SDXL - Selection issue

[sardaukar2030]

Hi there,

I’m a Krita user. Some time ago I installed the AI plugin, I was only using the default models. About a month ago I decided to install and try some additional models available at CivitAI. Saw a video who recommended to install SD1.5 Base models, also installed few LoRas. Always placing the files in their respective Comfy folders. The SD1.5 model I installed was the Juggernaut-Reborn Pruned-Model. Everything worked fine.

In order to increase the image quality I decided to try a SDXL model and I downloaded the Juggernaut XL-Ragnarok. I placed the SDXL model in the same folder where I have all the other checkpoints. After initiate Krita I noticed that the Juggernaut XL-Ragnarok model does not appeared in the main model selector window where all the default models and the others SD1.5 checkpoints are. I was not sure what happened, I did click on the gears and opened the AI settings. Inside the settings window, in the Styles tab section where I can see all the models installed (unless I uncheck “show pre-installed styles”). The Juggernaut XL-Ragnarok model was not listed there but it was in the “Model Checkpoint” selector… Why I can not select the Juggernaut XL directly from the main model selector window?. Is this normal?… Please let me know or explain me, if possible, probably there is something that I’m not understanding about models. Maybe I need to configure something.

I would like to use the Juggernaut XL-Ragnarok as a base model for better image composition and quality and then refine, detail and upscale the image with a SD1.5, as some workflows suggest. In my case, if I have a good quality image I can pass it to Juggernaut-Reborn to refine It, in combination of some of the suggested models showed in the Juggernaut reborn page, like AbsoluteReality or EpiCRealism.

Thank you for your help.

[minsky91]

You seem to confuse models (checkpoints) and Krita AI Styles. When you download a checkpoint and place it in the checkpoints folder, no new corresponding Style is created, but you will see the new checkpoint in the “Model Checkpoint” selector, just as you wrote. So, this behaviour of the plugin is normal.

In order to have a Style associated with a model/checkpoint, you need to create it explicitly or clone from an existing Style. Each Style is configured via a corresponding .json file, so the easiest way to create a Style is to find a .json file describing a style closest to your target one, make a copy of it and modify by replacing the old checkpoint with the new one. Or you do this in the Styles tab UI by using the Copy style feature, whichever you find more comvenient.

Krita AI cannot in principle create a style automatically for your checkpoint, there are too many parameters involved in the style definition. You have to do that yourself explicitly.

[sardaukar2030]

Hi minsky91,

Thank you very much for your help. Really appreciate It. I knew I was doing something wrong, I do not have very much experience understanding the AI files in Krita. I was reading carefully everything you said and it makes sense, I only need to digest It.

[minsky91]

Glad to be of help. I should have mentioned perhaps that you create a new Style by cloning an existing/built-in style preset for the same base model. In case of JuggernautXL_ragnarok the closest built-in preset would be Cinematic (XL).

It's a great realistic/universalist model btw, I use it all the time for img2img refinements and inpainting. But I guess it's time to move on to the latest crop of semi-large models like Flux 2 Klein, Qwen (Edit) and Z image turbo, they can save you tons of time for complex edits or style transfers, and add some quality too. But SDXL is sill the king in certain use cases - artistic style-based ones, for example. And the fastest one, for basic / hires refinements (if we drop SD 1.5 out of the picture for good, as we should)

[sardaukar2030]

Hi Minsky,

Your directions worked fine. Your explanation was pretty good, thank you. I appreciate your time to explain.

Agree, Cinematic (XL) is a good one, I saw that the Base model is ZavyChromaXL. At this point I think I like more the SDXL than the SD1.5... but I'm not an expert, only someone learning the basics.

Never have try the Flux models, Qwen or Z image yet, but I’m sure those are as good as you describe It. I also would like to know what AI creators use for animation. I can see there are some workflows but I still need to learn more. Sometimes it is difficult to find tutorials that explain all the process. I do not know if Krita can do good animations or it better to use other App. Need to find out about It too.

I have a laptop computer with two SSD. The main “Disc” is where I have Krita and the AI, I would like to move the Checkpoints and the LoRas to a new folder in the second “Disc” because I have more space there. Need to find an easy to follow tutorial to learn how to do that and move the heavy models there.