Website To Framevideo

Website to FrameVideo

Orchestrate the 7-step workflow to capture a website and produce a professional video. This skill is the workflow coordinator — each step loads the appropriate skill or reference for implementation details.

When To Use

• User provides a URL and wants a video (product demo, social ad, brand reel)
• User says "capture this site", "turn this into a video", "make a promo"
• User wants a video based on existing website content

Quick Start

Basic workflow for website → video:

# User provides URL
"Make a 30-second product demo video from https://example.com"

# System orchestrates:
1. Capture site (Step 0) → extract brand + content
2. Create DESIGN.md (Step 1) → lock visual identity
3. Align on message (Step 2) → get video brief approval 💬
4. Storyboard + script (Step 3) → get narrative approval 💬
5. Generate voiceover (Step 4) → produce audio + timing
6. Build HTML (Step 5) → create FrameVideo composition
7. Validate + deliver (Step 6) → QA + Studio URL

Key points:

• 💬 = requires user approval gate
• Each step loads specific skills/references progressively
• Autonomous mode still requires verification (no skipping QA)

Workflow Overview

Each step produces an artifact that gates the next. Gates marked 💬 require user approval. The workflow loads skills and references progressively — only read what's needed for the current step.

Step 0: Capture & Brand       → Site summary
Step 1: Brand Identity         → DESIGN.md
Step 2: Strategy & Messaging   → Video brief 💬
Step 3: Storyboard + Script    → STORYBOARD.md + SCRIPT.md 💬
Step 4: VO, Timing + Captions  → Audio + transcript 💬
Step 5: Build Compositions     → index.html + compositions
Step 6: Validate & Deliver     → Studio URL

Autonomous Mode Rules

If user signals autonomous mode ("decide for me", "surprise me"):

• ✅ Auto-decide: TTS provider, voice, colors, beat count, music, captions
• ❌ Still required: Asset audit, per-beat HTML review, DoD checklist, verification disclosure

Autonomous mode = decide preferences, NOT skip verification.

Step 0: Capture & Understand the Brand

Load: references/step-0-capture.md

Capture the website using appropriate tool, extract brand assets (colors, fonts, images, copy), understand what the product does and who it's for.

Gate: Print site summary — strategy-first (product value, audience, brand voice) before asset inventory.

Step 1: Brand Identity

Load: references/step-1-design.md

Write DESIGN.md covering visual identity: color palette, typography, component styles, layout principles. Use design-styles.json for exact computed values.

Gate: DESIGN.md exists with color palette, font choices, and do's/don'ts.

Step 2: Strategy & Messaging

Load: references/step-2-brief.md, scan references/capabilities.md TOC

Align with user on what the video must communicate. Parse user's prompt for video type and style. Lock the message and narrative arc before storyboarding.

Gate: Video type, duration, format, message, and narrative arc are locked.

Step 3: Storyboard + Script 💬

Load: references/step-3-storyboard.md

Write storyboard concept-first: message → narrative arc → beats → techniques per beat → brand accents. Write narration script to match. Present both to user for approval.

Gate: STORYBOARD.md + SCRIPT.md exist AND user has approved the plan.

Step 4: VO, Timing + Captions 💬

Load: references/step-4-vo.md Use: framevideo-media for TTS generation, framevideo-voiceover-ssml if pronunciation control needed

If no narration requested: ask about background music, skip to Step 5. Otherwise: choose TTS provider, generate audio, transcribe, map timestamps to beats.

Gate: Either (a) no narration, manual beat timings in storyboard, or (b) narration.wav + transcript.json exist and beat timings updated.

Step 5: Build Compositions

Load: framevideo skill (entire skill — every rule matters) Load: references/step-5-build.md Use: gsap for animation (default), other animation adapters as needed

Build index.html and beat compositions following storyboard architecture. Sub-agents run framevideo lint and framevideo snapshot on each beat.

Gate: Every compositions/beat-N.html has been read top-to-bottom by main agent against DESIGN.md and STORYBOARD.md.

Step 6: Validate & Deliver

Load: references/step-6-validate.md Use: framevideo-visual-qa for final review, framevideo-cli for validation

Run lint, validate, inspect. Take snapshots (formula: max(beats × 3, ceil(duration_seconds / 2))). Review each snapshot. Use visual QA to confirm brand assets don't obscure message and captions don't collide with content. Fix issues before delivering.

Deliver something you're proud of. Ask: would I post this with my name on it?

Gate: npx framevideo lint, validate, inspect pass with zero unaddressed issues, visual QA passes, final response includes Studio URL.

Quick Reference

Typical Video Formats

• Landscape: 1920x1080 (default)
• Portrait: 1080x1920 (Instagram Stories, TikTok)
• Square: 1080x1080 (Instagram feed)

Reference Files

Skills Loaded by Step

Integration

This skill orchestrates:

• framevideo — core composition authoring (Step 5)
• framevideo-cli — validation commands (Step 6)
• framevideo-media — TTS and transcription (Step 4)
• framevideo-voiceover-ssml — pronunciation control (Step 4, optional)
• framevideo-visual-qa — final quality review (Step 6)
• gsap — animation (Step 5, default adapter)
• chanjing-digital-human — if user wants digital human host (Step 4-5)

Validation

After completing workflow:

• All gates passed
• npx framevideo lint passes
• npx framevideo validate passes
• npx framevideo inspect passes
• Visual QA checklist complete
• Studio URL delivered
• "What I did NOT verify" section included in summary

Credits And References

• FrameVideo documentation: https://framevideo.dev
• Workflow references: references/ directory in this skill