Framefields: An Apache-2.0 Framework for GPU-Accelerated Video Compositing and 3D Rendering via WebGPU

Framefields integrates compositing, motion graphics, and 3D capabilities into a pure software engineering workflow, eliminating the need for traditional multi-gigabyte creative suites. By leveraging WebGPU, it bypasses headless browser overhead (such as Puppeteer or Chromium), enabling direct hardware framebuffer rendering with throughputs of 60–120+ FPS and a memory footprint of 200–400 MB per worker. Key architectural features include the Slug GPU algorithm for deterministic, frame-accurate text rendering from 10,000px down to 1px, and a library of over 50 WebGPU shaders for visual effects. The framework supports native 3D scene graphs with glTF/USDZ import, physical motion blur, and on-device AI vision models (RTMDet-Ins, RTMO) for pose estimation and segmentation. It is designed for cloud-native deployment and integrates with AI agents via a plugin system that instructs agents to use the npm package for scaffolding and rendering.