ChatGPT images as a text-to-motion compiler: generate many animations, pick the best

cocktailpeanut · x · 2026-11-06

cocktailpeanut argues that animation has always been treated as something you build first (Blender scenes, keyframes, mocap) because building from scratch was expensive — but when generation gets cheap enough, generating many alternatives and picking the best becomes more practical.

In practice, ChatGPT images can serve as a text-to-motion compiler: no Blender, no manual keyframes, no mocap reference. Just iterate on multiple versions via chat branches in natural language, then send the best one to a video generator.

Original post →

ChatGPT images as a text-to-motion compiler: generate many animations, pick the best

More from Multimodal