Reddit User Tests MiniMax H3 with t2va-then-ref2va Pipeline for Consistent Clips
urabewe · reddit · 2026-10-07
A Reddit user shares a three-clip 10s short test built with MiniMax H3 and Yue2 for music: the first clip uses text-to-video, the rest use ref2va with clips and voice from previous generations for consistency. Post work unified the dog barks, removed generated music and isolated vocals. Settings: 0.6mp, lcm/beta57, 4 steps with the DMAD 4-step LoRA.
More from Multimodal
- Spira Maxima ties open-source release to 2M views or 5K followers — nikola_mr64990 · 2026-10-07
- Google ships Nano Banana 2.1, jumps 80 points on LMArena's text-to-image board — danielrock · 2026-10-07
- DRAMA 1.0 launches: edit only the expression layer, swap 10 emotions in existing footage — nikola_mr64990 · 2026-10-07
- video-shotcraft hits 10.5k stars: cinematic product videos via Claude Code + Remotion — tom_doerr · 2026-10-07
- Third-party test: Google's Nano Banana 2.1 beats ChatGPT Images, 5x faster and cheaper — prajdabre · 2026-10-07
- Single Image to Full 3D Scene: Adaptive Chunking Extends Object Generators to Outdoor Rome — Jiraphon Yenphraphai · 2026-10-07