AI

AI YouTube Thumbnail Generator

Kacper Walczak · 06-03-2026

Generate YouTube thumbnails with Python, FastAPI and optional Ollama AI.

Creating YouTube thumbnails is usually a manual process.

Open Photoshop.
Pick fonts.
Add emojis.
Align text.

This small project automates the whole process using Python + Pillow.

Optionally it can also use Ollama AI to generate better titles.

Example UI

You open the web interface and generate thumbnails instantly.

The generator allows selecting different visual styles and generates thumbnails in seconds.

Architecture

The project is intentionally simple.

Frontend (HTML UI)
↓
FastAPI backend
↓
Thumbnail generator (Pillow)
↓
Optional Ollama AI text improvement

Everything runs locally.

No cloud required.

Project Structure

thumbgen/
├── main.py          # FastAPI backend
├── generator.py     # Pillow image generator
├── requirements.txt
├── static/
│   └── index.html   # frontend
└── output/          # generated thumbnails

Setup

pip install -r requirements.txt

Start the server:

uvicorn main:app --reload --port 8000

Open the UI:

http://localhost:8000

Thumbnail Styles

The generator includes several built-in styles.

Each style controls:

• background
• font
• color palette
• layout

Generating Thumbnails via API

You can also generate thumbnails via HTTP.

Example request:

curl -X POST http://localhost:8000/generate \
  -H "Content-Type: application/json" \
  -d '{"title": "Consciousness appears at 6Hz", "style": "neon_science", "emoji": "🧠"}'

The image will be saved in the output/ folder.

Using Ollama AI

The project optionally integrates with Ollama.

Install Ollama:

curl -fsSL https://ollama.com/install.sh | sh

Download a model:

ollama pull phi3:mini

Now the UI button "Suggest text (AI)" will generate titles for thumbnails.

Example flow:

Topic → AI → Title suggestion → Thumbnail generator

Recommended Models for CPU

If you run this on a laptop CPU:

For most setups phi3:mini is the best balance.

Why this tool exists

Thumbnail generators online often require:

• subscriptions
• cloud APIs
• uploading your data

This tool is:

• local
• simple
• extensible
• hackable

You can easily modify styles or integrate it into your own pipeline.

Conclusion

Combining small tools like:

• FastAPI
• Pillow
• Ollama

makes it easy to build practical AI-assisted utilities.

Instead of relying on big platforms, you can build your own tools that run completely locally.

READ

Latest readings

• Readings are sites which will help you with detailed

• information about given topic. Read latest ones from Learn.