Open space-nuko opened 1 year ago
I think it will be possible technically. However the composable-lora extension seems to specify the sub-prompt to apply LoRA with the insertion position of the prompt. So to support the feature with our extension, it may be required to have some new UI controls to specify the sub-prompt for each LoRA (or use the prompt like the Web UI implementation.)
Text box like sub prompts: [ ]
(and set 0 or 0,1 etc. for it) may work, but it may be better more flexible way...
Probably worth testing if the support is compatible with https://github.com/opparco/stable-diffusion-webui-two-shot also
Seems like this extension has been getting some traction lately:
https://github.com/opparco/stable-diffusion-webui-composable-lora
With some hacks it's able to apply LoRAs to specific parts of the image using attention maps. However it only supports the native LoRA implementation of webui. Would be nice to have it in additional_networks too