Gitframes is an open-source npm package designed to enable AI coding agents to generate videos programmatically by writing TypeScript code. According to the project documentation, it consolidates capabilities typically spread across three professional desktop applications—compositing and VFX (Photoshop-style), motion graphics and animation (After Effects-style), and 3D scene creation (Blender-style)—into a single lightweight software package.
The tool renders directly to GPU hardware via WebGPU, Metal, Vulkan, and Dawn in Node.js and modern browsers, avoiding the overhead of headless Chrome browsers. This architectural choice yields significant performance benefits: Gitframes achieves 60–120+ frames per second throughput with a memory footprint of approximately 200–400 MB per render, compared to 1.5–4.0 GB or more for Chromium-based solutions like Remotion.
Key technical capabilities include:
Typography & Animation: The Slug GPU algorithm evaluates glyph contours per-pixel in WGSL shader language, producing sharp text at any size (10 to 10,000 pixels) without texture atlases or scaling artifacts. The system supports After Effects-style range selectors and animators for text effects.
2D Effects & Compositing: Over 50 modular GPU shaders provide professional image processing—Curves, Levels, Selective Color, 3D LUTs, Film Grain, Halftone, Mesh Warp, and Screen-Space Relighting.
3D Graphics: Unified scene graph with camera controls, mesh loading (OBJ, FBX, glTF, STL, PLY, and others), physically-based rendering with glass and ambient occlusion, motion blur via physical shutter buffers, and depth compositing that nests 2D elements within 3D planes.
Audio & Procedural Sound: Multi-track mixing, deterministic procedural sound effects (whoosh, impact, riser), and reactive signals that sync visuals to audio. Frame-accurate audio beat grids eliminate timing drift.
Computer Vision: On-device ONNX neural models for object detection, instance segmentation, multi-person pose estimation, and person segmentation—all processing happens locally without cloud round trips.
Data Visualization: Built-in layer.chart method generates line, area, bar, scatter, candlestick, and pie/donut charts with staggered reveal animations.
The package is currently in beta, with APIs subject to change. According to the documentation, Gitframes is optimized for cloud-native and CI/CD deployment, with compact Docker images (~500 MB) suitable for serverless GPU render clusters on platforms like AWS, Modal, RunPod, and Kubernetes. AI agents can install Gitframes skills via Claude Code or other coding assistants using a single-line command.
Key facts
- Gitframes is an npm package for code-first video generation designed specifically for AI coding agents
- It consolidates compositing, motion graphics, and 3D capabilities into one lightweight tool, avoiding the need to install multi-gigabyte professional applications
- The tool uses native WebGPU rendering (via Dawn, Metal, or Vulkan) instead of headless browsers, achieving 60–120+ FPS throughput with 200–400 MB memory per render
- It includes 50+ GPU shaders for professional effects, analytic GPU-rendered typography via the Slug algorithm, 3D scene management with PBR rendering, and on-device ONNX vision models for object detection and pose tracking
- Memory footprint is significantly lower than browser-based alternatives like Remotion (1.5–4.0 GB), making it suitable for cloud and serverless deployments
- The package is currently in beta with APIs subject to change