The full guide from my reel: what the tool is, what it costs, and how to run it on your laptop.
The tool is AI-Youtube-Shorts-Generator, a free, open-source project on GitHub (MIT licence, about 5,700 stars and updated in late October 2026 when I checked). It does the job of paid clipping tools like Opus Clip: you give it a long video and it gives you back vertical shorts.
Link: github.com/Anil-matcha/AI-Youtube-Shorts-Generator
What it does
- Takes a YouTube link and transcribes the video with Whisper
- Uses an AI model to find the strongest moments: hooks, quotable lines, emotional peaks, practical tips
- Gives every clip a viral score, a suggested opening hook and a one-line reason
- Crops to 9:16 for Reels, Shorts and TikTok (local mode tracks faces to keep the speaker in frame)
- No watermark, no per-clip credits, and long videos are split into chunks automatically
What it costs
The code is free. You pay only for the AI it calls, using your own key. It has two modes:
- API mode (default): needs a MuAPI key, which handles download, transcription and cropping for you.
- Local mode: runs on your own computer with ffmpeg and only calls an AI model to pick the highlights. It works with an OpenAI or a Gemini API key.
What you need
- A laptop or desktop (this is not a phone app)
- Python 3.10 or newer
- For local mode: ffmpeg, plus an OpenAI or Gemini API key
How to set it up
1. Download the code and open the folder:
git clone https://github.com/Anil-matcha/AI-Youtube-Shorts-Generator.git
cd AI-Youtube-Shorts-Generator
2. Create a clean Python environment and install what it needs:
python3.10 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
pip install -r requirements-local.txt
(On Windows, activate with venv\Scripts\activate. The second install line is only for local mode.)
3. Create a file called .env in the folder. For local mode with Gemini, it looks like this (put your own key in, and never share it):
LLM_PROVIDER=gemini
GEMINI_API_KEY=your_gemini_key_here
LOCAL_WHISPER_MODEL=base
LOCAL_OUTPUT_DIR=output
4. Run it on a video:
python main.py "https://www.youtube.com/watch?v=VIDEO_ID" --mode local
Your clips land in the output folder. Leave out --mode local to use API mode instead.
Before you rely on it
- Only clip videos you own or have permission to reuse. Reposting someone else’s content can get your account struck.
- The AI picks good moments most of the time, not every time. Watch each clip before you post it.
- It is an open-source project, so setup can break when things update. Check the GitHub page for the latest instructions.
- Keep your API keys private. Never paste them into a chat, a screenshot or a public file.
Source: the project’s GitHub README (checked 1 November 2026).
Found this useful? Follow me on Instagram @kevinsingh.me for a new AI tool every day.
