Summary LiveTranslate offers real-time audio translation for Windows, targeting VTubers, livestreamers, and foreign media consumers. It captures system audio and microphone input, processes it via ASR (Whisper, SenseVoice, FunASR, Anime-Whisper), translates using LLM APIs, and displays results in a customizable, transparent overlay.

How It Works The pipeline captures audio (WASAPI loopback + mic), applies low-latency VAD, and feeds it to selected ASR engines. Transcribed text is translated via any OpenAI-compatible LLM API and streamed character-by-character to a transparent, always-on-top overlay.

Quick Start & Requirements

• OS: Windows 10/11
• Python: 3.10+
• GPU: NVIDIA + CUDA 12.6 (or 12.8 for RTX 50xx) recommended for ASR acceleration; CPU fallback available.
• Network: Required for LLM API access.
• Install: Clone repo, run install.bat for automated setup (Python, venv, PyTorch, dependencies). start.bat launches; update.bat updates.
• Manual Install: Requires manual venv, PyTorch (CUDA/CPU), and pip install -r requirements.txt.
• First Launch: Setup wizard for model sources (ModelScope/HuggingFace) and cache path. Auto-downloads VAD/ASR models (~1GB).
• Links: None provided in README.

Highlighted Details

• ASR Options: faster-whisper, SenseVoice, FunASR Nano, Anime-Whisper (Japanese anime/galgame).
• LLM Agnostic: Supports any OpenAI-compatible API (DeepSeek, Grok, Qwen, Ollama, vLLM, etc.).
• Streaming Display: Real-time, character-by-character translation output.
• Overlay: Transparent, always-on-top, click-through, draggable, 14 themes.
• Performance: CUDA acceleration for ASR; includes a benchmark tool.
• Audio Input: WASAPI loopback + optional microphone mix-in.

Maintenance & Community No details on contributors, sponsorships, community channels (Discord/Slack), or roadmap are provided in the README.

Licensing & Compatibility MIT License. No specific commercial use restrictions noted.

Limitations & Caveats Windows-only application. Newer NVIDIA GPUs (RTX 50xx) require CUDA 12.8. README does not specify known bugs or alpha/beta status.