mirror of
https://github.com/Dokploy/templates.git
synced 2026-07-08 15:35:24 +02:00
0aad2a0231
* feat: Add Kokoro TTS FastAPI template (#353) - Add CPU-optimized docker-compose.yml with source build - Add GPU-optimized docker-compose-gpu.yml for NVIDIA support - Add comprehensive template.toml with OpenAI-compatible API docs - Add kokoro-tts.svg logo and meta.json entry - Support streaming audio, timestamps, and multi-language TTS - Resolves #353 * updated the meta.json for the build errors * removed the docker-compose-gpu.yml file * Update docker-compose.yml --------- Co-authored-by: Mauricio Siu <47042324+Siumauricio@users.noreply.github.com>
73 lines
2.0 KiB
TOML
73 lines
2.0 KiB
TOML
[variables]
|
|
main_domain = "${domain}"
|
|
device = "cpu"
|
|
|
|
[config]
|
|
env = [
|
|
"DEVICE=${device}",
|
|
"MODEL_PATH=/app/models",
|
|
"HOST=0.0.0.0",
|
|
"PORT=8880",
|
|
"PYTHONUNBUFFERED=1",
|
|
"UV_SYSTEM_PYTHON=1"
|
|
]
|
|
|
|
[[config.domains]]
|
|
serviceName = "kokoro-tts"
|
|
port = 8880
|
|
host = "${main_domain}"
|
|
path = "/"
|
|
|
|
[[config.mounts]]
|
|
filePath = "README.md"
|
|
content = """# Kokoro TTS FastAPI
|
|
|
|
This template provides a Dockerized FastAPI wrapper for the Kokoro-82M text-to-speech model.
|
|
|
|
## Features
|
|
|
|
- Multi-language support (English, Japanese, Chinese)
|
|
- OpenAI-compatible speech endpoint
|
|
- CPU and GPU support
|
|
- Web interface for monitoring
|
|
- RESTful API with comprehensive documentation
|
|
- Streaming audio generation
|
|
- Word-level timestamps and phonemes
|
|
|
|
## Usage
|
|
|
|
- **Web Interface**: Access the web UI at `https://${main_domain}/web`
|
|
- **API Documentation**: Available at `https://${main_domain}/docs`
|
|
- **Health Check**: Monitor service health at `https://${main_domain}/health`
|
|
|
|
## Configuration
|
|
|
|
The service runs on port 8880 and supports both CPU and GPU inference.
|
|
For GPU support, ensure your Dokploy instance has NVIDIA GPU support enabled.
|
|
|
|
## API Endpoints
|
|
|
|
- `POST /v1/audio/speech` - Generate speech from text (OpenAI compatible)
|
|
- `POST /dev/captioned_speech` - Generate speech with timestamps
|
|
- `POST /dev/phonemize` - Convert text to phonemes
|
|
- `POST /dev/generate_from_phonemes` - Generate audio from phonemes
|
|
- `GET /health` - Health check endpoint
|
|
- `GET /docs` - Interactive API documentation
|
|
- `GET /web` - Web interface
|
|
|
|
## Model Information
|
|
|
|
- Model: Kokoro-82M
|
|
- License: Apache-2.0
|
|
- Repository: https://github.com/remsky/Kokoro-FastAPI
|
|
- HuggingFace Model: https://huggingface.co/hexgrad/Kokoro-82M
|
|
|
|
## Notes
|
|
|
|
- This template builds the image from source during deployment
|
|
- Uses CPU-optimized Dockerfile by default
|
|
- Initial build may take several minutes due to model download
|
|
- Ensure sufficient disk space for model storage
|
|
- For GPU support, change dockerfile path to `docker/gpu/Dockerfile` in docker-compose.yml
|
|
"""
|