We checked this claim: is it true that
Anthropic's Claude Opus produced every frame of an animated short using browser code?
Verified demonstrations confirmed the model wrote functional web code to generate each frame directly on a digital canvas
Anthropic model draws animated film with web code
Online demonstrations showed Claude Opus rendering an entire fruit fly cartoon using browser script rather than traditional video generation.
In a nutshell
Online demonstrations proved that Anthropic's Claude Opus model can animate an entire narrative cartoon by writing executable web code rather than generating video files, sparking efforts across the programming industry to turn automated browser scripts into full-scale multimedia production tools.
Highlights
- Kevin T. Ngo showed Claude Opus creating every frame of a fruit fly cartoon using pure JavaScript.
- The model drew visual assets directly on an HTML browser canvas without relying on pre-rendered images or diffusion models.
- Online communities quickly isolated the underlying prompt patterns used to govern color schemes, textures, and motion timing.
- Programmers expanded the technique into automated workflows that pair canvas animation scripts with browser-generated audio tracks.
From the Editor’s Diary
Generating animations through executable code turns video from a heavy computational output into lightweight software, making visual production far more adaptable even as human prompting remains necessary to hold complex scenes together.
Who's involved
Kevin T. Ngo
Digital creator who published the original browser animation test
goal → Demonstrating that artificial intelligence models can create motion pictures through code rather than video files
Anthropic
Artificial intelligence research firm and maker of the Claude model family
goal → Engineering foundation systems that solve complex programming and geometric reasoning tasks
Developer and Creative Coding Community
Software engineers and digital artists experimenting with automated animation scripts
goal → Testing whether script-driven graphics can reliably replace conventional production software
In short
Artificial intelligence can now generate motion pictures by writing functional software instructions rather than guessing pixels, shifting digital video creation from heavy image processors to simple web browsers. This method makes automated video production likely to expand into lightweight, interactive web formats, because procedural computer scripts bypass the massive computing power required by standard image generation tools. Whether this approach replaces traditional digital animation remains uncertain, because questions linger over the amount of human steering needed to link complex scenes together.
San Francisco-based AI research laboratory Anthropic saw its flagship system, Claude Opus, write pure JavaScript—the standard programming language that runs interactive elements inside web browsers—to illustrate a complete short cartoon depicting the life of a fruit fly. Instead of outputting pre-rendered video clips using diffusion systems, which synthesize frames by refining patterns of noise, the software generated mathematical vector commands directly onto an HTML canvas, an invisible digital sketchpad built into standard internet browsers.
The breakthrough spread rapidly across software development circles, drawing comparisons to early milestones in artificial intelligence, such as GPT-4 generating simple vector graphics of a unicorn. Yet while creative coders welcomed the model's grasp of physical composition and geometry, critics questioned whether human users had to heavily engineer the instructions behind the scenes to keep the motion consistent.
How it unfolded
Claude Opus illustrates fruit fly animation using browser code
Digital creator Kevin T. Ngo published a video recording of a cartoon titled "The life of a fruit fly," reporting that Anthropic's Claude Opus produced the complete JavaScript program required to draw every single frame on a browser canvas.
Online developer communities unpack prompt blueprints as workflow spreads
The clip triggered wide analysis across technical social networks regarding whether generating visuals with software code offers a cleaner alternative to pixel-based diffusion tools. Programmers on Reddit dissected the prompt recipes used to simulate textured print aesthetics, color palettes, and natural movement, while software guides soon broadened the technique to build interactive web graphics and three-dimensional animations.
Where things stand
The core claim stands verified: Claude Opus generated procedural web code that drew every frame of the animated sequence without using any pre-existing image files. Industry attention has since turned to building automated systems that combine these visual scripts with browser-generated audio tracks to output standard video files.