A simple FastAPI Server to run Silero TTS

Credit goes to the developers of Silero TTS
Silero PyTorch Page
Silero GitHub Page

This is primarily to serve the TTS extension in SillyTavern. The TTS module or server can be used any way you wish.

Installation

pip install silero-api-server

Starting Server

python -m silero_api_server will run on default ip and port (0.0.0.0:8001)

usage: silero_api_server [-h] [-o HOST] [-p PORT]

Run Silero within a FastAPI application

options:
  -h, --help            show this help message and exit
  -o HOST, --host HOST
  -p PORT, --port PORT
  -l LANG, --language LANG
  --show-languages

On first run of server, two operations occur automatically. These may take a minute or two.

1. The model will be downloaded
2. Voice samples will be generated.

Using v4 Models

This server supports both v3 and v4 Silero TTS models. To use v4 models, specify the model name in the language parameter:

# Russian v4 model
python -m silero_api_server --language "v4_ru.pt"

# Ukrainian v4 model  
python -m silero_api_server --language "v4_ua.pt"

# Uzbek v4 model
python -m silero_api_server --language "v4_uz.pt"

# Indic v4 model
python -m silero_api_server --language "v4_indic.pt"

# Cyrillic v4 model
python -m silero_api_server --language "v4_cyrillic.pt"

API Docs

API Docs can be accessed from http://localhost:8001/docs

Voice Samples

Samples are served statically by the web server at /samples/{speaker}.wav or callable from the API from /tts/sample?speaker={speaker} endpoint.

Selecting Language

Use command-line options or download and set the desired language using POST /tts/language with payload {"id":"languageId"}
List of language ids are available via GET /tts/language