Whisper Speech-to-Text API Service

Overview

This project is a lightweight, OpenAI-compatible API server for transcribing audio to text using the Whisper model. It's designed to run locally, making it easy to set up and use for speech-to-text tasks.

Features

OpenAI API Compatibility: Fully compatible with OpenAI's /v1/audio/transcriptions and /v1/models endpoints.
Local File Support: Transcribe audio files stored locally on your machine.
Multiple Input Methods: Supports:
- Local file paths
- Files accessible via URL
- Base64-encoded audio
- Multipart form uploads
Easy Setup: Designed to run as a local service with minimal configuration.
Hardware Optimization: Utilizes GPU (CUDA, MPS) or CPU for efficient processing.
Health Check: Includes a /health endpoint for service monitoring.

Recommended Model

For Russian language transcription, we recommend using the whisper-large-v3-russian model from Hugging Face. This model is fine-tuned specifically for Russian speech recognition and delivers high accuracy. For faster transcription with slightly lower accuracy, consider the whisper-large-v3-turbo-russian model, which is optimized for speed.

Perfect for local development or offline use cases where OpenAI's API isn't accessible.

Quick Start

Edit the Configuration File (config.json):

Set the path to your Whisper model (model_path).
Configure other parameters like language (language), chunk size (chunk_length_s), batch size (batch_size), and audio normalization settings.

{
    "service_port": 5042,
    "model_path": "/path/to/your/whisper-model",
    "language": "english",
    "chunk_length_s": 30,
    "batch_size": 16,
    "max_new_tokens": 256,
    "return_timestamps": false,
    "norm_level": "-0.5",
    "compand_params": "0.3,1 -90,-90,-70,-70,-60,-20,0,0 -5 0 0.2"
}

Run the Server:
- Simply execute the server.sh script:
```
./server.sh
```
- If you need to update the environment, use:
```
./server.sh --update
```

Use the API:

Once the server is running, you can send transcription requests.
Example request (curl):

curl -X POST -F file=@audio.mp3 http://localhost:5042/v1/audio/transcriptions | jq -r '.text'

Enjoy seamless audio-to-text transcription with your local Whisper API server!