Skip to main content
Video Lab is a local automation runtime for Codex, Claude Code, and similar agent environments. It validates and renders HTML compositions into MP4 files with Breeze TTS narration by default. Pass --no-audio when the output should be a video-only MP4. Human users normally interact with it through an agent workflow rather than by authoring project files manually.

​
Local requirements

Video Lab uses the runtime bundled with the installed Breeze CLI. Local rendering also needs Node.js, FFmpeg, ffprobe, and Playwright Chromium. On first use the runtime downloads Chromium, so run breeze video doctor once to set up and verify the environment.

​
Login and credits

These local commands run without a Breeze login and do not consume credits:
  • breeze video init
  • breeze video capture-url
  • breeze video validate
  • breeze video inspect
  • breeze video snapshot
  • breeze video doctor
Rendering commands require a logged-in Breeze profile or API key with at least 400 credits because they generate Breeze TTS narration:
  • breeze video render
  • breeze video preflight
  • breeze video batch
A --no-audio render still needs the account gate, but it produces a silent MP4 without calling Breeze TTS. 
breeze video doctor --agent

​
Agent workflow

breeze video init --project .local/breeze-video/demo --template authored --force --agent
breeze video validate --project .local/breeze-video/demo --agent
breeze video inspect --project .local/breeze-video/demo --agent
breeze video render --project .local/breeze-video/demo --no-audio --force --agent
breeze video render --project .local/breeze-video/demo --voice voc_... --force --agent
Use the --no-audio render for visual rehearsal. The final render omits --no-audio, generates narration from script.segments[].text, and fails if TTS cannot complete.

​
Visual safety rules

  • Place text on stable light areas, scrims, or solid panels.
  • Keep core titles, captions, and product details inside safe margins.
  • Do not let moving assets cover titles or body text in their final positions.
  • Use timeline_v1 when frame-by-frame animation matters.