# Videos, lectures and podcasts

> Turn what you watch and listen to into materials, with captions or on-device transcription.

Source: https://cognipage.io/docs/watch-along

Much of what people learn now is watched rather than read. So when a window you approved starts playing, CogniPage captures **what is said** instead of the page around it — and the lecture becomes a material like any other, with the same flashcards, the same recall schedule and the same place on your knowledge map.

**Where** Settings → Watch along · on by default

## How it works

1. **Play something in a window CogniPage can see**
   A lecture, a course video, a podcast episode — in a browser or a local player.
2. **Approve that specific piece of media**
   You are asked per video, not per site. Approving a lecture says nothing about the next video on the same platform: that gets its own question.
3. **The words become a material**
   Captions are read where the service publishes them — instantly, and at no cost. Where none exist, CogniPage can listen and write the transcript itself, on your computer.
4. **Adverts are left out**
   Ad breaks are detected and skipped, so a sponsor read never turns into a flashcard.

## What is recognised

- Course platforms — Coursera, Udemy, edX and any university running Open edX, Udacity, FutureLearn, Khan Academy, Alison, NPTEL
- Professional — LinkedIn Learning, Pluralsight, Skillshare, O'Reilly, MasterClass, Teachable, Thinkific
- Lecture capture — Panopto, Echo360, Microsoft Stream, Zoom recordings, Loom, Wistia, Kaltura
- Video & audio — YouTube, Vimeo, TED, Dailymotion, Spotify and Apple podcasts, Overcast, Pocket Casts
- On your machine — Any local video or audio file in VLC, IINA, mpv, QuickTime, Music or Podcasts

The pages you *read* around a course — the syllabus, the notes, the quizzes — keep being captured as reading, exactly as before.

## On-device transcription

**Transcribe when there are no captions** is off until you switch it on. Turn it on and CogniPage listens to the audio the approved application is playing and writes the transcript itself. Everything happens on your computer:

- No audio and no transcript ever leaves the machine.
- It costs no
  [AI credits](https://cognipage.io/docs/account)
  .
- No audio file is written to disk — only the resulting text is kept.
- Nothing is downloaded, ripped or saved from the service you are watching.

### Choosing a speech model

Switching transcription on reveals the model list: **Fastest** (32 MB), **Balanced** (57 MB, the default) and **Most accurate** (181 MB). A bigger model is more accurate and slower. Each shows whether it is downloaded, and you download or delete them here — nothing is fetched until you ask for it, and you can delete a model again at any time to reclaim the space.

> **Your microphone is never opened.**
>
> CogniPage requests no microphone permission at all. It taps the audio that one specific application you approved is already playing — a call, a notification or anything else on your machine is not part of it.

## Turning it off

Switch off **Capture what you watch and listen to** and CogniPage goes back to reading only what is written on screen. Transcription can be left off independently, in which case captions are still used wherever they are published.