But how do AI images and videos actually work? | Guest video by Welch Labs

At 0:00 the video opens with an introduction, then at 3:37 it explains CLIP, followed by a discussion of the shared embedding space at 6:25. At 8:16 the presenter dives into diffusion models and DDPM, and at 11:44 covers learning vector fields. The video continues with DDIM at 22:00, DALL·E 2 at 25:25, conditioning at 26:37, guidance at 30:02, and negative prompts at 33:39. The outro appears at 34:27, and a brief note about guest videos follows at 35:32. Throughout, the video explores how AI images and videos are generated by turning text into images using diffusion models, CLIP, and related techniques, including how noise videos are processed through a VAE to act like a video compressor.