Video to Text
Paste the link to a video and read what was said. Each speaker separated, every line timed, and subtitles you can download — from YouTube, TikTok, Instagram, Vimeo, X, or a direct link to a file.
A page with a video on it, or a direct link to a video or audio file. Up to 30 minutes long.
How to turn a video into text
- 1Paste the link. A YouTube watch page, a TikTok, an Instagram reel, a Vimeo page, an X post, a Loom recording — or a direct address ending in .mp4, .mov, .mp3 or .m4a. If it plays in a browser it is worth pasting.
- 2Leave the language alone, usually. Detect automatically reads the spoken language off the audio itself and is right nearly always. Naming it yourself helps for languages a detector confuses — Hindi against Urdu, one Arabic dialect against another.
- 3Press Transcribe and leave it. A two-minute reel is done in seconds; a half-hour interview takes a minute or two. The line shows which stage it is at — reading the link, pulling the audio, transcribing — so you can tell it is working.
- 4Read it, copy it, or download it. Plain text for a document, text with timestamps for quoting a moment, .srt for a video editor, .vtt for HTML5 video on a web page.
What you actually get
Not a wall of lowercase words. The transcript comes back punctuated, capitalised and paragraphed, with numbers written the way a person would write them — so it reads like something typed up rather than something a machine coughed out. Fillers like “um” and “uh” are left out, because nobody wants them in a written transcript.
Where more than one person is talking, the turns are separated and labelled Speaker 1, Speaker 2 and so on. It tells voices apart; it has no way of learning their names, so it does not pretend to. For an interview or a two-host podcast that distinction is the difference between a usable transcript and a paragraph that contradicts itself.
Every turn also keeps the second it was spoken at. That is what makes the subtitle downloads possible, and it is why the Timestamps button is worth pressing when you are hunting for the one sentence you half-remember from forty minutes of talking.
What it is good for
- Captions for your own videos. Transcribe the reel you just posted, download the .srt, and drop it onto the footage — captioned video is watched further by people scrolling with the sound off.
- Reading a long video quickly. An hour-long podcast is eight thousand words you can skim in two minutes. Searching text beats scrubbing a timeline.
- Quotes you can trust. Timestamps mean you can check the exact moment before you quote anyone.
- Turning talk into posts. One recorded conversation is a week of captions, a newsletter and a blog post, once the words exist as text.
- Notes from a lecture, a webinar or a meeting recording — anything where the information was spoken and you need it written.
- Watching what a competitor is saying, not just what they captioned. Pair it with the free engagement checker to see which of their videos earned the attention.
Accuracy, honestly
On clear speech it is very good — better than most people expect, and punctuated properly, which is the part cheap tools get wrong. It gets worse exactly where a human transcriber would: loud background music, several people talking over each other, a phone held at arm’s length in a restaurant, a heavy dialect, or technical jargon and brand names it has never seen spelled out. Proper nouns are the most common thing to fix.
So read it before you publish it. It is a first draft produced in a minute instead of an hour, not a finished document — and that is still the whole saving.
What happens to your video
Nothing of the video is kept. For a direct file link, the file is never downloaded here at all — it is read where it sits. For a page, the audio alone is pulled, converted to a small speech-quality stream, transcribed, and deleted the moment the words exist. No copy of anybody’s video lives on this site.
What is kept is the transcript, in your own account, so you can come back to it. Delete it whenever you like and it is gone. Nothing is shared with anyone, and nothing is used to train anything.
A note on what is yours to transcribe: this tool reads what is publicly available at a link you give it. Whether you may then republish someone else’s words is a question about their copyright, not about this tool — quoting a video with a credit is ordinary; reposting a whole transcript as your own is not.
When a link will not work
A handful of cases genuinely cannot be read, and the message says which rather than failing silently:
- Private, unlisted-with-a-sign-in, or members-only videos. If a browser would ask you to log in, so would we — and we do not have your login.
- Live streams still in progress. Transcription works on a finished recording; come back when it has ended.
- Videos the site has removed, or links that were mistyped.
- “Prove you are not a robot.” Big video sites sometimes challenge requests coming from a data centre rather than a home broadband line. There is nothing wrong with your link; try again later, or use a direct link to the file.
- Copy-protected video (a paid streaming service). Its audio is encrypted and cannot be read by anything but its own player.
- A playlist or a channel address. Paste the link to one video — guessing which of forty you meant would be worse than asking.
Free, with a sensible limit
Two transcriptions a day, up to thirty minutes of video each, on a free KEMETORY account. The account is not a marketing exercise: unlike the PDF resizer and the QR generator, which run entirely in your own browser and cost nothing to serve, transcription is billed by the minute of audio — so it has to be attached to somebody. A KEMETORY app account raises the ceiling to four hours a video, which covers a full podcast episode or a conference recording.
Questions about transcribing video
Is it free?
Yes. A free KEMETORY account gets two transcriptions a day, up to 30 minutes of video each. The account exists because transcription costs real money per minute of audio — it is what stops a stranger spending the budget by refreshing.
Which sites does it work with?
YouTube, TikTok, Instagram, Facebook, Vimeo, X, Twitch, Dailymotion, SoundCloud, Reddit, LinkedIn, Loom and many more, plus a direct link to a video or audio file (.mp4, .mov, .webm, .mp3, .m4a, .wav). If a page has a video on it, paste it and see.
Does it tell speakers apart?
Yes. An interview comes back split into turns labelled Speaker 1, Speaker 2 and so on. It separates voices — it has no way of learning their names, so it does not guess them.
Can I get .srt or .vtt subtitles?
Yes. Every turn keeps the time it was spoken, so the result downloads as .srt for a video editor or .vtt for HTML5 video, ready to drop onto your footage.
What languages work?
Dozens, including English, Arabic, Spanish, French, German, Italian, Portuguese, Dutch, Russian, Turkish, Hindi, Indonesian, Japanese, Korean and Chinese. Leave the language on Detect automatically and it is read off the audio.
How long does it take?
A short reel is seconds. A half-hour interview is a minute or two — most of which is pulling the audio down, not transcribing it. The line on screen says which stage it is at.
How accurate is it?
Very good on clear speech, with real punctuation and capitals. Background music, crosstalk, a distant microphone or heavy dialect all cost accuracy, and unusual names are the most common thing to correct. Read it before publishing it.
Do you store the video?
No. A direct file link is never downloaded at all; for a page, only the audio is pulled and it is deleted the moment the transcript exists. The transcript stays in your account until you delete it.
Why did my link fail?
Usually a private or members-only video, one the site has removed, a live stream that has not finished, or a site challenging a request from a data centre. The message says which. A direct link to the file itself avoids nearly all of these.
Can I transcribe my own recording?
Yes, if it has a web address — a public link from cloud storage, or any direct URL ending in .mp4 or .mp3. Paste the address and it is read straight from there.