TheDeathDragon/LiveTranslate
Real-time audio translation, captures system audio + mic, runs ASR (Whisper/SenseVoice), translates via LLM API with streaming display. Perfect for VTubers, livestreamers, and watching foreign content. Windows 实时音频翻译，ASR 语音识别后 LLM 流式翻译显示，适合 VTuber、主播和外语视频观看。
What it solves
LiveTranslate provides real-time audio translation for Windows users, allowing them to translate system audio (such as videos, livestreams, and voice chats) and microphone input on the fly without needing to modify the original media player.
How it works
The application captures system audio via WASAPI loopback and optional microphone input. It then processes the audio through a pipeline consisting of Voice Activity Detection (VAD) using Silero, Automatic Speech Recognition (ASR) via engines like faster-whisper or SenseVoice, and finally translation through any OpenAI-compatible LLM API. The resulting translation is displayed as a transparent, always-on-top overlay on the screen.
Who it’s for
It is designed for Windows users who need real-time subtitles for foreign language audio coming from any source on their computer, including gamers, viewers of international livestreams, and people in voice chats.
Highlights
- Versatile ASR Support: Compatible with multiple engines including faster-whisper, SenseVoice, FunASR Nano, and Anime-Whisper.
- Flexible Translation: Works with any OpenAI-compatible API (e.g., DeepSeek, GPT, Ollama, vLLM).
- Low Latency: Uses 32ms audio chunks and adaptive silence detection for fast processing.
- Customizable Overlay: Features a click-through, draggable transparent overlay with 14 different color themes.
- Hardware Acceleration: Supports CUDA for GPU-accelerated ASR inference.
- Remote Offloading: Ability to offload ASR tasks to a separate GPU machine via HTTP.
Related
- Project
- Project
- Project
- Project
- Project