✦ Arthemy Painter Illustrious - Quick Start ✦
With version 5.0, I've recalibrated the entire model using my custom merge tools.
My main objective was to provide more "composition freedom" and try to fix the "anime" aesthetic that most of the previous versions were naturally showcasing. This model has been fine-tuned in ComfyUI so it might not be optimized for other UIs.
As always, I strongly suggest to start from the following parameters - in order to have a reliable starting point before diving into your preferred style.
1_____Size: Around 1088x1344
2_____Prompt: Start from the prompts present here:
Quick Positive Prompt
oil_painting_(medium), (chiaroscuro), perspective, fantasy, (solo), ((sketch))
[Your Prompt Here!]Negative Prompt
2D, 3D, toon, limited palette, glowing
2_____CFG: 3.5 / 5.0
3_____Steps: 50
(I've noticed that 30 steps are not usually enough to solve the image completely)
4_____Scheduler / Sampler: Euler a, DEIS Karras
I hope you like my models and I can't wait to see what you all are going to create with it!
Description
Hello, this new version aimed to become:
- More stable (with less extra arms and hands)
- Easier to use (now it works well with minimal negative prompts)
- Less Anime (the previous version was less "western" than intended)
For the overall style:
- Many people asked me to make this model more similar to "Vintage fantasy book covers" or something with a retro vibe like that and I've tuned this model get more towards that aesthetic.
FAQ
Comments (8)
Another work of wonders, thanks for it Arthemy!
One word tho. Regarding Forge users most of the time they ll have "deepbooru: escape (\) brackets (so they are used as literal brackets and not for emphasis)" and ".,\/!?%^*;:{}=`~()" set as word delimiters in their settings. So when you recommend using "oil painting \(medium\)" you should instead recommend using "oil painting (medium)" for Forge users.
Hey there! Thank you very much for the feedback - it's curious because I'm using Forge and that's the syntax suggested by the Booru Tag auto-completer and it seems to work fine.
For now I updated the model suggesting the syntax you suggested, but I'll take a deeper look into it today!
@Arthemy No problem. Illustrious checkpoints are trained on Danbooru-style tags, so they expect the tag to be written exactly as it appears in the dataset unless your checkpoint having been trained using a specific tag (that would be oil paining \(style\) in this case). So if not using a a tool that is asking you to add "\" to Danbooru tags dont add them. Test generating images using "western comics (style)" instead of "western comics \(style\)" and you ll see that the western comics style is rendered (play with the tag weight if needed depending of the LoRa and checkpoint used).
@antarek I've tested it and you're right! I've given too much credit to my extension! Thank you for the feedback, it's truly appreciated! ^_^
@Arthemy My pleasure. Keep the good work on your checkpoints. They are absolute masterpieces.
I'd love to see your take on a stylized 3d model!
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.