All documentation

Auto-captions & Class Notes

One click after a class ends turns the recording into searchable captions plus a clean revision document.

Tutors · Institutions7 min read
Who this is for
Tutors and institution admins who record live classes and want the recording to be more useful to absent or revising students.

Overview

After a live class ends, the recording is uploaded to GCS. Once the recording URL is available, you can trigger captioning. The pipeline is:

  1. Mark captionsStatus = PROCESSING on the class.
  2. Stream the recording into OpenAI Whisper (whisper-1 with verbose_json + segment timestamps).
  3. Convert Whisper segments into WebVTT.
  4. Store the VTT + transcript URL on the class. Mark captionsStatus = READY.

On failure the status flips to FAILED, the original recording is unaffected, and you can retry.

Triggering caption generation

  1. Open the class detail page from your tutor dashboard.
  2. If the recording is uploaded, you'll see a Generate captions button next to the playback controls.
  3. Click it. The button changes to Processing… and the API call returns immediately — the actual transcription runs in the background.
Re-running is safe
If status is already PROCESSING, the API returns "already generating" without starting a second job. To force a re-transcription after failure, click again.

Status & timing

  • NONE — captions never generated.
  • PROCESSING — Whisper job in flight.
  • READY — VTT and transcript are saved.
  • FAILED — see the error log; retry is safe.

Typical timing: a 60-minute class transcribes in 3-6 minutes. Cost is a small Whisper API charge (~$0.36 per hour of audio at current OpenAI rates).

Caption format (WebVTT)

Captions are stored as standard WebVTT — one segment per Whisper chunk, timestamps formatted 00:MM:SS.sss. Example:

WEBVTT

00:00:00.000 --> 00:00:04.520
So today we're going to look at Newton's first law…

00:00:04.520 --> 00:00:09.140
An object at rest stays at rest unless acted on by a force.

Where students see captions

Once the status is READY, the class video player picks up the VTT track automatically and shows a "CC" toggle. Students can also download the full transcript from the class page.

AI-generated class notes

Captions are the input to AI study notes: a separate pipeline takes the transcript and produces a structured revision document (sections, bullet points, key formulae) ready for students to download as PDF or read inline.

Class notes are auto-generated for every class with a transcript; you can also trigger a regeneration from the class page if you've edited the transcript or want a fresh pass.


Header Logo