ReFrameVideo: open-source toolkit to move, remove and reframe any video subject
VoidEclipseAI · reddit · 2026-11-20
A solo developer released ReFrameVideo, an open-source toolkit for editing and regenerating video frames: move/swap people and objects, remove someone with background fill, resize or flip subjects, reframe shots, and edit long videos in segments.
- How it works: subjects become layers via SAM 3.1; edits become a rough cut with gaps painted gray plus a redraw mask; a MiniMax H3 LoRA the author trained treats gray areas as "empty, fill me"; an optional clean first frame improves stability.
- Limitations: best on static shots, slow pans, tracking shots; trained on live-action only; 4–5 min per 5s clip on a 4090; Windows only.
LoRA weights are on Hugging Face (ModelScope mirror); feedback and failure cases welcome.
More from Multimodal
- GPT-6.2 builds a walkable Louvre simulation recreating every painting — venturetwins · 2026-11-20
- Slopcore AI music video goes viral: 500 million copies of 'her' — what number are you? — Yamapama · 2026-11-20
- Suspected GPT-6 Astra leak: one-prompt iterative 3D scene edits with editable Blender files — ZeroStateReflex · 2026-11-20
- Image Models Now Ship Fine-tuned LLM Prompt Upsamplers — Why Not Fuse Them? — ostrisai · 2026-11-20
- How to make viral 2000s-style AI video game trailers: full workflow, prompts, models — Div_pradeep · 2026-11-20
- AI-generated meme: Jensen Huang-like figure flanked by 'AI race accelerates' newspaper — AIandDesign · 2026-11-20