stable diffusion
It’s a technique based on denoising diffusion, where images are put into a latent space before being worked on, then the model is trained to iteratively remove random noise from images. Eventually, we can give it pure tv static and it will give a realistic image. Add in a CLIP embedding that comes from the text description of the image as a hint during training, and we can type in our own prompts along with the tv static to influence what’s created.