SenseVoice-API ghcr.io/hsiang-han/sensevoice-api:latest https://ghcr.io bridge bash false https://github.com/hsiang-han/SenseVoice-API/issues https://github.com/hsiang-han/SenseVoice-API OpenAI-compatible Speech-to-Text API powered by SenseVoice-Small (Alibaba FunAudioLLM). Features: emotion detection (happy/sad/angry/neutral), audio event detection (Speech/Music/Applause/Laughter), auto language identification (zh/en/ja/ko/yue), word-level timestamps, WebSocket streaming. Ultra-low latency (~70ms for 10s audio on GPU). Supports both GPU and CPU mode (set DEVICE to cpu for CPU-only, no GPU required). REST API at /v1/audio/transcriptions. For GPU mode: remove --gpus all from Extra Parameters if running CPU-only. AI: Productivity: Tools: http://[IP]:[PORT:10095]/docs https://raw.githubusercontent.com/hsiang-han/unraid_templates/main/templates/sensevoice-api.xml https://raw.githubusercontent.com/hsiang-han/unraid_templates/main/assets/sensevoice-icon.png --gpus all 10095 /mnt/user/appdata/sensevoice-api/models cuda:0 FunAudioLLM/SenseVoiceSmall https://huggingface.co false 1 10095 all compute,utility