RefMod vs LoRA: free ComfyUI workflow simplifies MiniMax H3 character references
no3us · reddit · 2026-10-10
Developer vavo released RefMod Pilot, a free ComfyUI workflow for MiniMax H3 that offers a no-training alternative to character LoRAs: instead of learning adapter weights through training, H3's Full Reference mode encodes your images into a reusable reference file. Workflow: upload a few consistent photos, save the reference, then use it in a Characters → Scene → Output video workflow with aspect ratio and duration settings; technical nodes live in a subgraph.
The bundled orange robot example was built from six views, with the reference encoded in 5.7s on a 48GB Blackwell MIG instance (a single encoding measurement, not a video-gen benchmark). Underlying nodes come from the open-source ComfyUI-MiniMaxH3Mod project; code on GitHub, weights on Hugging Face.
More from Multimodal
- One prompt turned OpenAI's 722 math papers into a 92-second animated explainer via Opus — FinanceYF5 · 2026-10-10
- Dev recreates fictional ChatGPT UIs for Windows 98, XP, Vista and 7 — Midnight_Sun_BR · 2026-10-10
- MiniMax H3 removes people from videos inside ComfyUI — RobbaW · 2026-10-10
- BudgetPix: Google and UIUC's pixel diffusion model adapts compute per image, cutting tokens to 10% — CSProfKGD · 2026-10-10
- Free Female Face Prompt Builder adds natural-language-to-SD prompt conversion — dobelmont · 2026-10-10
- Model tuning could stop AI video from fighting 2D animation style, says Andrew Carr — andrew_n_carr · 2026-10-10