Developer recreates Cartesia's blog experiment, visualizing model router chunk maps on images
davidwromero · x · 2026-11-10
Developer alexdmartin314 reports that recreating Cartesia's blog post experiment was unexpectedly fun and plans to blog the recreation/improvement process. Shared visuals show three-panel outputs — router input frame, chunk maps with centers dotted, and chunk mean colors — where the model segments images (shark, beach scene, corkscrew) into semantic chunks. More findings promised in a follow-up thread.
More from Multimodal
- Real-Time AI Avatars Are Nearly Indistinguishable From Humans, VC Says — venturetwins · 2026-11-10
- Claude Fable 5.2 tops rivals in live comparisons: dodgeball, drumming octopus, fire-breathing dragon — ArtificialAnlys · 2026-11-10
- Fable 5.2 generates a voxel-style lake scene worth a look — Angaisb_ · 2026-11-10
- Lightweight Muse Ember one-shots an interactive 3D tea garden, Meta reshares test — AIatMeta · 2026-11-10
- 2-step LoRA slashes LTX 2.5 video renders to 2 steps with SoL refiner — TimeTruth2490 · 2026-11-10
- Artificial Analysis hosts first image and video model evaluation event on Nov 17 in SF — ArtificialAnlys · 2026-11-10