If you have time and let's not forget a Nvidia video card with plenty of VRAM, there are tons of tutorials on YouTube.
Start with Stable Diffusion webui, learn about different checkpoints and models, learn about txt2img.
Then move to img2img.
Learn about ControlNet, how it will map an existing image body and face position and use it to generate a new image.
Then you would create a rought image in Daz3D for example, use it as source and use AI to generate a comic style image while observing the source body and face.
You can use ReActor or similar to replace any face the AI generate with the source, that is how they keep the same faces all the time.
Face , body are background are done separately, merged for final image.
Note: at least this is how I would do it with my limited knowledge, maybe they have a different workflow.