This is a submission for the AssemblyAI Challenge : Sophisticated Speech-to-Text
What I Built
Scribe
Scribe is an Obsidian (an incredible graph based note taking application) plugin that not only records your voice and transcribes it, but answers questions with "Hey Scribe", summarizes what was said, pulls out insights and graphically maps out the main points.
Repo
https://github.com/Mikodin/obsidian-scribe
Key Features
- Robust voice recorder
- Handles network failures with grace
- Has high compression for cloud storage
- Doesn't lose audio files when something goes wrong
- Can pickup from any audio file
- Voice to text raw transcription & formatting with AssemblyAI
- Summarize, get insights, and chart out what was said with ChatGPT (sorry not LeMUR - maybe coming soon)
- Clean and simple UI
Demo
Stack
Essentially the Obsidian app is the stack. It's an electron app with an extensive ecosystem.
Obsidian Dev Docs Here
I was able to leverage React along with the MediaRecorder Api.
Journey
This is something that I've wanted for quite some time. I've been trying several transcription plugins over the last year in Obsidian, but none of them do what I wanted. They were either fickle, save multi MB audio files, or were too basic.
Scribe is my attempt at solving for this.
May it be of use for you!
Thank you to dev.to, and AssemblyAI - it's an honor to submit something here.
Top comments (2)
Definitely gonna be using this! Obsidian also my main note taking app! will take advantage while my $50 credit last š
Glad to hear!!
Iām super open for feedback as well, let me know what ya think!