Summary
Cole Medin demonstrates a fully open-source pipeline for generating complete AI videos end-to-end — including synchronized audio, animated scenes, and transitions — using Claude Code as the orchestrating agent. The workflow chains together four key technologies: HyperFrames for programmatic scene rendering (positioned as a more reliable successor to Remotion), ElevenLabs or the free Kokoro for voice generation, Archon as the workflow manager with parallel execution and durability via a Neon database backend, and Claude Code to drive the entire process from a two-sentence prompt.
The video walks through the full pipeline architecture: generating a unique execution ID, copying a video template into an isolated workspace, running a research pass with an anti-fabrication gate, scripting with embedded performance tags for natural speech cadence, generating audio, rendering scenes, and validating audio-visual sync before delivering the final output. Medin shows a 25-second Claude Code explainer video produced with no manual iteration as a live proof of concept.
The repository is publicly available and designed to be set up by any coding agent in under 10 minutes. Medin is clear that output quality isn’t production-ready out of the box — voice inflection and transitions still have rough edges — but argues the workflow has immediate practical value for team explainers and YouTube Shorts, with a customization-friendly template system for those who want to iterate further.
📺 Source: Cole Medin · Published May 14, 2026
🏷️ Format: Hands On Build
