Generate captions for a video
Drop in a video and get timed captions without uploading it anywhere. Save an SRT or VTT file, or send the words to the subtitle editor and burn them into the picture.
No file handy? . It runs entirely on this device.
About this tool
Captions decide whether a video gets watched. Most people scroll with the sound off, so a clip with words on screen keeps its audience and a clip without them does not. This tool listens to your video and writes those words with timings attached, using a speech model that runs inside your browser instead of on a server. Nothing is uploaded, so there is no queue, no file-size limit set by a server and no account.
You get two ways out: a subtitle file (SRT or VTT) for YouTube or another editor, or the video and its captions together in the subtitle editor, where you can style them and burn them permanently into the picture.
How to generate captions for a video
- Open your videoDrop an MP4, MOV, WebM or an audio file into the box. The file stays on your device.
- Pick a model and languageLeave the model on Balanced for most videos. Choose Accurate for quiet recordings or strong accents, and Fast for a quick draft. Choose the language spoken in the video.
- GenerateThe first run downloads the speech model once and the browser keeps it. After that, jobs start without downloading again.
- Check and exportRead through the timed lines and fix any word. Download an SRT or VTT file, or press “Open in subtitle editor” to style the captions and burn them in.
Features
- Speech recognition with Whisper that runs on your device, in the background so the page stays responsive.
- Three model sizes: Fast (42 MB), Balanced (76 MB) and Accurate (240 MB). They run on WebAssembly, so they work in any recent browser.
- Eleven languages to choose from, including English, Urdu, Hindi and Arabic. English clips use a faster, more accurate English-only model.
- Captions are split on natural sentence boundaries and wrapped to the line length you choose: 32, 42 or 56 characters.
- Every line is editable before you export.
- Export SRT, VTT or plain text. One press sends the video and its captions, timings included, into the subtitle editor.
- No account, no export limit, no watermark.
Supported formats
- Video in
- MP4, MOV, WebM, MKV and most files a browser can play.
- Audio in
- MP3, WAV, M4A, AAC, Ogg and FLAC. You can caption an audio file, but only a video can be sent on to the subtitle editor.
- Captions out
- SRT (with numbering and timings), VTT (with a WEBVTT header) and plain text, all UTF-8.
- Burn-in
- Done in the subtitle editor, which exports MP4 at Auto, Original, 720p or 1080p.
Tips
- Set the language to the one spoken in the video. A clip in another language will come out wrong if English is selected.
- Two people talking over each other is the biggest cause of wrong words. Where it matters, cut the clip at the overlap and caption each side separately.
- Background music and noise hurt accuracy. If the first result is poor, try the Accurate model.
- Long files take longer and need more memory. Trim a long recording first, or work on a computer rather than a phone.
- Always read the captions before you publish. Automatic captions are a first draft, not a final one.
Questions
Is this caption generator really free and watermark-free?
Yes. There is no account, no export limit and no watermark. The speech model runs in your browser on your own device, so there is no server bill to pass on to you.
Are my videos or audio uploaded to a server?
No. Your audio is read and transcribed inside your browser. The only thing downloaded is the speech model file itself, once. You can confirm this in the browser’s Network tab while a job runs.
Why does the first run download something?
The speech model has to reach your device before it can work. The default Balanced model is a 76 MB download. The browser keeps it after the first download, so later jobs skip the wait.
Can I get burned-in captions instead of a separate file?
Yes. Press “Open in subtitle editor” after generating captions. The video and its captions arrive in the editor with their timings, ready to style and export as an MP4 with the captions drawn on the picture.
Which formats can I export?
SRT, VTT and plain text. SRT and VTT open in YouTube, Premiere, DaVinci Resolve, Final Cut, CapCut and VLC.
Does it work with Urdu, Hindi or Arabic speech?
Hindi, Arabic and dozens of other languages work with the multilingual models. Urdu recognition is limited by the model itself, so expect to check and correct Urdu captions. Right-to-left text is handled in the subtitle editor.
How accurate are automatic captions?
Clean speech with little background noise is usually close to correct. Accents, overlapping speakers, music and crosstalk produce errors. Always read through before publishing; every line is editable.
Can I caption a long video?
Yes, but it runs on your device, so a long file takes proportionally longer and needs more memory. Phones are happiest with clips of a few minutes; for a long recording use a computer.