Note Transcriber
n8n, OpenRouter, AssemblyAI, Claude Opus
Problem
There are many tools online that transcribe online lectures and make notes, but I couldnt find any that could do the same for 5-6 hour long lectures and weren't behind an egregiously expensive pay wall.
To save those long hours of having to watch and comb through lectures, I wanted a tool that could make detailed and to-the-point notes, with the ability to tweak the structures of notes as I wished to. Since I couldnt find any tool to my liking, I decided to make my own on n8n!
Project Structure
My custom n8n workflow works as follows:
1) Upload a dropbox link to the video lecture.
2) The URL is sent to AssemblyAI through an API, where it's transcript is generated.
3) The transcript comes back, and depending on length, it may be chunked or sent as one block to Gemini (with a given system prompt).
4) Gemini then compiles notes as markdown as per it's instructions, generating an HTML that can be viewed and downloaded as a PDF.
Issues:
The massive length of videos (>1.5 Gbs) meant that I couldn't directly upload the lectures to n8n. This also meant that I couldn't provide Google Drive links to the lectures since direct download URLs didn't work.
Some transcripts were very lengthy and couldnt be sent as a complete chunk to Gemini, even with its 1.05M context window, for analysis.
Solutions:
I used Dropbox as a temporary storage for videos, as Dropbox can generate direct download links for videos regardless of file size.
I implemented a smart chunking system, where the rough token count of received transcripts is calculated, and if its above a certain threshold, the transcript is divided and sent chunk-by-chunk to Gemini Flash Preview for analysis.
Result
What used to require manual transcription, note-taking, and formatting across multiple tools now happens automatically; I drop a video in DropBox and get comprehensive PDF notes in under half the time it takes currently existing tools.
The intelligent chunking handles even the longest lectures, while Gemini transforms raw transcripts into structured, readable notes that capture both lecture content and Q&A discussions.
This automation eliminated the tedious task of turning lectures into reviewable material, maximising time to study the notes instead of spending it creating them.