Notebook Video v1.1.0 — Visual Director

Notebook Video v1.1.0 upgrades the official workflow to Visual Director v9.

Highlights

• Selects image-text, pure-text, or pure-graphic for every scene from narration meaning.
• Uses Codex built-in image generation when available, with complete provider-neutral fallbacks for other agents and local environments.
• Treats generated images as Remotion input layers; exact Chinese text, callouts, arrows, diagrams, timing, and exits remain editable code.
• Adds manifests/visual-assets.json for provider, prompt summary, crop policy, rights note, and SHA-256 provenance.
• Adds validate-visual-plan for scene coverage, start-inclusive/end-exclusive lifetimes, asset references, integrity, and complete exits.
• Refreshes the official 30-second Remotion exemplar and public demo with original image-plus-text composition.
• Recommends Microsoft Edge Read Aloud as the preferred zero-key Mandarin TTS adapter when appropriate, while keeping TTS provider-neutral.

Validation

• Skill structure and v9 contract validation passed.
• Caption sync: 91 word boundaries across 13 cues.
• Semantic captions: 13 cues and 10 protected phrases.
• Worktree and reachable Git history privacy scan: no findings.
• Clean project creation and npm ci: 185 locked packages installed.
• npm audit: 0 vulnerabilities.
• Pull request and post-merge main CI passed.

Upgrade

Pull the latest main branch or install from this release. Existing projects should compare their scene guards and manifests with the new official example before adopting validate-visual-plan.

The attached MP4 is the silent 30-second public visual demo. Source archives are generated automatically by GitHub.