Decoding Algorithmic Artistry: Visual Generative AI
summary
Explains how modern diffusion pipelines power DALL·E 3, MidJourney, Stable Diffusion and introduces OpenAI’s Sora text-to-video as a step-change; describes Nvidia Picasso and a Live Portrait 3D pipeline that produces photorealistic 3D faces at 24 fps on consumer hardware using Nerf encoders; shows a tennis policy learned from broadcast video that chains realistic rallies without annotated rewards; demos weaknesses (DALL·E repeatedly inserts chopsticks into ramen, occasional physics and flame artifacts) and mitigation plans like C2PA provenance and platform AI tagging.
AI-written summary · about
chapters (10)
- 0:00intro
- 0:20preamble
- 0:52introduction
- 1:47diffusion models
- 5:05current research
- 9:59future implications
- 13:06current limitations
- 18:06navigating ai growth
- 24:28summary
- 25:03thanks!
related talks
29mIntegrating Stable Diffusion Generative AI in Artist Workflows
47mAI Image Generation: Visual Artistry with Midjourney, Adobe Firefly, and DALL-E3
49mDALL·E V2 - a Robot and a Painter Doing Image Generation p…
36mTHE ROLE OF GENERATIVE AI IN REGENERATING OUR REALITY
36mArt in the Age of Algorithms
33mThe Dawn Of Universal Creativity
39mKeynote: Artistic Applications of Artificial Intelligence
36mPanel: Generative art for 21st century