Day-one MiniMax H3 text-to-image workflow shared, face consistency impresses
Cold_Pudding5326 · reddit · 2026-11-12
A Reddit user shared a day-one ComfyUI workflow for MiniMax H3's image model (t2i and i2i, hosted on Civitai). They report strong face consistency without needing a LoRA, comparable to nano banana or GPT image models, with the main weakness being prompt adherence. The workflow targets realistic portraits and is meant as a community base to improve on.
More from Multimodal
- One-line prompt challenge: Fable 5.2 told to make the worst possible video with its full powers — FinanceYF5 · 2026-11-12
- MiniMax spotlights user-made dark fantasy art generated with H3-Image — Hailuo_AI · 2026-11-12
- Fable 5.2 recreates a code-only generated video in Blender, this time as a loop — FinanceYF5 · 2026-11-12
- User showcases 8 H3-Image creations spanning Polaroid, vintage DV and CAD-style art — Hailuo_AI · 2026-11-12
- ComfyUI-Cleaner released to find and remove unused ComfyUI models — chribonn · 2026-11-12
- An AI video made with ByteDance's Seedance 2.5 is called a masterpiece — SimplyAnnisa · 2026-11-12