creating datasets of IA generated characters

[Hernan-Barrientos]

Hi everyone!
I was thinking a way to produce datasets of generated characters for fine tuning and today I founded a simple way to do it.
I did just a few tests using img2img, the idea is cropping every image in each angle and use it in dreambooth

[ghost]

Interesting! What kind of prompt do you feed it? Just the description of the character or also prompt engineering like "character design", "concept art" and such?

[Hernan-Barrientos]

I used simple prompts. the prompt of the first img was "photo of beautiful girl, red hair, blue pupils".
I think is very interesting that the IA create the same character instead of a random face in every head.

[ghost]

Yeah it's impressive! I'm sure that more words of the character design world would help. Could you share that template image?