Back to Discussions
Tips & Tricks· edited

Comparison of anime-style gothic lolita images generated by 7 free-version models

ねむ
Translated to English·

Once, while generating gothic lolita illustrations with several models, I found myself thinking:

"Each model produces a completely different picture. Actually, on this site, there aren't many posts that generate images with many models on the same theme and compile a comparison, are there? 🤔" (I could only confirm two posts in community submissions, and both compared only two models.)

So I tested how 7 models — Image V3, Anim4gine, Animij, WAI, Neta Yume, Image V2, and SDXL — generate different images! (I also had the motivation of finding out which model is best at generating cute anime-style gothic lolita illustrations 😅)

The common prompt was "anime style, cute girl, long black hair, black gothic lolita dress with lace on the collar, hem, and sleeves, dark night background with ruined medieval ruins, golden lace-patterned frame."

Note: since images attached to posts on this site are limited to 5, some are combined as two-in-one.

① Image V3
image
The fineness of the dress lace, the girl's composed and cute expression, the decadent atmosphere of the background, and the neatness of the golden frame are lovely. If you want a refined picture, V3 is the one 🌟

② Anim4gine
image
It produced a picture with a retro atmosphere reminiscent of anime from around the 2000s. The coloring is flat, but the painterly touch is extremely detailed, so the lace at the girl's collar, the cracks in the building, and the spiral pattern of the frame are very delicate. The girl's expression is the most smiling of the seven 😊

③ Animij ④ WAI
image
For Animij, the black and gold decorations on the dress and background are drawn sharply, and while the girl's expression is somewhat stern, it's great at depicting light and shadow vividly and realistically 🧛‍♀️

For WAI, the dress patterns are luxurious, and the golden frame arranged like windows also has a realistic texture and is gorgeous, while the girl's eyes have no highlights and she has a languid expression, which combined with the ruined background gives a very dark atmosphere 🔮

⑤ Neta Yume
image
It has anime-style coloring, a simple dress design, black hair drawn with beautiful shine, and sparkling eyes on the girl, with a clearly depicted background — the tarot-card-like backdrop is distinctive! I think this is the most adorable of all ☺️

⑥ Image V2 ⑦ SDXL
image
V2 features a dress that looks simple but has delicate patterns upon closer inspection (the black tights are also lovely), the girl's innocent expression, and a background with color unity and depth. I think its golden frame is the most gorgeous of the seven ✨

SDXL unifies the dress, hair accessories, and even the background in jet black and gold, and all of them are luxurious; the balance between light and dark colors throughout the picture is also good, and the girl's expression is nostalgic. This is gothic!!!👑

[Summary]

In this test of generating anime-style gothic lolita images with 7 models:
V3 gives refinement, Anim4gine gives retro anime feel, Animij gives vividness, WAI gives languor, Neta Yume gives cuteness, V2 gives innocence, and SDXL gives luxury! I think that with prompt ingenuity, any model can produce pictures with even more different atmospheres.

Personally, I like V2 and SDXL: V2 because it produced a picture closest to my mental blueprint, and SDXL because it can easily create pictures with luxurious decorations 🎀

Which one do you all prefer?

1

12 replies

Maybe it's because AI has its strengths and weaknesses?

ねむOP·

Thank you for your comment!
That's right. It seems like there are differences in how people interpret prompts, aren't there?

たぬー·

In terms of cuteness, Yume; for character personality, probably anim4. WAI's characters don't come alive without an emotional context or facial expression.

ねむOP·

Thank you for your comment!
I didn’t specify any expression this time, but you’re right—if you’re using WAI, adding an expression prompt would probably brighten it up 😄

Anim4gine is an anime in the style of Mobile Suit Gundam? That kind of feel? Or something?

ねむOP·

That's right! In terms of the series, it feels close to the art style of "Mobile Suit Zeta Gundam" ☺️

T
The Catist·

V3 is my favorite. I wonder if the drawing style is different.

ねむOP·

Thank you for the comment!
The V3 art has a slightly more mature vibe, which is nice, right? ✨
I think V3 is probably the model that focused the most on refining the facial contours 🖌️

image

ねむOP·

Don't worry, I also often run out of ideas😂
In times like that, I often do things like remaking my past works, you know?👍
image

まれん2·

Personally, I like WAI.
Neta Yume seems like it could produce something good if matched with the automatic prompt.
The slightly low quality of SDXL is put to good use, as expected 👏🖤

ねむOP·

WAI's art really brings out that dark and gorgeous feel the most, doesn't it!
Neta Yume goes well with automatic prompts, doesn't it~
SDXL has lower saturation than others, which conversely gives it an old-fashioned vibe, doesn't it✨

Leave a reply

Join the community to share your thoughts and ideas.

Sign in to join the discussion