You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Stable Diffusion CPP has a limited feature set, and in particular lacks Controlnet support for architectures other than SD1.5. However, adding in Canny and Depth Control LoRAs and third-party inpainting LoRAs appears to be able to rectify this gap for both SDXL and Flux.
I wrote some notes about this with reference to sd.cpp.gui.wx here that may be of some use. Even if some of these tools currently struggle on the CPP backend for reasons not yet recognized, I reckon that the process of fixing them up will be pretty reasonable and similar in complexity to the recent PR for SDXL embeddings. Any future UI would also need updating to understand when to call the correct tools, as I similarly suggested for Nexa SDK based on the structure used by @R3gm in stablepy.
The text was updated successfully, but these errors were encountered:
Stable Diffusion CPP has a limited feature set, and in particular lacks Controlnet support for architectures other than SD1.5. However, adding in Canny and Depth Control LoRAs and third-party inpainting LoRAs appears to be able to rectify this gap for both SDXL and Flux.
I wrote some notes about this with reference to sd.cpp.gui.wx here that may be of some use. Even if some of these tools currently struggle on the CPP backend for reasons not yet recognized, I reckon that the process of fixing them up will be pretty reasonable and similar in complexity to the recent PR for SDXL embeddings. Any future UI would also need updating to understand when to call the correct tools, as I similarly suggested for Nexa SDK based on the structure used by @R3gm in stablepy.
The text was updated successfully, but these errors were encountered: