Multimodal · Part 8Images, Speech & Video
Autoregressive Image Generation: Drawing Pictures as Tokens
VQ tokenizers, raster-order and masked transformers, next-scale prediction and unified models: how LLM-style models draw images, measured on a laptop.