Character-swap LoRA for MiniMax H3 preserves motion, camera and scenes
Round_Awareness5490 · reddit · 2026-10-13
A Redditor released a character-swap LoRA for MiniMax H3 (BFS Collection) that replaces the main character in a video while preserving motion, expressions, camera and scene. Trained with guide latents, it works best wired through ComfyUI's Add Guide node, with an optional Planner workflow included.
- Experimental: a LoRA, not a universal character-swap solution
- Works better on people and human-like characters; results vary by video, reference image, prompt and settings
- Available on HuggingFace (Alissonerdx/Minimax-H3-ComfyUI) and Civitai
More from Multimodal
- SpatialASM: Shanghai AI Lab trains LVLMs to actively mentalize viewpoints for spatial reasoning — Shanghai-AI-Laboratory · 2026-10-13
- Paper turned into ComfyUI plugin: more diverse images across diffusion models at no extra cost — michaelchchoi · 2026-10-13
- Ghibli-style "Where Do Cats Come From" AI animation short — hazadaza · 2026-10-13
- "Working from the jungle": another eerily realistic AI video — Beginning-Willow-801 · 2026-10-13
- LAION releases open Humaneness Voice Converter that preserves emotion and timing — lmoroney · 2026-10-13
- Keyframe Labs launches persona-2.0-live avatar model with word-level emotion control and 180ms latency — edgarpavlovsky · 2026-10-13