Back to AI News
The Decoder

Anthropic's Claude Fable 5.2 paints images with hundreds of thousands of simulated brushstrokes

Anthropic's Claude Fable 5.2 takes an unusual approach to image generation: instead of directly computing pixels, the language model writes Python code that applies simulated brushstrokes, complete with bristle texture and paint application, to a virtual canvas. According to Anthropic employee Jake Eaton, a single photorealistic parking garage scene required roughly 700,000 strokes. The approach differs fundamentally from how other AI image models produce output.

AI Brief

It documents a novel and concrete paradigm for AI image generation in which a language model programs simulated brushstrokes rather than predicting pixels, which is significant for anyone tracking generative model design.

Anthropic's Claude Fable 5.2 doesn't generate images the way other AI models do. Instead of computing pixels, the language model writes Python code that places hundreds of thousands of simulated brushstrokes on a virtual canvas, complete with bristle texture and paint application. A single photorealistic parking garage scene took roughly 700,000 strokes, according to Anthropic employee Jake Eaton.

The article Anthropic's Claude Fable 5.2 doesn't generate images, it paints them appeared first on The Decoder.

graczyk_demo.mp4
aris_zhu_fable52_painting.mp4
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!