I saw what you were talking about.
The position of the eyes is adjusted by me, so it’s my mistake on the paper doll. I was in a hurry to present the concept and didn’t notice, because making animated eyes and mouth takes time and here .
I heard you, but as I said, I initially planned to use a simpler style, but there, in addition to a large degree of deviation in color, when trying to create large sizes, in the best case, there will be a breast size of 5 from the questionnaire, which requires me to use a higher quality model.
And with a simple style on large sizes, the character begins to look like a balloon due to the lack of details. That is, there is no point in doing more than three out of the declared 5 using the neural network.
To me, I prefer the simplified style way more. The other two have this uncanny valley feel to them. And style 2 definitely looks like ‘standard AI generation’, which, I feel, would definitely offput some people.
As for my thoughts on the lack of detail and looking like balloons, this is a game about an evil spirit possessing girls to get fat, I think realism is a moot point. 
3 Likes
Aren’t they all generated?
If you had to use AI, i’d probably would like to see characters created with 1 or a mix of 1 and 3, honestly i don’t think that the hyper realistic look of the 2nd would fit the theme of the game.
Said so, i really like the base models of the previously shown images sets, but those models have some problems in the higher stages so idk.
4 Likes
2 I used to work on the clothes, it’s hard to make significant edits to a realistic style so that it’s not noticeable.
1 Like
The original style from the main post easily looks the best imo
4 Likes
Yes, based on the reviews I also came to this idea. I’ll work a little in it without transformations on the original sizes of the characters and make the environment, and then I’ll decide something further.
1 Like