Its the cute girl hike dev. Character design seems a bit inconsistent, especially when it comes to the face but its not that bad unlike many other bigger games where multiple artists are involved.
However you can also train your own simple character based model on AI art. So you generate a bunch of similar looking waifus and train your own character model that you can slap on top of the main lora. The thing about this is that it kitbashes all those images automatically into an consistent style and design, so your character will always look recognizable, even if the training data has waifus of slightly different styles. Dev should do that instead, then draw over to add some detail and things they like, also mandantory if you wanna attain copyright on your images. If you wanna have multiple characters in one image its best to create single ones, then bash them together in photoshop.
Hey, which AI model do you use? Is there a website or is this something you scripted yourself?
Pretty clear there are lots of img2img involved. The dev uses in painting and other things to add colors quickly, then overdraws hands and give it a thick cell shade. Done by hand and manual editing. You can also Photobash styles like the pros do. Lets say, you wanna make an anime hentai game, all you have to do is to draw a simple figure, then slap over some bits and pieces from other artists to get the style right. then img2image it Most artists do that already through the process differs individually. Most artists do stuff like this, since AI exists now and is fully legal in japan, most use it to skip a few steps, especially in shading.
In the devs case theres no lora or controlnet is involved here, unfortunately. The results would've been much better, especially since you can direct hands much better.