Add subtitles to a video, free
Drop in a video or a recording: the speech recognition runs in your browser, you can rewrite the text afterwards, then download it as an SRT or burn it into the picture.
Drop your video or audio file here
MP4, MOV, MKV, MP3, M4A, WAV and the rest — up to 500 MB and 2 hours. The file stays in your browser; we never upload it.
How are the subtitles made in your browser?
In three steps. First we pull the sound out of the video and convert it to 16 kHz single-channel audio — the format the speech-recognition model works on. Then we cut that audio into thirty-second windows and run each one through Whisper, an open speech-recognition model that returns not only the words but when they were spoken. Finally the text and the timestamps are assembled into subtitles — which is what you see in the editor.
All of it happens on your own machine. The file never reaches a server, and neither do the subtitles: once the page has loaded you could pull out the network cable and the processing would still finish. The price is the first run — the model has to be downloaded, and it is several tens of megabytes. After that it stays in your browser cache and the next video never repeats it.
What it is actually good for
Social video with the sound off. Most Reels and TikTok viewers scroll in silence. Burned-in subtitles work there because they cannot be switched off, and there is no player that fails to show them.
Lectures and recordings. An hour-long talk becomes searchable text in a few minutes. The plain-text export exists exactly for that: paste it into your notes and find where the part you needed was said.
Accessibility. YouTube, Vimeo and most players read an SRT directly, so the video becomes followable for deaf and hard-of-hearing viewers — without re-encoding the picture.
Where its limits are
Honestly: this is the smallest Whisper model, because it is the only one that fits in a browser. On clean, close-to-studio English it is very good; on Hungarian it is noticeably weaker — it often mangles proper nouns and technical terms, and the punctuation is hit and miss. That is why the page has an editor: we assume you will read it through, not that it arrives finished.
A noisy recording, people talking over each other, a loud music bed or a strong accent all make the result worse. It does not tell speakers apart. The speed depends on your machine: five minutes of audio takes twenty to thirty seconds on a current desktop and several times that on a phone.
Frequently asked questions
Is it really free, and is there a minute limit?
Free, with no sign-up, and no monthly minute allowance — the processing runs on your machine, so it costs us nothing. One file can be up to 500 MB and 2 hours.
Does my video get uploaded anywhere?
No. Not the video, not the audio, not the finished subtitles. It is only uploaded if you click “Save to FileMv” yourself, to get a link to it.
How big is that first download?
The speech-recognition model and its runtime come to 54–72 MB together, depending on whether your browser can use the graphics card. It is a one-off: after that it comes from the cache. Burn-in additionally needs a video encoder, another 31 MB — but only if you actually burn subtitles in.
SRT or burned-in subtitles?
SRT if you are uploading to YouTube or Vimeo: there the subtitles can be switched off, they are searchable, and the picture stays untouched. Burned-in if it is going to TikTok, Reels or Facebook, where the viewer cannot switch subtitles on.
How accurate are the Hungarian subtitles?
On a clean, single-speaker Hungarian recording most sentences are usable, but expect typos, missing accents and imprecise punctuation. Every line can be fixed in the editor, and the timing is generally more accurate than the text — which makes fixing it quick.