I use Audio Hijack recording meetings and remote podcasts from time to time, given the latest version, the `Transcribe` block now can generate a meeting script after each meeting. And with LM Studio’s local run server, I wrote a .py script to call LLM loaded in LM studio to summarize the meeting with TLDR, Action Items and Open Items.
You can give a try at this Repo: klynwuu/ollama-lm-studio-meeting-summarizer
As long as
- You have a generic python(3.10+ should be good) environment with common packages installed + `pip install openai`
- This script can handle .srt, .txt and .md files, as for how much words you need process depends on the model setting in lm studio, usually default context length is 4096 and based on your memory like mines is 128gb, I can process unto 13K+-ish words.
- My script is actually constructed in a way to be able to process various text tasks, you can modify both system prompt and user prompt in `prompt` folder.
Feel free to give a forkAfter run command `python3 meeting_summarizer_v1.py`, it will generate a meeting summary.md file in the same folder/TARGET_FOLDER where your meetings scripts reside.
Raycast Script Command
I also converted the script into a Raycast script command.py (basically by putting .env and prompts into one .py file)
So, everytime when I’m done with one meeting, I will use raycast quickly generate a meeting summary.md like below
