Skip to content
النسخة العربية

Video to text (transcript)

Turn speech in a video into a timestamped transcript on your device, then export TXT, Word, SRT or VTT. Nothing is uploaded.

Loading tool…

Drop a video or audio file hereMP4, MOV, WebM, MKV, MP3, M4A, WAV, OGG and more · up to 500.0 MB and 60 minutes on this device
  • Your audio never leaves this device: it is decoded and recognised inside this browser tab.
  • The first run downloads the speech model (44 MB fast, 81 MB more accurate, 247 MB best) and keeps it in your browser for next time.
  • Keep this tab open until it finishes. Long files take a while; speed depends on your computer.
  • Works best in Chrome or Edge on a desktop or laptop.

How it works

Choose a video and this tool writes down what is said, line by line with timestamps. It uses OpenAI Whisper, an open speech-recognition model, running inside your browser tab: the audio track is decoded on your device and recognised by a model downloaded to your browser, so the video itself is never uploaded anywhere. The transcript appears while it is being made, you can correct any line, click a time to jump the player to that moment, search it, and download it as a Word document, a PDF, plain text, text with timestamps or SRT/VTT subtitles, or download the video itself with captions: an MP4 with a subtitle track players can switch on, or with the captions burned into the picture for WhatsApp, Instagram and TikTok.

Three model sizes are offered. In our test on a single 2.1 GHz processor core, Fast (Whisper tiny, about 44 MB to download once) got through a 5-minute English recording at about 4 times real speed, so a 10-minute video takes roughly 2.5 minutes; More accurate (Whisper base, about 81 MB) made noticeably fewer mistakes at about 2 times real speed, roughly 5 minutes for the same video. Best (Whisper small, about 247 MB) is the one to use for Arabic: OpenAI's published results show it making far fewer mistakes than the smaller models in Arabic, and in our test it handled Arabic speech at about 0.6 times real speed (half the speed of More accurate). Speed depends on your processor, and the first run also waits for the download. Whisper handles 99 languages including English and Arabic, can translate speech from any of them into English, and works best on clear speech: music, crosstalk, heavy accents or a distant microphone all cost accuracy, so read the result through before publishing.

Long videos are processed in 30-second windows. When a sentence runs past the end of a window, the next window starts at the beginning of that sentence, so words are not cut in half or repeated. Files up to 60 minutes and 500 MB work on a desktop or laptop (Chrome or Edge recommended; keep the tab open until it finishes); on phones the limit is 20 minutes because mobile browsers have less memory.

Summary, action items and decisions

When the transcript is finished, a summary panel appears under it: a short summary, key points, action items and decisions, each with the time it was said (click it to play from there). Everything in it can be edited, and it downloads as Word, PDF or text next to the transcript and subtitle files.

It is a basic on-device summary: sentences are ranked by how central they are and picked from the transcript, and action items and decisions are sentences with phrases such as "we need to", "I will", "we decided", "يجب", "سوف" or "تم الاتفاق". It does not rewrite or invent anything, and it can be wrong, so read it through. A search box finds the passages that answer a question, with their time; the answers are passages from the transcript, not generated text.

How to use it

  1. Choose or drop a video file (MP4, MOV, WebM, MKV, AVI and more).
  2. Pick the spoken language, or leave it on automatic detection, and choose Fast, More accurate or Best (recommended for Arabic).
  3. Click Transcribe. The first time, the speech model downloads once; after that the lines appear as each part of the video is recognised.
  4. Click a timestamp to check a passage in the player and fix any line directly in the transcript. If the language is wrong, choose the right one and redo; the audio is not decoded again.
  5. Download the transcript as Word, PDF, plain text or text with timestamps, the subtitles as SRT or VTT, or the video as an MP4 with captions.

Frequently asked questions

Is my video uploaded to a server?

No. The audio is decoded and transcribed inside your browser tab. The only download is the speech model itself (from this site, once), which your browser keeps for next time. You can disconnect from the internet after the model has loaded and it will still work.

How accurate is it?

On clear speech in a common language Whisper is usually good, but it is not perfect: names, numbers and technical terms are the usual errors, and background music, overlapping speakers or strong accents reduce accuracy further. The More accurate model helps. Always proofread before you publish or quote the transcript.

How long does it take?

In our test on a single 2.1 GHz processor core, the Fast model transcribed English speech at about 4 times real speed (a 10-minute video in roughly 2.5 minutes) and More accurate at about 2 times real speed (roughly 5 minutes). Slower computers and phones take longer, and the first run adds the model download. A progress bar shows the time remaining.

Which languages are supported?

The 99 languages Whisper was trained on, including English, Arabic, Spanish, French, German, Hindi, Urdu, Turkish and Chinese. Quality is best for widely spoken languages; choosing the language yourself is more reliable than automatic detection for short or noisy clips.

What if it picks the wrong language?

Automatic detection listens to up to three stretches of speech, skipping silence and music intros, and leans towards Arabic on the Arabic page or when your browser is set to Arabic. It then shows what it detected, for example "Detected: Arabic (91%)". If that is wrong, open "Wrong language? Choose and redo", pick the language and redo; the tool remembers your choice for next time.

Can I download the video with the captions?

Yes. "MP4 with subtitles" adds a subtitle track (tagged with its language) that players switch on with their CC button; it usually takes seconds because the video is copied, not re-encoded. "MP4 with captions burned in" draws the captions into the picture with correctly joined, right-to-left Arabic, so they show everywhere, including WhatsApp and Instagram; it re-encodes the video on your device, up to 30 minutes on a computer.

Can I save the transcript as PDF or Word?

Yes. Both are made on your device. The PDF has the file name, date and every line with its time, with Arabic shaped and laid out right to left; the text can still be searched and copied. The Word file uses right-to-left paragraphs for Arabic.

Can it translate?

Yes, into English only: tick "Translate the speech into English" and Whisper writes an English version of what is said in any supported language. For other target languages, transcribe first and translate the text separately.

Is there a length limit?

Up to 60 minutes and 500 MB on a desktop or laptop, and 20 minutes on phones and tablets, because the decoded audio and the model have to fit in the browser's memory. For longer recordings, cut them into parts with the video trimmer first.

Does it tell speakers apart?

No. Whisper writes what is said but does not label who said it. Add speaker names while editing the lines if you need them.

Related tools