Generate subtitles for a video, automatically
No typing and no timing by hand: pick a video, tap Get the words, and download captions that follow what's said, in the video's own language.
Make subtitlesFree, no sign-up. Your video stays on your device; only the audio is sent, when you tap Get the words.
Transcript
…so we recorded the whole thing live, in one room, in one afternoon. That's why it sounds the way it does.
How it works
- 1 / 3
Pick the video
Any MP4, MOV, WebM or MKV, up to 500MB and 20 minutes. Trim to the part you need.
- 2 / 3
Extract, then tap Get the words
The audio is pulled out in your browser, then what's said is written down in a few seconds.
- 3 / 3
Tap Subtitles
Download an SRT or VTT file. If you trimmed the video, pick whether it's timed to the whole video or to your cut.
The language is detected for you. Clear speech works best; loud music or people talking over each other make it harder.
What people use it for
- Caption a reel, short or TikTok before you post it.
- Subtitle an interview or a talk.
- Make a tutorial easier to follow with the sound off.
- Start from a draft and fix the few words it misheard.
Questions
How do I generate subtitles for a video?
Pick the video on Extractune, extract the audio, and tap Get the words. When the transcript appears, tap Subtitles and choose SRT or VTT.
What's the difference between SRT and VTT?
Both are plain-text subtitle files with the same lines and timings. SRT works almost everywhere, including YouTube and most editors; VTT is the web's own format, for HTML video players.
Are the subtitles timed to my video?
Yes. Each caption follows when it's said, split into short lines that are easy to read. If you trimmed the video first, you choose whether the times match the whole video or just your cut.
Which languages work?
Most widely spoken languages. The language is detected from the audio itself, so there's nothing to choose.
Does my video get uploaded?
No, the video never leaves your device. When you tap Get the words, the extracted audio (compressed, never the video) is sent to be transcribed, and a 20-second clip to check whether it's a known song. Extractune doesn't keep either after the request.
More ways to use Extractune
Updated