Comment by gokuldas011011

I believe this could be achieved by providing a seed image to the diffusion model and generating memoji based on it. This way fine tuning isn't required.