I have been using Stable diffusion A1111 since 2022, and many things have changed since then. Back in the days, checkpoints consist of a text encoder, a Image auto encoder and decoder (VAE) and a U-net build in to the check point.
The perks of that, is that when you download a checkpoint there is no need to configure the minor stuff.
But, now with SDXL and with models like anima and other stuff. you need to download respective CLIP and VAE models. There are still many model that come with VAE, but to better switch models, I think it’s better to link the noodles with VAE loader, Clip loader.
Leave a Reply