🎬 Creator Copilot

An all-in-one toolkit that automates the tedious, repetitive parts of running a YouTube channel — so creators can spend more time actually making videos.

Built for the YouTube Automation Hackathon.

🔗 Live demo: add your Render URL here after deploying (see below) 📦 Source: https://github.com/kingxjayant/creator-copilot

---

Deploying your own copy (free, ~2 minutes, no credit card)

This app is a plain Flask app served with gunicorn in production, so it deploys cleanly to any free Python host. The repo already includes a render.yaml for one-click deploy on Render:

1. Push this repo to your own GitHub account (already done if you're reading this on GitHub).
2. Go to render.com → sign up (free, GitHub login works) → New + → Blueprint.
3. Connect your GitHub account and pick the creator-copilot repo. Render will detect render.yaml automatically and pre-fill the build/start commands.
4. Click Apply / Create. First deploy takes ~2-3 minutes.
5. You'll get a public URL like https://creator-copilot-xxxx.onrender.com — open it on any device, no login needed.

(Free-tier Render apps sleep after 15 minutes of inactivity and take ~30-50s to wake back up on the first request — this is normal and not a bug in the app.)

If you'd rather run it without Render, gunicorn app.main:app --bind 0.0.0.0:$PORT works on Railway, Fly.io, PythonAnywhere, or any other Python host too.

---

The problem

Every YouTube creator, from a first-time uploader to a full-time channel owner, spends a huge chunk of their time on work that isn't actually making videos:

• Staring at a blank box trying to write a title, description, and 20 tags
• Manually designing a thumbnail in Photoshop/Canva for every single upload
• Reading through hundreds of comments to find the handful of genuine questions worth replying to, buried among spam and trolls
• Squinting at a spreadsheet of view/like/comment numbers trying to figure out what's actually working

None of this is hard, exactly — it's just slow, repetitive, and pulls time away from content itself.

What Creator Copilot does

A single web app with four independent tools, each solving one of the pain points above:

Tool	Input	Output
📝 Title & Description Generator	A transcript or rough summary of your video	5 title options, a full SEO-optimized description, and 15-20 tags
🖼️ Thumbnail Maker	Any photo/video frame + a headline	A properly-sized (1280×720) thumbnail with bold, high-contrast overlay text, generated instantly
💬 Comment Moderator	A batch of raw comments (one per line)	Each comment sorted into Spam / Toxic / Genuine, with drafted replies for real questions
📊 Analytics Insights	Video stats copy-pasted from YouTube Studio	Plain-English findings: top performers, underperformers, and specific growth advice

The key design decision: it always works

Every one of the four tools is backed by a genuinely useful rule-based engine that runs with zero configuration and zero API cost:

• The metadata generator uses real NLP techniques (bigram frequency extraction with stopword filtering) to pull out the actual subject of a transcript, not just generic filler.
• The thumbnail maker is pure image processing (Pillow) — no AI needed at all, since deterministic, fast, and free is better here than an AI image call.
• The comment moderator uses pattern-based classification for spam/toxicity that catches the vast majority of real-world cases.
• The analytics tool does real statistics (median, engagement rate, outlier detection) on your actual numbers.

If you add an OPENAI_API_KEY, three of the four tools (metadata, comment replies, analytics advice) automatically upgrade to use GPT for noticeably higher-quality, more tailored output — but nothing ever breaks or shows an error if you don't have a key. This was a deliberate choice so that anyone judging or trying this project can clone it and see it fully working in under a minute, without needing to pay for or configure anything.

Tech stack

• Backend: Python + Flask
• Image processing: Pillow (thumbnail generation, text overlay, auto-contrast)
• AI (optional): OpenAI API (gpt-4o-mini by default, configurable)
• Frontend: Vanilla HTML/CSS/JS (no build step, no framework — keeps the project trivially easy to run)
• Tests: pytest, 20 tests covering every route and core algorithm

Running it locally

git clone <this-repo>
cd creator-copilot
python3 -m venv .venv
source .venv/bin/activate          # Windows: .venv\Scripts\activate
pip install -r requirements.txt

python3 -m app.main
# → open http://localhost:5000

That's it — every feature works immediately with no further setup.

Fastest way to see it working: every tab has a "Try with sample data" button — click it, then click the tool's main action button. You'll see real output in seconds without needing your own transcript, comments, image, or stats export.

Optional: enable AI mode

cp .env.example .env
# edit .env and paste your OpenAI API key
export $(cat .env | xargs)   # or use python-dotenv / your shell's preferred method
python3 -m app.main

The homepage shows a badge indicating whether AI mode is currently active.

Running the tests

pip install -r requirements.txt   # includes pytest
python3 -m pytest tests/ -v

All 20 tests should pass in well under a second — they run entirely against Flask's in-memory test client, no live server or network access required.

Project structure

creator-copilot/
├── app/
│   ├── main.py                  # Flask app + all routes
│   ├── ai_client.py              # OpenAI wrapper with graceful fallback
│   ├── metadata_generator.py     # Feature 1: titles/description/tags
│   ├── thumbnail_generator.py    # Feature 2: thumbnail image generation
│   ├── comment_moderator.py      # Feature 3: comment classification + replies
│   └── analytics_insights.py     # Feature 4: stats parsing + insights
├── templates/index.html          # Single-page frontend
├── static/style.css, app.js
├── tests/test_app.py             # 17 tests, route + unit level
├── requirements.txt
└── .env.example

What I'd build next

• Real YouTube Data API integration to pull a creator's actual channel stats automatically instead of copy-pasting
• Batch thumbnail generation from multiple video frames at once, auto-picking the highest-contrast frame
• A/B title testing tracker that remembers which generated title a creator picked and correlates it with eventual performance
• Browser extension version that overlays suggestions directly inside YouTube Studio

Built with

Python, Flask, Pillow, OpenAI API, vanilla JS/CSS, pytest.