StreamSnag Get StreamSnag

Launching soon. StreamSnag 1.6 is not released yet, and Pro is not on sale yet.Launching soon. 1.6 is not released. Pro is not on sale.

StreamSnagGuidesTranscribe a lecture

How to transcribe a lecture on your own computer

A transcript lets you search a lecture, skim it, and find the minute where something was explained. Here is how to get one when the lecture has captions, and when it has none.

From captions: freeNo captions: ProNo AI account either way

First, check for captions

Look for a CC or subtitles button in the player. Captions made automatically count too. In StreamSnag, a lecture with captions has a Save transcript button under its name, and a “Captions” line lower down.

Some course sites show their own transcript panel beside the video. If yours does, you can read it there. What you usually cannot do is keep it as a file next to the lecture, with times you can jump to. The steps below give you that.

If the lecture has captions

FreeNo AI

  1. Open the lecture and press play

    Players load the video and its captions when playback starts. A few seconds is enough.

  2. Click the StreamSnag icon and press Save transcript

    The button is under the lecture's name. If you use the lecture's path instead, it is the Transcript step, which reads “From its captions”.

  3. Open the file next to the video

    The popup says “Transcript saved.” and names the file, with Show file beside it. It is a text file named after the lecture: Markdown with times, or plain text if you prefer. The format is in Settings, under Downloads.

Captions are cut into short lines for the screen, and automatic ones repeat each line as it scrolls. StreamSnag joins them into paragraphs, drops the repeats, and puts the time at the start of each paragraph. No AI is involved, and the words are exactly the captions' words.

If there are no captions

ProNo AI account

StreamSnag Pro can transcribe the lecture on your own computer. The speech-to-text runs on your machine, the sound is never uploaded, and no AI account is used. Because nothing goes through us, there are no minutes to count: transcribe as many lectures as you like.

  1. Set up the helper

    Transcribing needs the optional helper program. On Windows, double-click Set up StreamSnag helper.cmd in the StreamSnag folder. On macOS and Linux, run sh setup.sh there. Then restart the browser.

  2. Set up transcription, once

    In Settings, open Local helper and press “Set up transcription”. The model it comes set to, Small, is for English: about 135 MB, one time, the speech-to-text program included. It runs on your computer from then on. There are two other choices. Tiny is the smallest download, about 55 MB, and can repeat phrases on noisy recordings. Best, about 515 MB, covers 25 European languages and finds the language by itself.

  3. Save the lecture

    Transcription works on the saved file, so download the video first.

  4. Press Transcribe on this computer

    Once the lecture is saved, the popup shows a card headed Transcript: “No captions here. Turns the speech into text on this computer and saves it next to the video. Nothing is uploaded.” Its button is “Transcribe on this computer”. The transcript is saved next to the video, in the same format as one made from captions.

How long does it take, and how good is it?

An hour of lecture took under a minute on our fast desktop. An older laptop can need 3 to 10 minutes for each hour. How good it is depends on the recording. In our own test on real lectures, Small made the fewest mistakes of the three. What every one of them gets wrong is names and rare terms: see check the names and terms.

If you need transcription because of a disability or another access need and cannot pay for Pro, write to support@streamsnag.app and we will send a key.

Make it automatic

Pro

If you save a lecture every week, Pro can do this without being asked. In Settings, under “After a download finishes”, choose what should happen: save the transcript, transcribe when there are no captions, and go on to the reading edition, the study page and the study pack. Everything there is off until you turn it on. On the lecture's path the same choices are one switch, called Autopilot.

Check the names and terms

Automatic captions and speech-to-text both mishear names and technical words. Speech models of this size misspell unusual drug and technical names. In our sample lecture the captions wrote “Frank Sterling” where the slide says Frank-Starling. Before you rely on a term, check it against the slides. If you make a reading edition and have it tidied with your own AI account, the tidy corrects misheard terms against the slides and lists what it changed.

The time at the start of each paragraph helps here. If a sentence looks wrong, go to that minute and listen.

What to do with a transcript