YouTube Reels Workflow

Turn long-form YouTube training videos into short-form vertical reels.

Automated pipeline: Download → Transcript → AI Clip Selection → Trim → Vertical Reformat → Subtitles → Export.

Quick Start

# 1. Clone this repo
git clone <repo-url> reels-workflow
cd reels-workflow

# 2. Run setup (installs everything)
./setup.sh

# 3. Configure API keys
cp config/config.example.yaml config/config.yaml
# Edit config.yaml with your keys

# 4. Test with one video
./scripts/download.sh "https://youtube.com/watch?v=VIDEO_ID"
./scripts/process.sh output/downloads/VIDEO_ID.mp4

What This Does

Input:  Long YouTube training video (30-60 min)
Output: 5-10 vertical reels (15-60 sec each)
        ├── 1080x1920 (9:16)
        ├── Burned-in subtitles
        ├── Hook text overlay
        └── CTA overlay

Requirements

• OS: Linux, macOS, or WSL2 on Windows
• Runtime: Python 3.10+, Node.js 18+
• Tools: ffmpeg, yt-dlp (installer handles these)
• API: Composio account (free tier works)

Repo Structure

reels-workflow/
├── README.md              ← You are here
├── setup.sh               ← One-command installer
├── config/
│   ├── config.example.yaml ← Template (copy to config.yaml)
│   └── ffmpeg.conf         ← FFmpeg presets
├── scripts/
│   ├── download.sh         ← Download YouTube video
│   ├── transcript.py       ← Extract transcript via Composio
│   ├── find_clips.py       ← AI finds highlight clips
│   ├── trim.py             ← Cut clips from source
│   ├── reformat.py         ← Convert to vertical 9:16
│   ├── add_text.py         ← Add text overlays
│   ├── add_subs.py         ← Burn subtitles
│   ├── export_reel.py      ← Final export with all processing
│   └── batch.sh            ← Run full pipeline on multiple videos
├── templates/
│   ├── hook_overlay.json   ← Text overlay template
│   └── subtitle_style.json ← Subtitle styling
├── output/
│   ├── downloads/          ← Raw YouTube downloads
│   ├── transcripts/        ← Extracted transcripts
│   ├── clips/              ← Trimmed clip segments
│   └── reels/              ← Final exported reels
└── WORKFLOW.md             ← Detailed agent instructions

Agent Instructions

If you're an AI agent setting this up, read WORKFLOW.md — it has step-by-step instructions for the full pipeline including Composio MCP setup, video processing, and reel creation.

Configuration

Copy config/config.example.yaml to config/config.yaml and fill in:

composio:
  api_key: "ck_your_key_here"    # From composio.dev dashboard

youtube:
  channel_id: "UC..."            # Your YouTube channel ID

reels:
  max_duration: 60               # Max reel length in seconds
  min_duration: 15               # Min reel length in seconds
  resolution: "1080x1920"        # Vertical format
  fps: 30
  text_hook: "🔥 Watch this!"    # Default hook text
  text_cta: "Follow for more!"   # Default CTA text

License

Internal use — Project Amazon PH

Tech Stack

Layer	Technology
Automation	n8n workflow
Integration	Reels/social media APIs
Scheduling	n8n cron trigger
Monitoring	n8n execution logs

📊 Codegraph

See codegraphs/reels-workflow.md for the full dependency graph.