coefficient selection (must match encoding)
         bits_per_channel: Bits per channel (LSB mode only)
         embed_mode: 'auto' (try both), 'lsb', or 'dct'
+        progress_file: Optional path to write progress JSON for UI polling
 
     Returns:
         Extracted data bytes, or None if extraction fails
@@ -863,7 +869,7 @@ def extract_from_image(
 
         if has_dct_support():
             debug.print("Auto-detect: LSB failed, trying DCT")
-            result = _extract_dct(image_data, pixel_key)
+            result = _extract_dct(image_data, pixel_key, progress_file)
             if result is not None:
                 debug.print("Auto-detect: DCT extraction succeeded")
                 return result
@@ -875,18 +881,22 @@ def extract_from_image(
     elif embed_mode == EMBED_MODE_DCT:
         if not has_dct_support():
             raise ImportError("scipy required for DCT mode")
-        return _extract_dct(image_data, pixel_key)
+        return _extract_dct(image_data, pixel_key, progress_file)
 
     # EXPLICIT LSB MODE
     else:
         return _extract_lsb(image_data, pixel_key, bits_per_channel)
 
 
-def _extract_dct(image_data: bytes, pixel_key: bytes) -> bytes | None:
+def _extract_dct(
+    image_data: bytes,
+    pixel_key: bytes,
+    progress_file: str | None = None,
+) -> bytes | None:
     """Extract using DCT mode."""
     try:
         dct_mod = _get_dct_module()
-        return dct_mod.extract_from_dct(image_data, pixel_key)
+        return dct_mod.extract_from_dct(image_data, pixel_key, progress_file)
     except Exception as e:
         debug.print(f"DCT extraction failed: {e}")
         return None
@@ -1087,7 +1097,7 @@ def peek_image(image_data: bytes) -> dict:
     except Exception:
         pass
 
-    # Try DCT extraction (requires scipy/jpegio)
+    # Try DCT extraction (requires scipy/jpeglib)
     try:
         from .dct_steganography import HAS_JPEGIO, HAS_SCIPY
 

src/stegasoo/utils.py

@@ -66,9 +66,15 @@ def read_image_exif(image_data: bytes) -> dict:
                 # Convert bytes to string if possible
                 elif isinstance(value, bytes):
                     try:
-                        result[tag] = value.decode("utf-8", errors="replace").strip("\x00")
+                        # Try to decode as ASCII/UTF-8 text
-                    except Exception:
+                        decoded = value.decode("utf-8", errors="strict").strip("\x00")
-                        result[tag] = f"<{len(value)} bytes>"
+                        # Only keep if it looks like printable text
+                        if decoded.isprintable() or all(c.isspace() or c.isprintable() for c in decoded):
+                            result[tag] = decoded
+                        else:
+                            result[tag] = f"<{len(value)} bytes binary>"
+                    except (UnicodeDecodeError, Exception):
+                        result[tag] = f"<{len(value)} bytes binary>"
                 # Handle tuples of IFDRational
                 elif isinstance(value, tuple) and value and hasattr(value[0], "numerator"):
                     result[tag] = [float(v) for v in value]

test_data/phonebooth.pem

@@ -0,0 +1,30 @@
+-----BEGIN ENCRYPTED PRIVATE KEY-----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+-----END ENCRYPTED PRIVATE KEY-----