LectureScribe AI
Turns long lecture recordings into synced transcripts and study notes, fast and close to free.
- Domain
- Transcription
- Built with
- Python, FastAPI, Groq Whisper
- Source
- github.com/notsointresting/lecturescribe-ai
The problem
Reviewing hours of lecture audio is slow, and good transcription is either expensive or a pain to set up locally.
How it works
- A small FastAPI back end and a zero-dependency vanilla JS front end with no build step.
- Whisper transcription and translation on Groq’s LPUs.
- FFmpeg extracts and chunks audio for large files.
- A pool of API keys with round-robin balancing and automatic failover on rate limits, with progress streamed over Server-Sent Events.
Where it landed
Exports SRT, VTT, TXT, JSON and Markdown study notes, with click-to-seek playback and vocabulary presets for CS, maths, medicine and business.
Stack
- Python
- FastAPI
- Groq Whisper
- SSE
- FFmpeg
- Vanilla JS