Minimax H3 fast i2v tested: 8 steps at 3.20s/it, workflow and prompts shared
Time-Ad-7720 · reddit · 2026-09-19
A Redditor tested the Minimax H3 fast i2v model (fastvideofasth38stepv2prunedint8convrot) locally, hitting 3.20s/it at 8 sampling steps, and published a downloadable workflow.
The demo composes three 3-second POV shots into a matchcut sequence, with full prompts included. Key techniques: start/end frame constraints for precise motion control, explicit bans on pauses/camera moves/cuts for continuity, specified audio cues, and style/label/HUD consistency instructions. The background song was a human+AI hybrid production.
More from Multimodal
- Experimental video mixes visual styles, generated with Seedance 2.5 on Adobe Firefly — umesh_ai · 2026-11-04
- Hailuo AI claims H3 scored perfect on language tests; Japanese creator praises its Japanese signage fidelity — Hailuo_AI · 2026-11-04
- YANAMI: 400K-dialog multimodal dataset shows models still fail at stateful multiturn interaction — Shengyu Wang · 2026-11-04
- Open-source Idol model turns a single image into 3D in 45s on 8GB VRAM — Promptmethus · 2026-11-04
- ComfyUI v0.44.0 released with new Patina texture nodes — Fcking_Chuck · 2026-11-04
- Krea 3 technical report released: a deep dive into data, architecture and training — bdsqlsz · 2026-11-04