I first happened upon these features in ChatGPT and in CapCut and was interested in trying to create something on my own, locally, in my own style. I'm not quite sure where to start though. I'm a bit of a beginner in Comfy and A1111.
I'm aware there are some doll style loras out there, for example https://civitai.com/models/309747/dolly-merge-xl
But, while those loras are good for generating from scratch, I'm wondering how I can do what ChatGPT and CapCut are doing, which is creating a doll style image out of a reference image. I don't know if it's a specific workflow I should use, or if I need to find a diffusion model that is trained to do this already.
Eventually I'd like to experiment with the style of toy/doll, but for now I’d settle with getting a basic workflow up and running and identifying the models/loras I need to work with.