YouTube Video Creator

Create complete YouTube videos programmatically — from a topic or brief to a fully assembled, uploadable video with AI-generated images, narration, Ken Burns effects, and background music.

Quick Start

# 1. Create project directory
mkdir -p ~/videos/my-project/audio

# 2. Write a script (see JSON format below)
# 3. Generate visuals (screenshots, HTML slides, or AI images per scene)

# 4. Generate narration
bun .claude/scripts/generate-narration.ts --script ~/videos/my-project/script.json --output-dir ~/videos/my-project/audio

# 5. Assemble video
bun .claude/scripts/assemble-video.ts --script ~/videos/my-project/script.json --output ~/videos/my-project/final.mp4

# 6. Upload (optional, needs youtube-auth.ts setup first)
bun .claude/scripts/upload-youtube.ts --video ~/videos/my-project/final.mp4 --script ~/videos/my-project/script.json

Pipeline Overview

Topic/Brief
    |
    v
[Write script.json] --- scenes[], narration, image prompts, metadata
    |
    v
[Generate visuals] --- AI images (generate-image skill) or HTML screenshots
[Generate narration] --- ElevenLabs TTS (generate-narration.ts)
[Optional: AI video clips] --- FAL AI Kling/Wan (image-to-video skill)
    |
    v
[Assemble video] --- FFmpeg: Ken Burns + crossfades + narration + music
    |
    v
[Upload to YouTube] --- YouTube Data API v3 (resumable upload)

Video Script JSON Format

{
  "title": "Video Title",
  "description": "YouTube video description",
  "tags": ["tag1", "tag2"],
  "category": "28",
  "privacy": "private",
  "scenes": [
    {
      "id": "scene-1",
      "narration": "Text to speak",
      "imagePrompt": "Prompt for AI image generation",
      "type": "ken-burns",
      "imagePath": null,
      "audioPath": null,
      "audioDuration": null,
      "videoClipPath": null
    }
  ],
  "backgroundMusic": "/path/to/music.mp3",
  "backgroundMusicVolume": 0.05,
  "voice": {
    "voiceId": "21m00Tcm4TlvDq8ikWAM",
    "modelId": "eleven_multilingual_v2",
    "stability": 0.5,
    "similarityBoost": 0.75,
    "speed": 1.0
  }
}

Scene Types

Type	Description	Visual Source
ken-burns	Pan/zoom on static image (default)	imagePath
ai-video	Full motion AI video clip	videoClipPath

Visual Strategies

AI Images (most scenes): Use generate-image skill for abstract concepts, backgrounds, illustrations.

HTML Screenshots (for text/data/UI): Write an HTML file, screenshot with headless browser. Great for code snippets, data tables, charts, UI mockups.

AI Video Clips (1-2 hero scenes): Use image-to-video skill for scenes needing motion.

Scripts Reference

Script	Purpose	Key Args
generate-narration.ts	ElevenLabs TTS	--text, --script, --voice, --list-voices
assemble-video.ts	FFmpeg assembly	--script, --output, --resolution, --transition
youtube-auth.ts	OAuth2 setup	--credentials (one-time)
upload-youtube.ts	YouTube upload	--video, --script, --privacy, --thumbnail

Narration

# Standalone
bun .claude/scripts/generate-narration.ts --text "Hello" --output ./hello.mp3

# Batch (from script)
bun .claude/scripts/generate-narration.ts --script ./script.json --output-dir ./audio

# List voices
bun .claude/scripts/generate-narration.ts --list-voices

Popular ElevenLabs Voices

Name	ID	Type
Rachel	21m00Tcm4TlvDq8ikWAM	Female, calm, narration
Adam	pNInz6obpgDQGcFmaJgB	Male, deep, authoritative
Josh	TxGEqnHWrfWFTfGW9XjX	Male, casual, friendly

Assembly Options

bun .claude/scripts/assemble-video.ts \
  --script ./script.json \
  --output ./final.mp4 \
  --resolution 1920x1080 \
  --transition fade \
  --fps 30

Ken Burns presets rotate automatically: center zoom, left pan, right pan, top pan.

YouTube Setup

One-Time Auth

bun .claude/scripts/youtube-auth.ts --credentials ./client_secret.json

Upload

bun .claude/scripts/upload-youtube.ts \
  --video ./final.mp4 \
  --script ./script.json \
  --privacy private

Note: Unverified OAuth apps can only upload as private. Submit your Google Cloud app for verification to upload public videos.

Cost Estimate

Component	Per Scene	8-Scene Video
AI Image (Flux 2)	~$0.05	~$0.40
Narration (ElevenLabs)	~$0.15-0.25	~$1.50
AI Video Clip (optional)	~$0.35	~$0.70 (2 clips)
FFmpeg Assembly	Free	Free
Total		~$2-3

Prerequisites

• ffmpeg: brew install ffmpeg
• ELEVENLABS_API_KEY: ElevenLabs account
• FAL_KEY: FAL AI account
• YouTube upload: Google Cloud project with YouTube Data API v3 + OAuth credentials