Open EnzoAndree opened 1 year ago
There's another implementation of CLIP Guided Stable Diffusion for reference here as well: https://github.com/aicrumb/doohickey/blob/main/Doohickey_Diffusion.ipynb
I'm surprised this hasn't be added yet. Being stuck to just ViTL14 is, to say the least, unfortunately. There is such better CLIP models with much higher fidelity and creative ability. Even Diffusers has support for changing the CLIP model now, and one can freely reference them for automatic downloading.
Is your feature request related to a problem? Please describe. It would be great to add support for using CLIP in WebUI. It seems to have better results at the cost of more processing time.
Describe the solution you'd like an option that allows using CLIP in the image generation tabs.
Additional context Here is an implementation.