Look up diffusion models.
Indeed; but I've only heard of them being used for images rather than text. Why?