Audiobook Maker Engines

Docker images for Audiobook Maker TTS/STT/Text/Audio engines.

Note: For issues, feature requests, and discussions please use the main repository. This repo only contains the engine Docker images.

Available Engines

Production Engines

Engine	Type	GPU	Platforms	Description
chatterbox	TTS	Yes	amd64	Chatterbox multilingual TTS with voice cloning
vibevoice	TTS	Yes	amd64	Microsoft VibeVoice - expressive multi-speaker TTS
xtts	TTS	Yes	amd64	Coqui XTTS v2 - multilingual TTS with voice cloning
whisper	STT	Yes	amd64	OpenAI Whisper - speech recognition (GPU, :latest)
whisper	STT	No	amd64, arm64	OpenAI Whisper - speech recognition (CPU, :cpu)
silero-vad	Audio	No	amd64, arm64	Silero VAD - voice activity detection
spacy	Text	No	amd64, arm64	spaCy NLP - text processing and sentence splitting

Debug/Test Engines

Lightweight mock engines for testing the pipeline without GPU or ML dependencies.

Note: Debug engines are only visible in development builds (npm run dev:tauri). They are automatically hidden from the catalog in frontend production releases.

Engine	Type	Platforms	Description
debug-tts	TTS	amd64, arm64	Generates sine wave tones
debug-stt	STT	amd64, arm64	Returns mock transcription
debug-text	Text	amd64, arm64	Punctuation-based segmentation
debug-audio	Audio	amd64, arm64	Returns mock quality metrics

Usage

Pull an image

docker pull ghcr.io/digijoe79/audiobook-maker-engines/debug-tts:latest
docker pull ghcr.io/digijoe79/audiobook-maker-engines/vibevoice:latest

Run an engine

# CPU engine
docker run -d -p 8766:8766 ghcr.io/digijoe79/audiobook-maker-engines/debug-tts:latest

# GPU engine (requires NVIDIA Container Toolkit)
docker run -d --gpus all -p 8766:8766 \
  -v /path/to/samples:/app/samples \
  -v /path/to/models:/app/external_models \
  ghcr.io/digijoe79/audiobook-maker-engines/vibevoice:latest

Online Catalog

The catalog.yaml file contains metadata about all available engines. It is automatically generated and attached to each release.

Fetch latest catalog:

https://github.com/DigiJoe79/audiobook-maker-engines/releases/latest/download/catalog.yaml

CI/CD Workflows

Workflow	Trigger	Purpose
build-single.yml	Manual	Build and push a single engine image to GHCR
release-catalog.yml	Release	Generate and upload catalog.yaml

Recommended workflow:

1. Build & test image locally
2. Push to GHCR via build-single.yml
3. Pull & verify from GHCR
4. Create release to publish catalog.yaml

Local Development

Subprocess Mode

Run engines as local subprocesses instead of Docker containers:

# Clone this repo into the main audiobook-maker backend
cd /path/to/audiobook-maker
git clone https://github.com/DigiJoe79/audiobook-maker-engines.git backend/engines

# Set up a virtual environment for an engine
cd backend/engines/tts/debug-tts
python -m venv venv
venv/Scripts/pip install -r requirements.txt  # Windows
# or: venv/bin/pip install -r requirements.txt  # Linux/Mac

# Run the engine
venv/Scripts/python server.py --port 8766

Building Images Locally

# From repo root
docker build -t audiobook-maker/debug-tts:latest -f tts/debug-tts/Dockerfile .
docker build -t audiobook-maker/vibevoice:latest -f tts/vibevoice/Dockerfile .

License

MIT License - see LICENSE