YouTube Transcript
The YouTube Transcript screen fetches the captions a YouTube video already has and turns them into a readable transcript. It is the one screen in the app that works without a licence, and it takes seconds rather than the minutes a full transcription needs — because nothing is transcribed, only retrieved.
Getting a transcript
Paste a YouTube link into the box at the top of the screen and click Get transcript. Links in any of the usual shapes work — youtube.com/watch?v=…, youtu.be/…, Shorts, embed and live URLs — and a bare video ID is accepted too. Extra parameters such as &t=90s or a playlist ID are ignored; only the video itself is fetched.
The app looks the video up, shows its title, channel, duration and thumbnail, and picks the most useful caption track automatically: your transcription language from Settings if the video has it, then English, then any human-written track, then an automatic one. If the video offers more than one track, a dropdown appears beside the button so you can switch — the transcript is re-fetched for the track you choose.
The first run downloads a helper
YouTube no longer serves caption data to plain web requests, so the app uses yt-dlp to fetch it. The first time you use the screen this is downloaded automatically — about 35 MB, once — and the button says Install & get transcript so you know it is coming. You can also install it ahead of time from the callout on the screen, or from the Dependencies screen where it sits alongside FFmpeg and the other managed tools.
Only the caption file is downloaded. The video itself is never fetched, which is why a transcript arrives in seconds regardless of how long the video is.
Human-written vs automatic captions
Tracks a person wrote and uploaded are listed first, under the name YouTube gives them (for example English - CC). Tracks YouTube generated by speech recognition are marked (auto-generated).
The difference matters. Human-written captions are punctuated and usually accurate. Automatic captions carry no punctuation or capitalisation at all, and YouTube repeats each line as it scrolls up the screen — the app removes those repeats exactly, so a track that arrives as 1,300 fragments becomes the ~650 lines that were actually spoken.
YouTube also offers machine translations of its automatic captions into 150-odd languages. These are deliberately left out of the list: they are machine translations of a machine transcription, and the app's own Translate screen does a better job from a clean source.
Reading and using the transcript
Captions are written to be read a line at a time on screen, so they arrive chopped into short fragments. The transcript view stitches them back into paragraphs using sentence endings, pauses in the speech and a maximum length, which is what makes the result readable rather than a list of scraps.
The Timestamps toggle shows or hides the time beside each paragraph. Clicking a timestamp opens the video at that exact moment on YouTube.
| Copy | Copies what is on screen — with timestamps if they are shown, plain prose if they are not. |
|---|---|
| Save | Writes an .srt, .vtt or .txt file. SRT and VTT keep the original caption timing, cue for cue. TXT saves the readable paragraphs. |
| Open in Editor | Saves the transcript, then opens it in Edit Subtitles as a subtitle-only file, ready to correct, clean up or translate. |
Saved files are named after the video with the language appended — Video title.en.srt — so the language is picked up automatically if you later translate the file.
Using Whisper instead of YouTube's captions
The source dropdown lists the video's caption tracks and, at the bottom, AI transcription (Whisper). Choosing it downloads the audio and transcribes it locally instead of using anything YouTube provides.
This is worth reaching for even when the video does have captions. YouTube's automatic captions carry no punctuation or capitalisation and are often wrong on names and technical terms; Whisper produces punctuated, properly segmented text. The trade is time — minutes rather than seconds — plus an activated licence, FFmpeg and a Whisper model. Fetching captions stays free either way.
Videos with no caption track at all have Whisper as their only option, and the screen says so.