Repolode
Submit
snakers4

snakers4/silero-models

Pre-trained text-to-speech models from Silero, loadable via PyTorch Hub or pip, covering Russian and many Turkic, Caucasian and other CIS languages.

View on GitHub
Stars
6,141
Forks
376
Open issues
13
Last commit
7 days ago

What is it?

Silero Models is a collection of pre-trained text-to-speech models that can be loaded with one line of code through torch.hub.load or the silero pip package. Models are downloaded on demand and can also be cached and run standalone with PyTorch and the Python standard library.

The model lineup is organized by version and language group. The V5 line includes Russian models with automatic stress and homograph handling, plus large CIS collections covering languages such as Tatar, Kazakh, Ukrainian, Uzbek, Chuvash, Yakut and several Caucasian languages. Older V3 and V4 models add English, German, Spanish, French and Indic languages. Models support 8000, 24000 and 48000 Hz sampling rates, and V5 models support SSML markup.

The repository includes Colab notebooks with usage examples. Licensing differs by model group: most models are under CC BY-NC 4.0, while the CIS base models are under MIT.

Key features

  • 01Loads through torch.hub or a pip package with a single line of code.
  • 02Covers Russian with automatic stress and homograph handling in V5 models.
  • 03Provides many voices across Turkic, Caucasian and other CIS languages.
  • 04Supports SSML markup in V5 models.
  • 05Outputs audio at 8000, 24000 or 48000 Hz and can run on CPU or GPU.

Use cases

  • Adding Russian or CIS-language speech output to a Python application.
  • Generating voice prompts or audio samples for multilingual projects.
  • Running offline TTS on a local machine with PyTorch and no external API.

Quick start

  1. 1

    Install the silero package from PyPI

    pip install silero
  2. 2

    Import silero_tts, load a model such as v5_ru, and call apply_tts on the example text

Things to know

  • The models are licensed under CC BY-NC 4.0, which is non-commercial only.
  • Models are downloaded on demand, so first use needs an internet connection.

In collections

AI video & audio tools

Topics

#text-to-speech#pytorch#speech-synthesis#pretrained-models#russian#speech#armenian#azerbaijani#belarus#colab

FAQ

What is silero-models?
Pre-trained text-to-speech models from Silero, loadable via PyTorch Hub or pip, covering Russian and many Turkic, Caucasian and other CIS languages. Silero Models is a collection of pre-trained text-to-speech models that can be loaded with one line of code through torch.hub.load or the silero pip package. Models are downloaded on demand and can also be cached and run standalone with PyTorch and the Python standard library.
Is silero-models free to use? What license does it have?
silero-models is open source under the Other license. Check the license terms before using it in commercial products.
Is silero-models still maintained?
Actively maintained. The last commit to silero-models was on Oct 15, 2026.
What can I use silero-models for?
Adding Russian or CIS-language speech output to a Python application. Generating voice prompts or audio samples for multilingual projects. Running offline TTS on a local machine with PyTorch and no external API.
How do I install silero-models?
Install the silero package from PyPI: pip install silero Import silero_tts, load a model such as v5_ru, and call apply_tts on the example text
Where can I get silero-models?
The source code and installation instructions are on GitHub: https://github.com/snakers4/silero-models

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects