Cheaper Qwen VAE for Anima (and it's training)

r/StableDiffusion
Open Source AI AI Research

Just a modification of Qwen Image VAE that allows you to not waste time on parts that are useless in case of non-video models. I have tried it with lora