This is a general purpose model, aimed to be able to perform at least reasonably well at different styles and subjects.
If you want to focus on just a specific artstyle, you're probably better of with a model trained specifically on that ;)
For VAE I reccomend XL_VAE_C F1 or F0.51 but that's just my personal preference :)
My main focus with this model was to depict a broad range of subjects and themes, without being too bland.
I mainly enjoy doing fantasy and sci-fi stuff, interesting characters, weird hybrid-creatures and robots, so these are things that I tested more throughly than others and what this model is pretty good at.
It covers a pretty good range of artists, settings, painterly and anime styles, and also does relatively decent realism, although the latter seems to require more extensive prompting, to get really good results.
It also does subjects from many popular franchises and settings pretty well, although not without shortcomings. For example if you wanted to do races from a popular roleplaying game, it has a fairly good idea of what "Drow", "Tieflings" or "Tabaxi" should look like, but has no clue about "Githyanki" or "Aarakocra".
Hands and weird eyes: While this is something almost all models struggle with to some degree, this model seems to have a harder time with it than some of the source-models that went into the mix. So I can't blame those for that ;)
I did my best to minimize those problems, but in the end you're going to have to generate a few more images than with other models, until you hit a seed where everything looks right.
Adding things like "bad hands" or "broken fingers" to the negative prompt can reduce these problems to some degree, you probably know the drill.
???? stuff: Even if it is not specifically geared towards it, like many other models this can produce some ???? content, often unprompted so. So if you are averse to seeing the occasional ??????, you should add address that in your negative prompt.
If you are looking for a model that is good at this kind of stuff you will probably be disappointed though. This model is not meant for this, and will probably not give you what you want.
The following Models went into this mix in varying proportions:
ZavyChromaXL by Zavy (Thanks to Zavy for giving me the permission to publish this merge <3 )
Yamer Grimlock Singularity Void and Anime Unstable Illustrator Stage ÉNA by Yamer
Realcartoon XL v5 by 7Withefire7
ICBINP-v6 by residentchiefnz (in v03)
Check out these models and creators if you have a moment, because they do amazing shit!
Updated a few of the base-models and merged ICBINP-v6 by residentchiefnz into the mix.
This makes this merge a little biased towards realism, but non-photorealistic artworks are still very much possible. You might need to add "photo" to your negative prompt in cases were it still tries to do realism and you don't want that.
1. Os direitos dos modelos repostados pertencem aos criadores originais.
2. Criadores originais que desejam reivindicar seus modelos devem contatar a equipe do SeaArt AI pelos canais oficiais. Clique para reivindicar
