Video Factory — setup & how to use

What you get

• runner.mjs — the engine (queue, TTS lane, render pool)
• Live dashboard at http://localhost:4300
• Boot: Mac ./start.command · Windows node runner.mjs
• Empty folders ready: building, queue, work, done, output, …

Requirements

• Mac or Windows — voice differs (Mac MLX · Windows Chapter 4B)
• Node.js 20+ — check with node -v
• HyperFrames CLI — npx hyperframes --help
• Local voice clone wired to your OS path for auto-narration

1. Download & unzip

1. Log in → open Library → Views on Autopilot
2. Under additional material, download Video Factory (.zip)
3. Unzip somewhere easy, e.g. ~/Desktop/video-factory or C:\video-factory

2. Start the factory

cd video-factory
# Mac:
chmod +x start.command && ./start.command
# Windows:
node runner.mjs

Dashboard: http://localhost:4300/dashboard.html. Leave the terminal open — closing it stops the factory.

3. Fire ideas (any terminal)

1. In any Claude Code session, say what videos you want — one idea or a batch of 10.
2. Claude captures each idea as building/<slug>/ (shows as Building on the dashboard immediately).
3. When the HyperFrames project is ready, Claude moves it to queue/.
4. The factory voices + renders 3 at a time. Finished MP4s land in output/ and ~/Desktop/ready-to-post/.

Lifecycle: Building → Queued → Voicing → Rendering → Done

What a job is

Any HyperFrames project folder in queue/. Optional job.json:

{
  "title": "My product reel",
  "voice": true,
  "skill": "product-launch-video",
  "renderArgs": ["--quality", "high"]
}

If the folder has SCRIPT.md + STORYBOARD.md + audio_meta.json, the factory runs local voice first, then renders.

Reality check

One Mac renders 3 videos at once; voice is one-at-a-time. ~100 videos drains in about 2 hours unattended if you keep the queue fed. Firehose, not teleportation — which is how 200 in a day is real.