sudoping01/adia-inference-server

Inference server of wolof tts Adia # Adia_TTS Wolof API ## Description Adia_TTS Wolof API is production ready inference server that provides Text-to-Speech capabilities for the Wolof language. This project solves the accessibility challenges faced by Wolof speakers in digital environments by offering a reliable speech synthesis solution. The API works with Adia_TTS model and implements text segmentation strategies to overcome the model's 200-character limitation, making it possible to process texts of any length with natural-sounding results. ## Table of Contents - Features - Model Information - Installation - Usage - API Endpoints - Technical Details - Limitations - License - Citation ## Features * Wolof speech synthesis with natural voice quality * Voice style customization through text descriptions * Automatic text segmentation for long inputs * Audio concatenation with smooth transitions * Batch processing for documents * GPU acceleration support * FastAPI-based HTTP endpoints ## Installation ### Prerequisites * Docker * NVIDIA GPU with CUDA support (strongly recommended) * Hugging Face API token (Optional) ### Step-by-Step Installation 1. Clone this repository: ```bash git clone github.com cd adia-tts-inference-server ``` 2. Build the Docker image: ```bash docker build -t adia-tts-wolof . ``` 3. Run the container with your Hugging Face token: ```bash docker run --gpus all -p 8080:8080 -e HF_TOKEN=your_hf_token -d adia-tts-wolof ``` 4. Verify the installation: ```bash curl localhost ``` ## Usage ### Basic Speech Synthesis Send a POST request to the `/predict` endpoint with Wolof text: ```bash curl -X POST localhost \ -H "Content-Type: application/json" \ -d '{ "text": "Entreprenariat ci Senegal dafa am solo lool ci yokkuteg koom-koom, di gëna yokk liggéey ak indi gis-gis yu bees ci dëkk bi." }' \ --output speech.wav ``` ### Voice Style Customization You can customize the voice style using the d …