Péter Major

Ah, ok there was a mistake in the code, that made UI weird in case of only one controlnet mode was available. I pushed a fix for it. You otherwise...

The fixed version is released. Currently only canny, depth, hed and openpose have feature extractors, so all others need an input image already in the target format.

I plan to do more bugfixing and documentation on the weekend. Also writing guides for controlnet ready model conversion.

Thanks for the info. The mask invert thing is a good idea, and it should be very easy to add it. For the external location I might look into it...

I could add it to the lib yes, but I have many things planned, however it would easy to add especially with the example provided. I will think about it....

I also need this feature.

Yes, it is known that the safety checker has some false positives. This is just the original one by StabilityAI, when I have more time I will look into finding...

How much VRAM do you have? Maybe it is using the integrated GPU, you could check in settings.

Still to be done. You can start auto-generation by pressing the generate while generating, and press it again to stop.

Thanks, I will look into it.