How to caption datasets for motion LoRA training on Minimax H3?
Ambitious_Corgi5723 · reddit · 2026-11-06
A Reddit user details the struggle of preparing caption datasets for motion LoRA training on Minimax H3: motion-training docs are scarce versus character LoRAs, and H3's changed prompt handling makes it unclear whether captions should be long exhaustive paragraphs, motion-only, or deliberately omit motion so it bakes in. They also break motion into macro (large body movement) and micro (muscle/f tissue vibration) layers and ask about trigger tags — after a week, results are still poor.
More from Multimodal
- GPT Image 2.6 tops Image Arena, up to 2.6x faster, redefines Pareto frontier — cantrell · 2026-11-06
- FLUX 3 Image Hits Artificial Analysis Leaderboards, Strongest in Animation & Gaming — ArtificialAnlys · 2026-11-06
- FLUX 3 models beat FLUX.2 in all 19 Text-to-Image categories; Klein tops $9/1k price frontier — ArtificialAnlys · 2026-11-06
- Minimax character-swap LoRA nails maskless character tracking on first tries — tindell- · 2026-11-06
- AI-animated short STORM-BORN shows cinematic painted wolf close-up — RealisticValuable484 · 2026-11-06
- Baroque Mozart AI music video blends Midjourney images with Kling AI motion — creatoroff · 2026-11-06