Skip to content

AI Infrastructure Video Pipeline

A pipeline that produces a 1080p narrated explainer video about AI infrastructure — LLM Gateway, AI Gateway, MCP Gateway, MCP Registry, and Skill Registry — using HyperFrames for composition and ElevenLabs for text-to-speech narration.

What it produces

A 3-minute, fully-animated 1920×1080 MP4 video (video/hf/final.mp4) with:

  • 7 scenes, each as a self-contained HyperFrames block
  • Per-scene ElevenLabs TTS narration with 600 ms lead / 900 ms trail padding
  • 400 ms crossfade overlap between scenes
  • Orsted-inspired split-panel layout (text left, image right)
  • GSAP entrance animations and Ken Burns image effect per scene
  • Final fade-to-navy closing card

Quick overview

video/hf/
├── build.py          # Single source of truth — run this first
├── index.html        # Generated host composition (slim, 44 lines)
├── final.mp4         # Rendered output
├── scenes/           # One HyperFrames block per scene (generated)
├── audio/            # ElevenLabs MP3s per scene (generated, cached)
└── img/              # PNG image assets (tracked in git)

Two-step workflow

# 1. Generate audio + emit index.html + scene blocks
python3 video/hf/build.py

# 2. Render to MP4
npx hyperframes render video/hf -o video/hf/final.mp4

See Quick Start for full setup, or jump to Architecture to understand how the pieces fit together.