π₯ [Updated] Style Explorer: 3,000+ Danbooru Artist Comparison
Visual library for Anima 2B model to compare and find artist styles.
Database: 3,000 artists (Target: 5,000).
- Features: Search, Favorites system, Mobile-friendly.
- Performance: Fast loading via WebP previews.
Update: Raw Styles & New Character Benchmark
Major overhaul based on community feedback:
Raw Aesthetics:
Removed all quality tags (masterpiece, score_9, etc.) from the positive prompt to show authentic artist influence.Benchmark Character:
New standardized character designed for optimal anatomy and color readability.Accuracy:
Improved comparison of how different artist styles actually affect the generated image.Status:
3,000 image previews re-generated to this new standard.
I think your quality tags in examples overpower the artists too much. (Are you using Pony model score tags?)
I think your quality tags in examples overpower the artists too much. (Are you using Pony model score tags?)
I'll run tests, and if so, I'll regenerate. Thanks for pointing that out.
You've already made a lot of images, and I'd like to suggest that prompts like "black lips" seem to mask the artist's unique way of expressing his lips and interfere with his evaluation.
You've got too strong quality tags, do NOT use the score tags, they change artist/image composition a lot. Case in point to see if they're overpowering: bkub. It should be a simple drawing. Yours is not. Sorry to say, but you're going to have to redo, as some of your images are VERY far from the actual artist style.
You've got too strong quality tags, do NOT use the score tags, they change artist/image composition a lot.
Do you mean not to use score tags only in the main prompt, but what about the negative prompt?
Do you mean not to use score tags only in the main prompt, but what about the negative prompt?
From my testing, putting score_1, score_2, score_3 in negatives is fine, it just gets rid of the worst of the worst. But when put 7,8,9 or masterpiece in the positives, it's going to affect the artist style and is not "neutral" anymore. I've done a preview set for myself, and my current positive prompt is 'highres, <artist tag>, <prompt>'. The images might not be the flashiest, but you do get closest to neutral artist output.
Do you mean not to use score tags only in the main prompt, but what about the negative prompt?
From my testing, putting score_1, score_2, score_3 in negatives is fine, it just gets rid of the worst of the worst. But when put 7,8,9 or masterpiece in the positives, it's going to affect the artist style and is not "neutral" anymore. I've done a preview set for myself, and my current positive prompt is 'highres, <artist tag>, <prompt>'. The images might not be the flashiest, but you do get closest to neutral artist output.
What's your entire negative prompt? I am interested in reducing mine.
What's your entire negative prompt? I am interested in reducing mine.
I'm using lowres, bad anatomy, bad hands, extra fingers, missing fingers, low quality, glitch, jpeg_artifacts
I want to remove the most glaring "wrongs", but if an artist is sketch, crude in style , or not aesthetic pleasing, i do not want to cover that with a negative, so as minimal as possible.
I think just putting in the artist and nothing else in the prompt would be best.
If an artist name is not recognized, you'll end up with latent noise, and you can't tell if it's real or not. my current prompt is highres,
@artist
, standing, contrapposto, confident, outside Just having neutral prompt having a person standing, will tell a lot. You can easily tell H artists, the body types in their respective works, or the shading used. Less is better, but you need a reference point.
Update: Raw Styles & New Character Benchmark
Major overhaul based on community feedback:
Raw Aesthetics:
Removed all quality tags (masterpiece, score_9, etc.) from the positive prompt to show authentic artist influence.Benchmark Character:
New standardized character designed for optimal anatomy and color readability.Accuracy:
Improved comparison of how different artist styles actually affect the generated image.Status:
3,000 image previews re-generated to this new standard.
Note: This is the final version of the benchmark. It has been tested and approved for its balance of readability and style consistency. No further changes to the character or settings are planned.
Much better! Thanks for making the changes.

