An end-to-end AI-powered automation system that converts long YouTube videos into short-form clips (YouTube Shorts / TikTok) with zero manual editing.
Built as a portfolio project demonstrating AI integration, cloud deployment, and workflow automation.
- API Health: https://yt-long-to-short-automation-system.onrender.com/health
- Trigger: Drop a YouTube URL into Google Sheets → clip appears in YouTube drafts automatically
``` Google Sheets (input: YouTube URL) ↓ n8n (workflow automation + scheduling) ↓ Flask REST API (Render cloud deployment) ↓ yt-dlp → download video ↓ AssemblyAI → AI transcription ↓ ffmpeg → slice into 20s clip ↓ YouTube Data API → upload as Private draft ↓ Google Sheets (updated with clip URL) ✅ ```
| Layer | Technology |
|---|---|
| Language | Python 3.12 |
| API Framework | Flask + Gunicorn |
| Video Download | yt-dlp |
| AI Transcription | AssemblyAI |
| Video Processing | ffmpeg |
| Workflow Automation | n8n |
| Cloud Deployment | Render |
| Keep-Alive | cron-job.org |
| Storage Trigger | Google Sheets API |
| Output | YouTube Data API v3 |
``` yt-long-to-short-automation-system/ ├── src/ │ ├── downloader/ # yt-dlp video download │ ├── transcriber/ # AssemblyAI transcription │ ├── clipper/ # clip detection + ffmpeg slicing │ ├── captioner/ # caption burning (local only) │ └── uploader/ # YouTube upload (via n8n) ├── screenshots/ # portfolio screenshots ├── output/ # generated clips (gitignored) ├── api.py # Flask REST API ├── main.py # local pipeline runner ├── Procfile # Render deployment config ├── requirements.txt # Python dependencies └── .env # secrets (gitignored) ```
- Python 3.12+
- ffmpeg installed
- AssemblyAI API key
- Google Cloud project with YouTube Data API v3
```bash git clone https://github.com/MDanieechi14/yt-long-to-short-automation-system cd yt-long-to-short-automation-system python -m venv venv source venv/bin/activate pip install -r requirements.txt ```
Create a `.env` file: ``` ASSEMBLYAI_API_KEY=your_key_here ```
```bash python main.py ```
```bash python api.py ```
- Push repo to GitHub
- Connect to Render
- Set build command: `pip install -r requirements.txt`
- Set start command: `gunicorn api:app --bind 0.0.0.0:$PORT --timeout 600`
- Add environment variable: `ASSEMBLYAI_API_KEY`
- Set up cron-job.org to ping `/health` every 5 minutes
| Node | Purpose |
|---|---|
| Google Sheets Trigger | Detects new YouTube URL in sheet |
| HTTP Request (POST) | Sends URL to Flask API |
| HTTP Request (GET) | Downloads clip binary |
| Google Sheets Update | Writes clip URL + status |
| YouTube Upload | Uploads clip as Private draft |
| Limitation | Cause | Solution |
|---|---|---|
| Low resolution output (320p) | Render free tier 512MB RAM | Upgrade to Render $7/month (2GB RAM) |
| Single clip per video | Memory constraint | Multiple clips available on paid tier |
| No captions on cloud | Memory constraint | Full captions run locally via main.py |
| Cold start delay (~30s) | Render free tier sleep | cron-job.org keep-alive ping every 5min |
| No persistent storage | Render free tier ephemeral storage | Use Cloudinary or S3 for production |
- Add caption burning on cloud (paid tier)
- Multiple clips per video
- Vertical crop for true Shorts format (9:16)
- Thumbnail generation
- TikTok upload integration
- Slack/email notification when clips are ready
- Web UI for non-technical users
- All API keys stored in `.env` (gitignored)
- `.env` never committed to GitHub
- YouTube OAuth uses Google secure OAuth 2.0 flow
Built by @MDanieechi14 as a portfolio project for AI automation / workflow engineering roles. EOF