Summarize with local AI
- 1 min read
In this tutorial, I’ll present a simple way to auto-transcribe and summarize your media using local AI.
Example use
Create a summary of a podcast episode and a transcript to view before listening. Figure out if you want to spend next N hours on this episode and if it delivers what you want.
Required tools
- Python3
- Whisper
- KoboldCpp
- Local LLM, eg. Gemma 27b.
- Prefect
Architecture
We will be using Prefect to orchestrate the actions.
Transcribing the media
Whisper is used to generate .txt files with the transcript from audio/video file.
Summarizing the transcript
We will connect to a local LLM using KoboldCpp API to ask a local LLM to summarize the transcript. You can alternatively use the API of your favorite LLM provider to do the same.
Storing the summary
The summary will be stored locally in the MD file.