https://github.com/mqlvin/brain-bites
Shorten educational videos 🎓 into short, rich videos ⚡ using AI 🤖
https://github.com/mqlvin/brain-bites
ai artificial-intelligence content-summarization content-summarizer ffmpeg flask gpt gpt-4 inversity openai openai-api openai-chatgpt python video-processing youtube yt-dlp
Last synced: 3 months ago
JSON representation
Shorten educational videos 🎓 into short, rich videos ⚡ using AI 🤖
- Host: GitHub
- URL: https://github.com/mqlvin/brain-bites
- Owner: Mqlvin
- Created: 2024-11-20T13:30:36.000Z (over 1 year ago)
- Default Branch: master
- Last Pushed: 2024-12-02T09:22:20.000Z (over 1 year ago)
- Last Synced: 2025-07-17T16:53:38.667Z (12 months ago)
- Topics: ai, artificial-intelligence, content-summarization, content-summarizer, ffmpeg, flask, gpt, gpt-4, inversity, openai, openai-api, openai-chatgpt, python, video-processing, youtube, yt-dlp
- Language: Python
- Homepage:
- Size: 113 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Brain Bites
Shorten long documentaries 📽️ into short, education-packed videos ⚡.
## Usage / Running
To run the project, you will need the `openai` and `dotenv` Python libraries.
You can either `pip install` them, or can install them otherwise appropriate to your operating system.
You will also need `yt-dlp`, `ffplay` and `ffmpeg` available on your system's PATH, so it is executable from the terminal.
The project will require an environment variable keystore (`.env` file) with the following keys:
- `OPENAI_KEY` - Your OpenAI API key.
- `RUNTIME_DIR` (optional) - The runtime directory, defaults to `./runtime`
## The Vision
The current project trajectory is to use AI and LLM to trim YouTube videos from lengthy documentaries into short and fast paced educational videos.
This is to make videos more engaging and practical for usage, such as in a classroom, where time and attention spans are limited.
The content is implemented in a frontend, scrolling-style web application.
## The Process
- We feed **OpenAI's GPT-4** with YouTube transcripts, asking the LLM to regenerate chapterized, concise transcript.
- A comparison between the full and trimmed transcript will occur using a custom text-comparison algorithm, finding the removed sentences from the transcript.
- The video will be split into chapters based on the shortened transcript using **ffmpeg**.
- A quiz and chapter summaries are also generated to aid with learning, again using **OpenAI's GPT-4**.
- The content generated is served on-demand by the frontend
## Inversity
This is an [Inversity](https://inversity.co/) project application.
We were tasked with "How can AI make education more engaging, effective and equitable?"