search

AUTOMATIC1111 / stable-diffusion-webui

Stable Diffusion web UI
GNU Affero General Public License v3.0
141.4k stars 26.72k forks source link

[Feature Request]: Add support for "CLIP + VQGAN" Model #5126

Open YakuzaSuske opened 1 year ago

YakuzaSuske commented 1 year ago

Is there an existing issue for this?

What would your feature do ?

Add model support for the CLIP + VQGAN model.

Proposed workflow

  1. Go to model selection.
  2. Press model name (vqgan_imagenet_f16_16384.ckpt)
  3. Generate

Additional information

currently it says this when placed in the models. image after doing --disable-safe-unpickle it then says. image Currently in this zip https://drive.google.com/file/d/1-FVkiV8jYFiLh8XB48LHrHmx0TDtl_9J/view?usp=sharing it includes the model and the yaml file. image

jyap808 commented 1 year ago

Is this related to Stable Diffusion?

YakuzaSuske commented 1 year ago

Is this related to Stable Diffusion?

Yes but it's an older model that is not used anymore. Tho it's not "Stable Diffusion" it's a different thing called "Clip guided diffusion".

leohu1 commented 1 year ago

Stable Diffusion does better that the old models, so I don't think it will be supported

YakuzaSuske commented 1 year ago

Stable Diffusion does better that the old models, so I don't think it will be supported

I know, but it's still nice to have that compatibility, i want to fool around and see if i can train some embeds or HN with the old models and create a new style.