← Home
TWIML · May 15, 2025 · 50m

Diffusion Models: Beyond Image Generation

How diffusion models are being applied beyond image generation: video, 3D modeling, protein design, and audio synthesis.

This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.

Canon

Diffusion models succeed in environments with continuous, structured data (images, audio, 3D shapes) and struggle in environments with discrete, sequential data (text, code). The data environment determines which architecture works.
Charrington traces the biographical arc of generative model architectures, showing how each generation solved the previous generation's problems while introducing new limitations.
Was this useful?