The router used in the upcoming new release of the TTS Arena.

• Support for multiple TTS providers (ElevenLabs, PlayHT, Papla, CosyVoice, Hume, Kokoro, Maya Research Veena)
• RESTful API with FastAPI
• Provider and model selection
• Default voice selection

1. Clone the repository
2. Install dependencies:

   pip install -r requirements.txt
   

3. Create a .env file with your API keys:

   ELEVENLABS_API_KEY=your_elevenlabs_api_key
   PLAYHT_API_KEY=your_playht_api_key
   PLAYHT_USER_ID=your_playht_user_id
   PAPLA_API_KEY=your_papla_api_key
   CARTESIA_API_KEY=your_cartesia_api_key
   HF_TOKEN=your_huggingface_token # For Kokoro, CosyVoice spaces
   HUME_API_KEY=your_hume_api_key
   VEENA_API_KEY=your_veena_api_key # For Maya Research Veena TTS
   VOCU_API_KEY=your_vocu_api_key # For Vocu TTS
   PARMESAN_API_KEY=your_parmesan_api_key # For Parmesan TTS
   ZEROGPU_TOKENS=hf_xxx,hf_xxx
   

   (The ZEROGPU_TOKENS variable should be a comma-separated list of Hugging Face tokens for multiple Hugging Face accounts to avoid ZeroGPU rate limiting)

uvicorn app:app --host 0.0.0.0 --port 8000

Or use the Python script:

python app.py

Lists all available TTS providers.

Response:

{
  "providers": ["elevenlabs", "playht"]
}

Lists all available models for a specific provider.

Response:

{
  "models": [
    {
      "id": "eleven_multilingual_v2",
      "name": "Multilingual v2",
      "description": "Multilingual speech generation model"
    },
    ...
  ]
}

Generates TTS audio from text using the specified provider and model.

Request:

{
  "text": "Hello, this is a test.",
  "provider": "elevenlabs",
  "model": "eleven_multilingual_v2"
}

Response:

{
  "status": "success",
  "provider": "elevenlabs",
  "model": "eleven_multilingual_v2",
  "audio_data": "base64_encoded_audio_data",
  "extension": "mp3"
}

To add a new TTS provider:

1. Create a new file in the tts_providers directory, e.g., tts_providers/new_provider.py
2. Implement the TTSProvider interface
3. Register the provider using the @register_provider decorator
4. Add the import to tts_providers/base.py

The API includes robust error handling:

• Provider-specific errors are logged and don't crash the server
• If a provider fails to initialize, it's marked as unavailable but doesn't affect other providers

Unless otherwise specified, the code is licensed under the MIT license.

Note that the code in the spaces directory may be subject to different licenses. For example, the CosyVoice space is licensed under the Apache License 2.0.