hugpadhugpad

Efficient-Large-Model / H3-to-LTX-Latent-Adapter

Other195M48.8K5Updated Sep 10, 2026
Hugging Face

Downloads

48.8K

Likes

5

Last updated

Sep 10, 2026

License

Not available

An original BF16 inference adapter that converts normalized MiniMax-H3 video latents into normalized LTX-2.5 Conv VideoVAE latents. It is used by Sol-H3-Spark in Sana.

Tasks & library

  • minimax-h3

Tags

  • minimax-h3
  • ltx-2.5
  • latent-adapter
  • video

Technical specifications

Parameters195M
Libraryminimax-h3
LanguagesNot available
LicenseNot available