I admitted I was wrong about the SDXL-based anime models in the article I wrote a few days ago, especially about Animagine XL v3. It turns out that I didn't maximize the potential of the model.
This is what my generation looks like with my previous prompting style:
1girl, solo, anya melfissa, brown hair, brown eyes, wavy hair, small breasts, dress, seductive smile,white background, simple background,masterpiece, best quality,Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name,Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 583364425, Size: 1024x1536, Model hash: 1449e5b0b9, Model: animagine-xl-3.0, Clip skip: 2, ENSD: 31337, Version: v1.7.0, Hashes: {"model": "1449e5b0b9"}Messy, bad eyes, unnecessary details, pale colors. Maybe some of you say this is good, but this is not my preference.
Now compare it with this one:
1girl, solo, anya melfissa,brown hair, brown eyes, wavy hair, small breasts, dress, seductive smile,anime coloring, white background, simple background,masterpiece, best quality,Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name,Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 583364425, Size: 1024x1536, Model hash: 1449e5b0b9, Model: animagine-xl-3.0, Clip skip: 2, ENSD: 31337, Version: v1.7.0, Hashes: {"model": "1449e5b0b9"}I've added a style to the prompt, which is anime coloring. Great color, better detail, tidier.
Hands are still bad, Stable Diffusion classic problem. Let's not focus on that issue.
Maybe we can use a popular artist's name for style?
Asanagi:
1girl, solo, anya melfissa,brown hair, brown eyes, wavy hair, small breasts, dress, seductive smile, asanagi, white background, simple background, masterpiece, best quality,Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name,Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 583364425, Size: 1024x1536, Model hash: 1449e5b0b9, Model: animagine-xl-3.0, Clip skip: 2, ENSD: 31337, Version: v1.7.0, Hashes: {"model": "1449e5b0b9"}Mika Pikazo:
1girl, solo, anya melfissa,brown hair, brown eyes, wavy hair, small breasts, dress, seductive smile,mika pikazo, white background, simple background,masterpiece, best quality,Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name,Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 583364425, Size: 1024x1536, Model hash: 1449e5b0b9, Model: animagine-xl-3.0, Clip skip: 2, ENSD: 31337, Version: v1.7.0 Or heck, something in between:
1girl, solo, anya melfissa,brown hair, brown eyes, wavy hair, small breasts, dress, seductive smile,asanagi, mika pikazo, white background, simple background,masterpiece, best quality,Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name,Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 583364425, Size: 1024x1536, Model hash: 1449e5b0b9, Model: animagine-xl-3.0, Clip skip: 2, ENSD: 31337, Version: v1.7.0It is THAT powerful if you know how to use it properly.
Even Animagine XL v3 can do something like this:
Pixel art:
1girl, solo, anya melfissa,brown hair, brown eyes, wavy hair, small breasts, dress, seductive smile,pixel art, white background, simple background,masterpiece, best quality,Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name,Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 583364425, Size: 1024x1536, Model hash: 1449e5b0b9, Model: animagine-xl-3.0, Clip skip: 2, ENSD: 31337, Version: v1.7.0Blue theme + Asanagi:
1girl, solo, anya melfissa,brown hair, purple eyes, wavy hair, small breasts, dress, seductive smile,blue theme, monochrome, asanagi, white background, simple background,masterpiece, best quality,Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artist name,Steps: 30, Sampler: Euler a, CFG scale: 7, Seed: 583364425, Size: 1024x1536, Model hash: 1449e5b0b9, Model: animagine-xl-3.0, Clip skip: 2, ENSD: 31337, Version: v1.7.0I think I will stop here. You definitely got my point if you read until this part.
It’s not the issue of the model, it’s our skill issue. We can very much generate better images if we know how to use the model correctly.
(P.S: I'm sorry for all the Melfriends seeing this. I mostly generated Anya images with brown eyes instead of purple in this post, I forgot that Anya has purple eyes.)
Kayfa out.