At a glance
- Accepts
- MP4, WebM and MOV directly; AVI, WMV, FLV, 3GP, MPEG-TS and MKV or MOV with AC-3, E-AC-3 or ALAC audio through an optional one-time FFmpeg download (about 10 MB) that also runs in the browser
- Outputs
- A timestamped transcript you can copy, or download as .txt, .srt, .vtt or Markdown notes (.md); optional speaker labels (Speaker 1, Speaker 2, renamable), as WebVTT voice tags in .vtt
- Where it runs
- Entirely in your browser, on your device
- What leaves your device
- Nothing - the file never leaves your device
- Limits
- First run downloads the chosen Whisper model, from about 40 MB to several hundred; you choose the spoken language (no automatic detection); formats the browser can't read need the FFmpeg converter, which takes videos up to 4 hours on a computer and 90 minutes on a phone; optional speaker labels can be wrong where people talk over each other or sound alike; on-screen text is not read
- Account
- Not required - there is no sign-up
- Price
- Free, with no watermark and no usage cap
- Made by
- Visrut
- Last updated
- 11 October 2026
This turns the speech in a video into text and subtitle files without sending the video anywhere. Your browser extracts the audio track itself, and OpenAI's Whisper model, running inside the tab, transcribes it with timestamps. That makes it suitable for the videos you would rather not upload: unreleased content, client recordings, internal meetings, lectures with students in them.
How to use it
- Drop in a video: MP4, MOV and WebM are read directly; AVI, WMV, FLV, 3GP and unusual MKV files go through a one-time converter download.
- Pick a model. Tiny English is enough for clear English speech; move up for accents, music under the dialogue, or other languages, and choose the spoken language.
- Press transcribe, then copy the text or download it as plain text, Markdown notes with timestamped headings, SRT subtitles or WebVTT subtitles.
SRT or VTT - which subtitle file?
SRT is the one almost everything accepts: YouTube, Vimeo, Premiere Pro, DaVinci Resolve, Final Cut (via import) and desktop players like VLC. WebVTT is the web's native format, the one an HTML5 <track> element loads. The timings and text are identical; only the layout of the file differs.
Honest limits
- Only the audio is used. Text that appears on screen is not read.
- Whisper sometimes invents words over silence or music. Speaker labels are optional and can be wrong where people talk over each other or sound alike. Check subtitles before publishing them.
- Subtitle cues follow Whisper's phrases, which can run longer than a subtitle style guide would allow.
- Browsers that can't read the audio a piece at a time (Safari before version 26, Firefox on Android, some formats) read the whole video into memory instead. Low-memory mode avoids that but has to play the video through (at double speed in Chrome and Edge, normal speed in Safari), and the tab has to stay open while it plays.
- Formats the browser can't read need the converter (about 10 MB, once). It holds the whole soundtrack in memory, so it takes videos up to 4 hours long on a computer and 90 minutes on a phone.
- The first run on a model downloads it. Tiny is tens of megabytes; large v3 turbo is over 1 GB and is not offered on phones.
Frequently asked questions
Is my video uploaded anywhere?
No. Your browser pulls the audio track out of the video and Whisper transcribes it inside the tab. The only download is the speech model itself, the first time you use it - the video never leaves your device.
Which video formats work?
Most of them. MP4 with AAC audio and WebM with Opus or Vorbis are read directly in Chrome, Edge, Firefox and Safari. For formats the browser can't read by itself, such as AVI, WMV, FLV, 3GP with AMR audio, or MKV and MOV files with AC-3, E-AC-3 or ALAC sound, the page offers an extra converter: FFmpeg built for browsers, about 10 MB, downloaded once when you ask for it. It runs in your browser like the rest, so the video is still not uploaded, and it reads an hour of audio in seconds on a laptop. Files the browser can play but not decode directly (some MOV files in Safari) can also be read by playing them silently in the background instead, which takes half the video's length in Chrome and Edge and its full length in Safari. If even the converter finds no sound, the page says so rather than guessing.
Can I make subtitles for my video?
Yes. Whisper returns timestamps for each phrase, so you can download the transcript as an SRT file (for YouTube, Premiere, DaVinci Resolve and most players) or a WebVTT file (for HTML5 video on the web), as well as plain text.
How big can the video be?
There is no fixed limit. In current browsers (Chrome, Edge, Firefox on desktop, and Safari 26 or later, including on iPhone) the audio track is read a piece at a time, so a long video needs no more memory than a short one. Older browsers and some formats fall back to reading the whole file at once; on a phone the page checks the size first and offers low-memory mode, which plays the video silently and keeps only the sound. Long videos still take a long time to transcribe on a phone, so a phone is offered recordings over 30 minutes in 30-minute parts, with each part's transcript ready to download before the next. On a phone the large model is not offered at all, because it needs more memory than a phone browser gives one tab.
Which model should I pick?
Tiny English is fast and usually enough for clear English speech. Pick Base or Large v3 turbo for accents, background music, several people talking, or any language other than English, and choose the spoken language. Bigger models download more the first time: Tiny is tens of megabytes, Base around 80 to 150 MB, and Large v3 turbo over 1 GB, which is why phones get Tiny and Base only.
Which languages does it understand?
Whisper's multilingual models (Tiny multilingual, Base and Large v3 turbo) transcribe dozens of languages, and the page offers 30 common ones. You pick the spoken language in the Spoken language list, which starts at your browser's language; it does not detect the language by itself. Tiny English and the other English-only models understand English only. Base or Large v3 turbo are noticeably better than Tiny for languages other than English.
Does it say who is speaking?
It can. Tick Label speakers before transcribing (or press Label speakers afterwards) and a second pass in your browser works out who is speaking when, using two small open models (about 12 MB, downloaded once). The transcript and the .txt and .md files get "Speaker 1:" style labels, the SRT cues start with the speaker's name and the WebVTT cues carry a voice tag (<v Speaker 1>). Rename a speaker and every download uses the name. Labels can be wrong where people talk over each other or sound alike, so check them before publishing subtitles.