MiniMax H3 LoRA training still fails to preserve character similarity
Kitchen_Carpenter195 · reddit · 2026-08-04
A user reports mixed results training a character LoRA for MiniMax H3 with the real ai-toolkit.
- For WAN 2.2, they say 30–40 images already produce solid text-to-video quality.
- For MiniMax H3, the trained character does not resemble the reference images well.
- They ask whether anyone has had success with LoRA training on H3 yet.
The post is essentially a practical report on fine-tuning quality for a multimodal video model.
More from Multimodal
- Reddit users say MiniMax H3 can generate 10-second clips with unusually few limits — bickid · 2026-08-04
- MiniMax H3 users suggest a simple contrast-node fix for crushed blacks — Cequejedisestvrai · 2026-08-04
- NVIDIA posts NemotronLabs VoiceChat 11B on Hugging Face as a full-duplex model — adefa · 2026-08-04
- MiniMax H3 video generation takes 546 seconds for a 10-second clip on RTX 4070 Super — anyup88 · 2026-08-04
- MiniMax H3 text-to-image takes 8–9 minutes for a 10-second run on a 3090 Ti — iChrist · 2026-08-04
- Reddit user claims a free 2× speedup for MiniMax H3 and shares a YouTube guide — foxdit · 2026-08-04