Messing around with MiniMax H3
retained text
Tried MiniMax H3 for the first time about 2 days ago, I've been practicing how to train LoRAs for this model with ai-toolkit. These were some of the first clips that I was able to produce locally. Meant to jump in sooner but man there are so many configurations and model variants that I needed some time to figure it out. These were all text to video without first frame or references. Blown away by how well it picked up on the characteristics of the style I was going for.
document fetched 2026-10-24 10:34 · canonical https://reddit.com/r/StableDiffusion/comments/1xdla8a/messing_around_with_minimax_h3