[[Projects]] "Thunder" is a tiny script that transcribes voice memos (long form) and adds it to Obsidian as a dedicated note. The name "Thunder" was inspired by the [countersign](<https://en.wikipedia.org/wiki/Countersign_(military)>) the allied forces used during World War II (Thunder... Flash!). The idea being once "thunder" was called, the response (flash) would be a transcribe of that call. ## How it works ### Record A [macOS Shortcut](https://support.apple.com/en-ca/guide/shortcuts-mac/apd5ba077760/6.0/mac/13.0) records a voice snippet from my microphone. ![[2023-01-22 - Thunder - Shortcut.png]] That audio file is saved into a designated folder ("Voice memos") on my computer. ### Transcribe A [macOS Automator](https://support.apple.com/en-ca/guide/automator/welcome/mac) workflow executes a shell command (`thunder`). It is triggered when a new file is added to "Voice memos". ![[2023-01-22 - Thunder - Automator.png]] `thunder` uses [Whisper by OpenAI](https://openai.com/blog/whisper/) to transcribe the recorded audio snippet, generating a `.txt` file. ### Summarize `thunder` uses [GPT-3 powered by OpenAI](https://beta.openai.com/docs/guides/completion) to summarize the voice memo. This summary is added to the top of the note under the section "Summary (AI)". A title is generated from the summary, which is used for both the title and file name of the markdown file. ### Note The prepared markdown file (summary and transcribe) is added to my Obsidian in a designated folder for further processing. ## Source code The source code is private and the link below is for personal reference only. https://github.com/ItsJonQ/thunder ## See also - [[Vox]] - [[Grams]]