Plans to support video models (Hunyuan, Wan, etc)

#3
by Wi-zz - opened

Currently VAE decoding of these models is especially slow, so this could be extremely useful. Thanks for your hard work so far.

I started on a small Hunyuan VAE this weekend https://github.com/madebyollin/taehv, will consider making one for Wan as well (are the input/output shapes the same as Hunyuan?)

Added Wan weights to the TAEHV repo as well. The quality is still a bit iffy but it should suffice for decoding fullres preview videos

Your need to confirm your account before you can post a new comment.

Sign up or log in to comment