AI shorts generator

A free open-source project designed to turn youtube-videos into viral short videos. Highlight detection, subtitles, translation, voiceover, all in one for your content: no pre-clip credits or any watermarks. Designed for creators who want an alternative to short-video SaaS tools like OpusClip or Vidyo.ai for free.

Examples of a processed video

<table> <tr> <td align="center"> <img width="360" height="670" alt="image" src="static/ex1.png" ><br> _{<i>"The Speech that Made Obama President"</i>} </td> <td align="center"> <img width="298" height="533" alt="image" src="static/ex2.png" ><br> _{<i>"How Tom Overcame Social Anxiety - The Mindset That Changed Everything"</i>} </td> <td align="center"> <img width="298" height="533" alt="image" src="static/ex3.png" ><br> _{<i>"How to stay calm when you know you'll be stressed | Daniel Levitin | TED"</i>} </td> </tr> </table>

Features 🪁

• API: Freely use this generator in your own projects via our API.
• Convenient: Paste a YouTube link (any length!) and get a ready-to-post 9:16 short.
• Hooks option: When enabled, adds a context-aware AI-generated hook at the start of the clip
• Web version: Besides the CLI, you can also generate videos on a local website
• Smart Highlight Selection: Finds the most viral, hot moments from your video automatically based on algorithm

How It Works

1. Download: Fetches the source video from YouTube, or uses the local file path directly
2. Transcribe: faster-whisper produces a timestamped transcript locally — same step regardless of which LLM_PROVIDER is selected
3. Detect content type: The chosen LLM classifies the video (podcast, interview, tutorial, vlog, etc.) and pacing, so the highlight prompt can be tuned per content style
4. Highlight ranking: The chosen LLM (openai / gemini / muapi) scans the transcript through a virality framework — hook moments, emotional peaks, opinion bombs, revelations, conflict, quotables, story peaks, practical value — and emits ranked candidates with scores 0–100
5. Dedupe: Overlapping candidates are collapsed by score
6. Top-N selection: The top --n candidates are selected
7. Auto-crop: Each highlight is rendered as a vertical short at the requested --ratio, with an optional AI-generated hook at the start unless --no-hook is set

Requirements

• Python 3.10+
• Any LLM API key(OpenAI/Gemini/MuAPI)
• requirements.txt file dependencies

Quick start

1. Clone the repo:

git clone https://github.com/OStudi/short-video-generator-AI.git
cd short-video-generator-AI

2. Create and activate a virtual environment:

Windows:

python -m venv venv
venv\Scripts\activate

Linux/MacOS:

python -m venv venv
venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Set up .env

# Used LLM provider(openai/gemini/MuAPI)
LLM_PROVIDER=openai

# Enter the API key for the chosen provider
OPENAI_API_KEY=your_openai_key_here
OPENAI_MODEL=gpt-4o-mini          # optional
GEMINI_API_KEY=your_gemini_key_here
GEMINI_MODEL=gemini-2.5-flash      # optional
MUAPI_API_KEY=your_muapi_key_here

# Whisper settings
LOCAL_WHISPER_MODEL=base # tiny / base / small / medium / large-v3
LOCAL_WHISPER_DEVICE=auto    # auto / cpu / cuda

If you uncertain about the provider: | Provider | Free tier? | Get key | |---|---|---| | Gemini | ✅ Yes, but daily limit | https://aistudio.google.com | | OpenAI | ❌ Paid | https://platform.openai.com | | MuAPI | ❌ Paid, but pay-per-use, no subscription | https://muapi.ai |

Usage

Basic usage:

python main.py "https://www.youtube.com/watch?v=video_id" 

Renderred clips are saved to output folder

With flags:

python main.py "https://www.youtube.com/watch?v=video_id" \
      --n 3 \
      --ratio 9:16 \
      --resolution 1080 \

You can also provide a local video file instead of a YouTube link, for example:

python main.py "/Users/Admin/Folder/video.mp4" \
      --n 4 \
      --ratio 9:16 \
      --resolution 720 \
      --language zh \

CLI Flags

| Flag | Default | Notes | |------|---------|-------| | --n | 3 | How many clips to render | | --ratio | 9:16 | Any ratio / 9:16 for short videos / 1:1 for square | | --resolution | 720 | Source video download resolution: 360 / 480 / 720 / 1080 | | --language | auto | Force Whisper language code (e.g. en) | | --no-hook | - | Excludes AI generated hook from the start of the clip |

Web version set-up

Instead of the CLI, you can use a local web interface to queue multiple videos at once and adjust flags visually. The LLM provider and API keys are still configured in .env — the web version only handles video input and render flags.

python3 server.py

Serve the frontend:

cd web
python3 -m http.server 8000

Then, open in your browser:

http://localhost:8000/shorts-generator-ui.html

Contributing

Feel free to open an issue or fork the repository for your projects.

License

This project is licensed under the MIT License.