janvarev/Irene-Voice-Assistant

Ирина - русский голосовой ассистент для работы оффлайн. Поддерживает скиллы через плагины.

What it solves

Irene is a Russian-language voice assistant designed to work primarily offline, providing a private and local alternative to cloud-based assistants. It allows users to control their computer, set timers, and interact with various services through voice commands without requiring a constant internet connection.

How it works

The system uses a modular architecture based on the Jaa.py engine to support plugins (skills). It processes voice input through several supported Speech-to-Text (STT) engines, including Vosk (with a high-performance streaming model in version 13.0), GigaSTT, and SpeechRecognition. For output, it utilizes Text-to-Speech (TTS) engines like pyttsx3.

Starting with version 12.0, it supports "AI plugins" that use LLM tool-calling (via OpenAI-compatible endpoints like VseGPT.ru, Ollama, or LMStudio) to interpret free-form natural language commands and trigger specific functions.

Who it’s for

  • Users who want a local, offline Russian voice assistant for privacy or reliability.
  • Developers looking to build custom voice-controlled skills using a plugin system.
  • Smart home enthusiasts who want to integrate voice control with Home Assistant.

Highlights

  • Offline First: Core functionality and STT/TTS can run entirely on the local machine.
  • Extensible Plugin System: Easy to add new skills via the Jaa.py engine.
  • LLM Integration: Supports tool-calling for natural language command processing.
  • High-Accuracy STT: Integration with Vosk 0.56, which provides low-latency streaming and high accuracy.
  • Multi-platform Support: Installation guides and Docker images available for Windows, Linux, macOS, and ARM devices (Raspberry Pi).

Related

  • Project
    smixs/pohuy

    A collection of plugins and prompt styles that enable AI agents to communicate using idiomatic, profane Russian engineering slang for more direct and concise technical communication.

  • Project
    ShayneP/local-voice-ai

    A private, low-latency voice assistant that runs locally on your hardware using a stack of speech recognition, LLM, and speech generation models.

  • Project
    snakers4/silero-models

    A collection of fast, end-to-end text-to-speech models supporting a wide variety of languages, designed for efficient execution on CPU and GPU.

  • Project
    kyutai-labs/unmute

    A system that enables text LLMs to listen and speak by wrapping them in low-latency speech-to-text and text-to-speech models.

  • Project
    Natively-AI-assistant/natively-cluely-ai-assistant

    An AI interview copilot and meeting assistant that provides real-time, undetectable transcription and guidance using local STT and flexible LLM integration.