Community LoRA turns MiniMax-H3 33B video model into image style transfer
linoy_tsaban · x · 2026-11-10
A community LoRA, minimax-h3-image-restyle (Apache-2.0) by witcherderivia on Hugging Face, tames the 33B MiniMax-H3 video generation model for content-preserving style transfer on images, with new updated weights released.
Demo images show multiple reference-style pairs: input content (Witcher screenshots, idol group photos, a pencil horse sketch, an illustrated girl) plus style references produce outputs that keep composition and subjects while adopting the target style — covering flat illustration, macaron-doll textures, and painterly looks. Inference code and usage instructions are to be released.
More from Multimodal
- pi-H3 generates 768p video in 2-4 steps: 10s clip in 9.2s on a single B200 — GordonWetzstein · 2026-11-11
- MiniMax H3 Image arrives on fal: open-weights model with native 4K generation and editing — noahsolomon · 2026-11-11
- Fable 5.2 Demo: Hoodie Models Step Out of Product Strips and Dance on Fashion Sites — EricBuess · 2026-11-11
- MiniMax H3 Image: recommended generation sizes shared from official docs — fruesome · 2026-11-11
- One prompt, one tiny planet: Fable 5.2 builds and films a walkable 3D world — prasenx · 2026-11-11
- Video Decompiler: Open-Source Skill Has Fable 5.2 Code Every Frame of a Video — brandon_galang · 2026-11-11