search

kfirgoldberg / ConceptLab

Official Implementation for "ConceptLab: Creative Generation using Diffusion Prior Constraints"
https://kfirgoldberg.github.io/ConceptLab/
MIT License
245 stars 19 forks source link

Is this only available for Kandinsky? #3

Closed olliacc closed 10 months ago

olliacc commented 11 months ago

Is this only available for Kandinsky and not any of the stable Diffusion models?

eladrich commented 11 months ago

Yep, the optimization process is dependent on the diffusion prior model used by Kandinsky. You can however try and replace the Unet part with Stable Diffusion by using something like IPAdapter, see this demo for example https://github.com/tencent-ailab/IP-Adapter/blob/main/ip_adapter_t2i_demo.ipynb