Minimax character-swap LoRA nails maskless character tracking on first tries
tindell- · reddit · 2026-11-06
A Redditor combined Minimax's stock ref2vid workflow with a character-swap LoRA and character sheets to turn live-action footage into a TMNT-style sequence. Each shot was generated separately and reassembled in an NLE, with almost all clips succeeding on the first attempt — and the model tracked both characters accurately without any masking.
More from Multimodal
- Nunchux Tops 10 Artificial Analysis Leaderboards for Image & Video Generation Speed — junyanz89 · 2026-11-06
- "The Sunset Clause": a short film on regulation dystopia made with Claude and Magnific MCP — AIandDesign · 2026-11-06
- Omni Flash 1.2 demo builds a world that gets painted into existence — AI_Andrew · 2026-11-06
- Recraft's agentic image editing wins praise: smooth iterative tweaks, hi-res and vector export — aziz4ai · 2026-11-06
- Pai v1.0: Open-Source App Merges Painting and AI Generation in One Workspace — g45t345rt · 2026-11-06
- ChatGPT's New GIF Feature Flops: Flickering Frames, Mutated Limbs, No Seamless Loops — Pentunee · 2026-11-06