Somnai releases Disco Diffusion, the first tool to combine diffusion models with CLIP for text-guided image and animation generation. Unlike VQGAN+CLIP's iterative optimization, diffusion models generate images by gradually denoising random noise, producing higher quality results with more coherent structure. Somnai adds 2D animation techniques, and Adam Letts contributes 3D animation. The tool evolves rapidly through six major versions in eight months. Disco Diffusion becomes the training ground for the generation of AI artists who will later define the AI filmmaking movement. Riccardo Fusetti's AIFF 2023 Grand Prix winner 'Generation' will be made with this lineage of tools.