aitubenotes

How do I get notes from a YouTube video?

Paste the video URL into aitubenotes and submit it. The video is downloaded, the audio is transcribed, and the frames are read for text and diagrams. You get back a Markdown table of the notes, a short summary, and the whole run as JSON. Results are emailed when the run finishes, so you can close the tab.

What does aitubenotes read that a transcript tool does not?

The screen. Transcription tools work from audio, so anything a presenter showed but did not read aloud is lost: a slide, a command, an error message, an axis label, a line of code. aitubenotes runs optical character recognition and a vision model over the deduplicated frames and folds that text into the notes alongside the transcript.

Does it work on videos with slides or code on screen?

That is the case it is built for. Dense, visual material is where a transcript alone loses the most: lectures with slides, conference talks, coding walkthroughs, chart-heavy analysis. Frames are deduplicated first, so a slide held on screen for two minutes is read once rather than sixty times, and a static talking head costs almost nothing to process.

How long does a video take to process?

It depends on how much changes on screen rather than on the running time. The work is proportional to the number of distinct frames, so a slide deck shown at a steady pace costs more than a video of the same length pointed at one speaker. Processing runs on a GPU and the result is emailed when it is done.

What do I get back?

Three files, all downloadable from the job page. The notes are a Markdown table with a row per readable paragraph: what was said, the text read off the screen while it was said, and the lesson and action item drawn from the two, against the timestamp it came from. Subtitle cues are merged into paragraphs first, so it reads as a document rather than a subtitle file.