I thought about the relationship between the old model and the current model of this site.
ねThis site previously had models like Anime, Anime V2, Default, Realistic, and Neta Lumina.
I think many people know Neta Lumina is the predecessor of Neta Yume, but I also think many people don't know the details of the other four models 🤔
So, from past images of those four models, I've thought about the lineage relationship between each model and the current models 💡
This is just my speculation, so if you have other opinions, please write them in the comments 🖋️
[Consideration ①: Anime is the same lineage as Anim4gine]
Looking at an Anime image I discovered today, the atmosphere is very similar to current model Anim4gine's images.
Below are a screenshot of that image and an image I previously created with Anim4gine.
By the way, the former image is also one of my past works, and I remember that when I created it, the models on this site were only Image V2, Anime, and SDXL.
Therefore, I think Anime is either identical to the later Anim4gine, or one is a derivative of the other.
[Consideration ②: Anime V2 is the same lineage as Image V2]
The characteristics of Anime V2 images are very similar to those of the current model Image V2.
Below are a screenshot of one of those images and an image I previously created with Image V2.
Incidentally, it seems Image V2 came first (the first image on this site, a turtle illustration, was also generated by Image V2 🐢 See here).
Also, I've found that Anime V2 and Image V2 coexisted for a time (for example, my previously tagged "Amami Oshima Series" was generated with three models: Anime V2, Image V2, and Default 🏝️ See here).
Therefore, I believe Anime V2 is a derivative of Image V2.
[Consideration ③: Default is the same lineage as SDXL, but not completely identical]
In a previous post by Maren-san, SNoW-san said that Default is SDXL ().
However, while they are definitely in the same lineage, I think they are not completely identical.
The reason is that, for example, with anime-style images, many Default works feature faint outline lines and a watercolor-painting-like touch, whereas SDXL works feature sharp outline lines and vivid colors.
Below is one Default work and an image generated by SDXL with the same prompt.
By the way, since Default came first, I think it's likely that the "SDXL" on this site was created by adjusting the output parameters or something of Default.
[Consideration ④: Realistic is the same lineage as Default and SDXL]
I recently discovered an anime-style image generated by a model called Realistic. Below is a screenshot.
Now, when I saw this, I thought:
"Could this be an evolved version of Default?"
The slightly unstable depiction of hands, the look of eyes lacking highlights, yet at the same time the richness of color and detail like SDXL...
Putting these together, I believe Realistic is a derivative of Default and in the same lineage as SDXL.
[In Conclusion]
Probably, on this site in the past, multiple derivative models of a certain model appeared, and those derivative models or the original model were discontinued, leading to the current model lineup ⚙️
This was a speculation that came to me after repeatedly failing to reproduce Default's pale colors with SDXL 😇