5.8M-param Speech Event Detector adds laughs, breaths and背景 sounds to Whisper transcripts
Ayush Kumar Bar · hf · 2026-10-16
theboringai-work open-sourced Speech Event Detector with a paper: a 5.8M-parameter conv-Transformer that outputs probabilities for 54 event classes (20 speaker vocalizations, 34 background sounds) every 40ms, post-processed into timestamped events merged into Whisper transcripts. It captures non-verbal cues like breaths, laughs and coughs plus background audio, trained on seven public datasets with unified multi-source annotations and synthetic augmentation. Code and weights are on GitHub and Hugging Face.
More from Multimodal
- Tripo hosts first world-building hackathon in SF with World Labs and Runway talks — mhdfaran · 2026-10-16
- Kling 4.0 boosts text-to-video with character consistency and lip sync — CurieuxExplorer · 2026-10-16
- Huawei's SUPRA-Tex turns one image into relightable 3D materials with WYSIWYG fidelity — Chen Yang · 2026-10-16
- Giant Michael Jackson AI fashion images show the power of extreme proportions — OVolosin82152 · 2026-10-16
- BFL says Flux 3 Klein complaints stem from a Comfy workflow bug, fix underway — desktop4070 · 2026-10-16
- MiniMax H3 + RVC-project combo delivers surprisingly good voice results — ColdComplaint8 · 2026-10-16