This repository serves as a curated collection of resources for Russian speech technology, targeting researchers, developers, and enthusiasts in the field. It aims to consolidate links to datasets, models, courses, and tools for speech recognition, synthesis, voice conversion, and related linguistic tasks, providing a centralized hub for the Russian-speaking speech tech community.

How It Works

The project is a curated list of external links, organized by functional area within speech technology. It aggregates resources such as Telegram channels for community discussion, GitHub repositories for code and models, Hugging Face for pre-trained models, and academic courses for learning. The organization by task (e.g., ASR, TTS, voice conversion, normalization) allows users to quickly find relevant tools and data.

Quick Start & Requirements

This is a curated list of links, not a runnable project. Users will need to navigate to the provided links for specific tools and models. Requirements vary per linked resource, often including Python, specific ML frameworks (PyTorch, TensorFlow), and potentially GPU hardware for model training or inference.

Highlighted Details

• Extensive links to Russian speech datasets, including Common Voice, LibriSpeech, and custom datasets.
• Comprehensive list of Russian ASR models from Vosk, Hugging Face (Whisper, Wav2Vec2), and Salute.
• Resources for Text-to-Speech (TTS) cover various models like HiFi-GAN, FastSpeech, and VITS, with links to pre-trained checkpoints.
• Includes specialized tools for Russian linguistic processing, such as stress assignment, phonetization, and text normalization.

Maintenance & Community

The project is maintained by alphacep and links to several Telegram communities for discussion and news, including groups focused on speech recognition and general speech technology.

Licensing & Compatibility

The repository itself is a list of links and does not have a specific license. The licenses of the linked projects vary, with many projects hosted on GitHub being open-source (e.g., MIT, Apache). Users must consult the licenses of individual linked resources for usage terms.

Limitations & Caveats

As a curated list, the project's quality and currency depend on the linked resources. Some links may become outdated or lead to unmaintained projects. The README contains a significant amount of informal Russian text and Telegram links, which may require translation or context for non-native speakers.