| name | article-to-video |
| description | Complete article-to-video pipeline for VoiceForge. Covers project setup, information extraction, visual planning, TTS synthesis, animation code generation, sample validation, post-production, and Remotion export. Strict five-phase sequential execution. |
| argument-hint | <article.md> [project-name] [tts-provider] |
Article to Video — VoiceForge Pipeline
Transform articles into complete Remotion video projects with audio-synced animation.
When to use
Use this skill whenever working on the VoiceForge article-to-video pipeline, including any of:
- Setting up a new video project from an article
- Planning visual scenes from an article
- Generating narration and TTS audio
- Creating animation code for scenes
- Merging audio timing with visual planning
- Rendering sample clips for validation
- Post-production audio polish
- Exporting to Remotion protocol
Iron Rule
先音频，后动画；先样片，后全片。 Load the production order for the full 11-step methodology and phase mapping.
Audio-Driven Animation
When generating animation code or working with the animation engine (renderAt(t), timeline JSON, sync-player.js, iframe sync protocol), load the audio-driven animation guide.
Animation Engineering Standards
Every animation must pass the quality gate. Load the animation engineering standards before writing any animation code. Key rules:
- Primary visual is motion/physics, not text slides
- At least one real physics principle per animation
- At least 4 visual depth items (gradients, shadows, glow, particles, etc.)
renderAt(t) must be pure and idempotent- All elements pre-created, no DOM mutations during render
- Formulas via KaTeX, never plain text
- Minimum viable library stack
Five-Phase Pipeline
The pipeline has 5 strict sequential phases. Never parallelize across phases.
Phase 0: Project Setup
Lock down specs before any content work begins.
Phase 1: Content → Scene Pages
Extract information and plan visual layout before any narration or animation.
Phase 2: Audio → Timelines
Generate narration text, synthesize TTS, build multi-level timelines.
Phase 3: Animation Code Generation
Merge audio truth with visual planning, then generate React/Remotion components.
Phase 4: Sample → Full → Polish → QA → Export
Internal sequence is strict: sample must pass before full render.
How to use
Read individual rule files for detailed schemas, constraints, and examples. The main SKILL.md provides routing — the rules/ files contain the actual implementation specifications.
Quick Start
article_path = $ARGUMENTS[0]
project_name = $ARGUMENTS[1] or derive from filename
tts_provider = $ARGUMENTS[2] or "kokoro"
output_dir = voiceforge-remotion/public/{project_name}
Phase Completion Gates
| Phase | Gate Condition |
|---|
| 0 | project_config.json exists with locked specs |
| 1 | information_graph.json + visual_plan.json + scene_outline.json exist, all 10+ dimensions filled per scene |
| 2 | Every scene has audio.wav, timing from actual TTS, multi-level timelines ready |
| 3 | Every scene has animation component code consuming beats + phrases + visual_objects |
| 4 | Sample passes 3-pass review → full render → post-production → QC passes → remotion_project.json v5.0 exported |
Error Handling
- Each Phase validates output before proceeding
- Phase 2 (TTS) failure: run
voiceforge info, suggest different provider
- Phase 3 (code gen) failure: generate static components first, add animation incrementally
- Phase 4 (sample) failure: fix at source phase, re-render sample — never skip sample gate
- Any Phase failure stops pipeline — never continue with incomplete data
Prohibitions
- No content work before Phase 0 completes (specs must be locked)
- No narration before Phase 1 completes
- No animation code before Phase 2 completes
- No Phase 1 / Phase 2 parallelization
- No duration estimation from text — timelines must come from actual TTS audio
- No full render before sample passes 3-pass review
- No export before QC passes
