Rest API server for Silero TTS

Server for running Silero TTS models (or other compatible models) with OpenTTS-like API. This server allows users to generate speech using different models and provides an easy-to-use REST API.

Features

• Generate speech from text using multiple TTS models.
• Support for custom model uploads.
• OpenAPI documentation available.
• Full OpenTTS API compatibility

Supported languages:

• 🇩🇪 German (de)
• 🇬🇧 English (en)
• 🇪🇸 Spanish (es)
• 🇫🇷 French (fr)
• 🇮🇳 Indic scripts (indic)
• 🇷🇺 Russian (ru)
• 🇷🇺 Tatar (tt)
• 🇺🇦 Ukrainian (ua)
• 🇺🇿 Uzbek (uz)
• 🇷🇺 Kalmyk (xal)
• 🌐 Other cyrillic languages

Dependencies

• FastAPI
• swagger-ui
• pydub
• PyTorch

API endpoints

/api/tts
Generate speach.
If you use not local model, it will be downloaded automatically.
/api/languages
Get available languages list (From only local models!)
/api/voices
Get available speakers list (From only local models!)
/upload_model/
Upload custom model. Admin auth required!
/remove_file_cache/
Remove cached files. Admin auth required!
/openapi
OpenAPI docs

For authorization used key in HTTP headers: api_key: <64-byte token>

More information you can find at /openapi or swagger.yaml

Licenses

AGPLv3

To use Silero models you need to accept Silero's license.

Running in container

Building image:

podman build -t silero-rest-api-server:latest .

Container setup

From local image:

podman run -d --name silero-rest-api-server-container -p 5500:5500 -e SILERO_LICENCE=ACCEPTED silero-rest-api-server

From repository:

podman run -d --name silero-rest-api-server-container -p 5500:5500 -e SILERO_LICENCE=ACCEPTED yaroslavk1231/silero-rest-api-server

API KEY

podman logs silero-rest-api-server-container | grep "API KEY"

How to run

To setup

python3 -m venv venv
source venv/bin/activate
pip3 install --no-cache-dir torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
pip3 install --no-cache-dir -r requirements.txt

To run

source venv/bin/activate
SILERO_LICENCE="ACCEPTED" uvicorn main:app --host 0.0.0.0 --port 5500

Usage examples

Getting speakers list example

curl -X 'GET' \
  'http://127.0.0.1:5500/api/voices?language=en&locale=en-indic' \
  -H 'accept: */*'

Getting languages list example

curl -X 'GET' \
  'http://127.0.0.1:5500/api/languages' \
  -H 'accept: */*'

Example of speach generation.

curl -X 'GET' 'http://127.0.0.1:5500/api/tts?voice=v3_en%3Aen_64&text=Hello%20World%21' -H 'accept: */*' --output test.wav

Out: Audio file "test.wav"

Model uploading example.

curl -X POST "http://localhost:5500/upload_model/" -H "Content-Type: multipart/form-data" -H "api_key:<YOUR API KEY>" -F "file=@v3_en.pt"

Config files:

config.json

{
  "max_char_length": 600,  # Maximum size of text block. If text is larger than this value, it will be divided to blocks.
  "sample_rate": 48000,  # Sampling rate: 48000 or 24000 or 8000
  "threads_limit": 6,  # Maximum threads usage for models
  "min_model_version": 3,  # Minimum model version (Using models below version 3 may cause problems)

  "offline_mode": false,  # Prohibit server to download new models from models.silero.ai/models/tts
  "update_voices": true  # Save list of local speakers to speakers.json
}

langs.json

List of official SileroTTS models. Can be updated automatically.

{
    "model_name.pt": "download_url",
    ...
}

Project structure

• main.py - api server and main functions
• admin.py - functions for admin auth
• config.py - functions for work with config.json
• tts.py - function for work with models and pytorch

Used Projects

This project incorporates some code from the silero-api-server(MIT).

This app downloads add runs Silero TTS models provided by CC-BY-NC(or other) license.