Musubi Tuner adds VAE image support for MiniMax H3

bdsqlsz · x · 2026-08-15

The dev branch of Musubi Tuner now includes VAE image support and inference scripts for the MiniMax H3 model. Demonstrations show image editing capabilities via fl2va and ref2va models without additional training.

Original post →

More from Multimodal

Multimodal channel →