Descriptions:
Nate Herk demonstrates a Claude Code-orchestrated pipeline that automates the full production process for course video content — from raw script to finished, edited video — by chaining together three external services: HeyGen for AI avatar generation, ElevenLabs for voice cloning, and Remotion with ffmpeg for final editing and motion graphics.
The pipeline works by having Claude Code monitor a Google Drive folder for new scripts, chunk them into 45–60 second segments (the sweet spot before ElevenLabs voice quality degrades and before HeyGen’s 3-minute generation cap), submit each chunk to the respective APIs, collect the resulting video clips, and stitch everything into a finished product. A key constraint discovered during development: HeyGen’s newest Avatar 5 model — substantially more natural in lip movement and gesture than Avatar 3 or 4 — is not yet available via API, requiring Avatar 3/4 for automated batch runs while Avatar 5 is reserved for manual generation. ElevenLabs also imposes a 5,000-character input cap per call.
The practical result: Herk queued lessons 5.0 through 5.4 of a course before bed and woke up to finished video files. He’s candid that the approach is best suited for course content, short-form video, and advertisements rather than a primary YouTube presence, and that his own scripts remain human-written throughout. For creators managing high-volume content production, the video offers a concrete, replicable architecture with honest coverage of current API limitations.
📺 Source: Nate Herk | AI Automation · Published April 15, 2026
🏷️ Format: Hands On Build
