Nice extension

[light-and-ray]

Hello, I did't find a discussion page, so I made an issue. I was planning to make an extension for lcm hires fix. Generally it should was a copy of default hires fix with two additional options: cfg scale and suffix for prompt. Because upscaling with lcm require cfg scale set to 1.0 and add lora in the prompt, and it is not possible with default hires fix. But, as I see, your extension should do both of these features :)

So, I want to wish you good luck with it. It will be a great extension! Do not abandon it

[w-e-w]

I was planning to finish this one after 1.7 released (which was just yesterday) because of additional UI features is there any other parameters that you want to adjust in high res pass apart from CFG for LCM

[light-and-ray]

is there any other parameters that you want to adjust in high res pass apart from CFG for LCM

Yes auto detection of sd model type (1.x, ssd-1b, sdxl), and put automatically different suffixes provided in settings, if textbox is empty. Plus setting to disable this behavior

[light-and-ray]

Hm, basically, it can be implemented as add-on for your extension. Especially with 1.7.0 features of extension loading order

[w-e-w]

Yes auto detection of sd model type (1.x, ssd-1b, sdxl), and put automatically different suffixes provided in settings, if textbox is empty. Plus setting to disable this behavior

I think that would be useful but I'm not sure if this fits with the extension maybe it should be something that's implemented in web UI

also I'm not sure there's any method to quickly tell apart the model type aside from actually reading the model (I'm not familiar with the metadata structure)

if there is a fast way of turning four different types of models then good but if there isn't then the results can be cached after the model is loaded once

[JPGranizo]

@w-e-w Hi! This is a cool addon! Is it finished yet?

[w-e-w]

Hi! This is a cool addon! Is it finished yet?

I think it's mostly finished but it might be some kinks I need to iron out not planning to add more features to the initial release

---

currently seems to be facing a strange issue with cacheing upscale images for some strange it's reason causes the image to not be pixel perfect the same as if you would to generate them manually or without cacheing

and theres also seems to be some strange issues with img2img it'self in webui (which also affects hires fix) (this is not caused by the extension)

currently trying to work that out these out

these issues don't really impact the use in general but it dose impact a bit on the reproducibility of images so ideally I would like to solve those issues before I release this extension

note I'm beginning to suspect this is not an actual bug, but some sort of precision issue with GPU if this is the case then nothing is wrong with the code

---

also recently new feature were added in dev branch, which would impact compatibility with this extension I've already made two PR (pending merge) to fix the issues https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14645 https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/14626

not that it matters for now because those changes are going to be in webui 1.8 not the current master 1.7

---

I am not happy with the current name of this extension I'm very bad at name and stuff so I asked around and got some suggested names currently still undecided on the final name if you have some name suggestions please tell

[TooAiArt]

This extension is brutal, I love it :))

[w-e-w]

This extension is brutal, I love it :))

can't say I ever seen anyone use the term "brutal" to describe their appreciation, but I will take it