Creating a textual inversion embedding records the location/form of a concept inside of Stable Diffusion’s latent space. You can ‘apply’ them to images using img2img so that the resulting image also exhibits the concept recorded by the embedding.

Making flat images 3D