⬢github Python · 91 ★ +40 since we first saw it · pushed 1 d ago · MIT
KoljaB/RealtimeVAD
Faster speech off signals for spec gen voice pipelines
RealtimeVAD is a Python voice activity detection library that detects when a speaker has stopped talking about 100 ms earlier than Silero VAD. It layers a fast level-based check over Silero, emitting start/pause/end events every 10 ms so voice assistants can begin transcription, LLM calls, and speech output sooner. It runs on one CPU thread with only numpy and onnxruntime.
Why now: It was just shared on Hacker News as a Show HN, highlighting the trade-off of cutting ~110 ms of end-of-speech latency at the cost of more false positives in word gaps.
Who it is for: Developers building real-time voice assistants or voice pipelines (e.g. STT/LLM/TTS stacks, LiveKit apps) who want lower response latency.
Stars over our 265 snapshots: 51 to 91, since 2 d ago.
Where people talked about it
API: https://socialmediatrends-api.osmike.com/v1/repos/KoljaB/RealtimeVAD