Seeking Minimax H3 inpainting workflow for motion preservation
Bulky-Schedule8456 · reddit · 2026-08-30
The author asks if anyone has implemented an inpainting workflow for Minimax H3. While Minimax's native capability is good for subject replacement, it sometimes alters movement or expressions.
The goal is to strictly maintain the original motions and trajectories while changing only the masked area (e.g., hair color or robotic hand). The author found a SAM workflow on Civitai for highlighting but not true masked generation, and references a Hugging Face space, seeking a better solution.
More from Multimodal
- The math: 5 rerolls at 480p plus a 4K upscale costs $5 vs $9 at 720p — HeyAmit_ · 2026-10-15
- Dreamina workflow: generate at 480p, upscale to 4K, cut video costs 44% — HeyAmit_ · 2026-10-15
- One-prompt GPT-6 Astra writes two Palestrina-style motets in LilyPond — deanwball · 2026-10-15
- DiffSynth-Studio Adds OPD Workflow: Few-Step Diffusion Training in ~2 GPU-Hours — bdsqlsz · 2026-10-15
- AI film and TV hit the market: Prime Video picks up AI spin-off, pilots screen at Mipcom — lmoroney · 2026-10-15
- Ramayana film taps DNEG's massive VFX and AI-assisted production for global release — CurieuxExplorer · 2026-10-15