Show HN: Jevcut – an auto-clipper where code picks the cut points, not the model

Category: ai-ml

Tags: video-editing, auto-clipping, llm-judge, transcription, content-repurposing

Score: 7.0/10 (Innovation: 7, Technical: 7, Documentation: 8, Utility: 6)

Jevcut is an auto-clipper that turns long videos into short, self-contained clips by having code enumerate candidate cut points (sentence ends, pauses, speaker changes) and using a language model only to judge the resulting clips, never to name timestamps. It is interesting for its measured, benchmark-backed approach to the boundary problem in clip detection, with explicit standalone gates, a cascade architecture, and a dedicated evaluation harness.

Target audience: content creators, video editors, podcasters, media engineers

Repository: https://github.com/VBS2004/jevcut · Python · MIT · 4 stars

View on Hacker News