Now someone just needs to pipe the output into stable diffusion.