aiVideoSummarizer is a local-first tool that takes a video link, extracts the transcript, and generates concise summaries using a local LLM. The pipeline is designed for privacy, reproducibility, and flexibility, making it useful for research, note-taking, or quickly understanding long-form content.
- Transcript Extraction: Fetches transcripts from supported video sources (YouTube, Vimeo).
- Chunking: Splits transcripts into manageable segments for processing without context loss.
- Local LLM Integration: Summarizes text using a local model via Ollama.
- Customizable Output: Summaries can be short abstracts, detailed notes, or structured outlines.
- Privacy First: No data is sent to external APIs unless configured by the user.
Download and install Ollama from https://ollama.com/download.
python -m venv aiVideoSummarizerEnvWindows:
source aiVideoSummarizerEnv/Scripts/activate # WindowsmacOS/Linux:
source aiVideoSummarizerEnv/bin/activate # macOS/Linuxgit clone https://github.com/yourusername/aiVideoSummarizer.git
cd aiVideoSummarizerpip install -r requirements.txtpython main.py-
Add video URLs (YouTube or Vimeo) to
urls.txt, one per line. -
By default, the script uses the
mistralmodel. If your computer is slow, switch tollama3.2inmain.py. -
Edit main.py to change the prompt, and configure the output of the LLM.
-
Summaries, transcripts, and audio files are saved in the following folders:
./assets/audios./assets/transcripts./assets/summaries
Input: 45-minute technical lecture Output (short summary):
Prompt:
Give me the main focus, what was covered, and all key takeaways from the lecture.
Output:
- Main focus: Basics of signal processing
- Covered: Fourier transforms, FIR vs IIR filters, noise reduction
- Key takeaway: Real-world filtering involves balancing precision vs efficiency
Creative Commons Zero v1.0 Universal – free for personal use.