plugin-video

Converts attached videos into image frames with ffmpeg before they reach the model, so any vision-capable LLM can 'watch' them. Requires ffmpeg/ffprobe and a vision model. Version 1.1 Changelog: v1.1 - Persistent frame storage (default): frames are copied to LM Studio's user-files cache so follow-up questions about the same video keep working across turns/restarts; per-chat registry tracks files. New /video commands: list, on <name>, off [name], purge <name|all>. Max 'Frame count' and 'Max frames per video' raised from 100/400 to 1000. v1.0 - Initial release: ffmpeg frame extraction (fixed-FPS or fixed-count sampling, resolution cap) for vision models.