Comparaison des images photoréalistes générées par les 7 modèles de la version gratuite

Do you think, "Even if the model is different, the photorealistic images generated by AI all look the same"? You do, don't you?

But guess what? Each model actually has distinct characteristics!

It's no surprise if you didn't know. On this site, there's barely any prior research comparing the photorealistic image features of each model—mostly just comparisons of mecha images 🤖 (see Yamabico's ).

So this time, I systematically investigated the characteristics of photorealistic images from seven free models available on this site 📷️.

The common prompt is:
"Photorealistic style, park, trees, autumn leaves, black streetlight, night, bench, starry sky"
and the seed value is also the same 🍁🌃.

(Since only up to 5 images can be attached to a post, I've concatenated two images in some cases.)

① Image V3

The contrast between light and dark is precise and clear; the shapes of branches, leaves, and the bench are well-formed. It feels like a photo—excellent realism! 📷️

② Anim4gine

The streetlight is a bit glitchy, but the lighting effect has a softness, and the leaf details are fine. It’s like a hand-painted painting! 🖌️

③ Animij & WAI

Animij has leaves rendered with soft gradients; the contrast between the dark sky, dark trees, and autumn leaves is vivid—a picture-book atmosphere! 📔

WAI's leaf details are a bit rough, but the road texture is fine. The streetlight placement is tidy; excellent compositional stability! ⚖️

④ Neta Yume

The streetlight and bench glitched, but the leaf shapes are sharp, and the yellow and orange colors are vivid. The starry sky is also realistic; the whole image is like a scene from an anime! 🎥

⑤ Image V2 & SDXL

Image V2 has very neat leaf shapes and streetlight decorations, with fine color usage creating delicate gradients. Amazing detail precision! 🔍

SDXL has some instability in the streetlight and branch rendering, but the light bokeh, rich background, and leaf light-and-shade rendering are impressive. The overall feel is fantastical! 🔮

[Summary]

📷️ Image V3 excels in realism,
🖌️ Anim4gine in painterly quality,
📔 Animij in color contrast,
⚖️ WAI in compositional stability,
🎥 Neta Yume in anime-like feel,
🔍️ Image V2 in detail precision,
🔮 SDXL in richness of information!

This is just one test case; changing conditions might reveal different features from each model, so it's worth experimenting 👍.

Finally, I personally prefer Image V3 for its photographic atmosphere and Image V2 for its lovely fine details. Which one do you like?