Speech to Text
This free speech to text tool transcribes audio and video files, or a recording from your microphone, entirely on your device with the open Whisper model: no upload, no signup and no daily limit. It detects about 99 languages and exports TXT, SRT and VTT subtitles. A live dictation mode types as you speak.
Recorded and transcribed on this device.
Checking this device…
The transcript appears here, with a time for every sentence. Click a time to hear that part.
Files up to 2 hours. Speed depends on the device: with a graphics card it is usually several times faster than the audio itself.
Live dictation uses your browser’s speech service (Google in Chrome, Microsoft in Edge, Apple in Safari). To keep audio on your device, record in the first tab.
Say “period”, “comma”, “question mark”, “exclamation mark”, “colon”, “new line” or “new paragraph” to punctuate.
How do you transcribe audio to text for free?
Drop the file and wait a few seconds. The open Whisper speech model runs in your browser, on your own device, so the audio is never uploaded, there is no daily limit and no minute cap, and you do not need an account.
- Drop an audio or video file (MP3, M4A, WAV, a WhatsApp voice note, an MP4), or press Record from microphone. No file at hand? Try a sample voice memo.
- Leave the language on automatic, or pick it, and choose Fast or More accurate. Tick Translate to English to get English text from speech in another language.
- The transcript fills in piece by piece, with a time for every sentence. Click a time to hear that part, and fix any line.
- Press Copy text, or download TXT, SRT or VTT. The downloads use your edits.
How accurate is it?
We read three known scripts with a computer voice, transcribed them here and counted every word that came out wrong:
| Recording | Mode | Wrong words | Right | Time, model already on a laptop with a graphics card |
|---|---|---|---|---|
| English, 28 s | Fast | 3 of 76 | 96% | 5 s |
| English, 28 s | More accurate | 1 of 76 | 99% | 8 s |
| Portuguese, 24 s | Fast | 2 of 65 | 97% | 2 s |
| Spanish, 17 s | Fast | 3 of 51 | 94% | 2 s |
The language was detected correctly every time. The mistakes were the kind to expect: "homepage" heard as a made-up word, "shared" as "share", "sexta" as "cista". A computer voice is clean and steady, so real recordings with background noise, strong accents or people talking over each other will have more errors. That is where More accurate earns its larger download.
Fast or More accurate?
| Fast | More accurate | |
|---|---|---|
| Model | Whisper base | Whisper small |
| One-time download, on the processor | 77 MB | 249 MB |
| One-time download, with a graphics card | 206 MB | 410 MB, or 586 MB on cards without half-precision support |
| Best for | clear recordings, quick notes | noise, accents, languages other than English |
The download happens once. The browser keeps the model, so the next transcription starts right away and works with no internet connection. When the model is already on your device, the page says so.
How do you make subtitles from a video?
Drop the video, check the transcript, fix any word, then download SRT or VTT. Each subtitle is cut for reading: at most two lines of about 42 characters, at most 7 seconds on screen, and always inside the length of the audio, so the last line never appears after the video has ended. Add the file to YouTube, Vimeo, a video editor or VLC alongside the video.
Is my audio uploaded?
Files and microphone recordings in the first tab never leave your device: the only download is the model itself, once. Live dictation is the exception. It uses the browser's own speech feature, which runs on the device in recent Chrome versions after a one-time voice pack, and otherwise sends the audio to Google, Microsoft or Apple. The page tells you which one applies before you start.
What can't it do?
- Tell speakers apart. The transcript has times, but no Speaker 1 and Speaker 2.
- Go past 2 hours. A longer file is transcribed up to the 2-hour mark, and the page says so. Long files work best on a computer, which has more memory than a phone.
- Dictate live in Firefox, which offers no speech recognition to websites. Recording in the first tab works there.
- Hear what is not there. Loud music, crosstalk or a distant voice lower accuracy; More accurate helps, but read the result.
To hear text read aloud instead, use the text to speech tool; to count the words of a transcript, the word counter.
Frequently asked questions
Is this speech to text really free and unlimited?
Yes. The transcription model runs in your browser, on your own device, so there is no server bill per minute to cap. Transcribe as many files as you like, each up to two hours long, with no signup, no watermark and no credit card. A longer file is cut at two hours, and the page tells you.
Is my audio uploaded anywhere?
Files and recordings in the first tab never leave your device. The model is downloaded once and then runs locally, even offline. Live dictation is different: it uses the browser’s speech feature, which in some browsers sends audio to Google, Microsoft or Apple, unless the page shows it runs on this device.
How do I transcribe an audio or video file to text?
Open the first tab and drop the file, such as an MP3, M4A, WAV, WhatsApp voice note or an MP4 video. The language is detected automatically and the text appears piece by piece with a time for each sentence. Edit any line, then copy the text or download TXT, SRT or VTT.
How do I make subtitles from a video?
Drop the video file, wait for the transcript, fix any word you want, then download SRT or VTT. Each subtitle holds at most two short lines and seven seconds, and never runs past the end of the audio. Upload the file to YouTube, Vimeo or a video editor alongside your video, or add it in VLC.
Which languages does it understand?
The Whisper model covers about 99 languages and detects the spoken one by itself. Accuracy is highest in widely spoken languages such as English, Spanish, Portuguese, French and German. You can also pick the language by hand, and tick Translate to English to get an English transcript from speech in another language.
What is the difference between Fast and More accurate?
Fast uses Whisper base, a smaller model that downloads quickly and runs several times faster than the audio. More accurate uses Whisper small, a larger download that makes fewer mistakes on noisy recordings, strong accents and languages other than English. On a clean English test, Fast missed 3 of 76 words and More accurate 1.
Why does the first transcription take longer?
The first time, the browser downloads the model: about 77 to 206 MB for Fast and 249 to 586 MB for More accurate, depending on the device. It is stored by the browser, so later transcriptions start right away and work offline. A graphics card makes it several times faster than real time.
How does live dictation work?
Press Start dictation and speak. Words appear as you talk, and you can edit them. Say period, comma, question mark, new line or new paragraph to punctuate. Recent Chrome versions can run it on the device after a one-time voice pack download; otherwise the browser’s online speech service does the work.
Can it tell who is speaking?
Not yet. The transcript is one continuous text with times, without speaker names such as Speaker 1 and Speaker 2. Separating speakers needs an additional model. For interviews and meetings, the clickable times make it quick to find each answer and label it yourself while editing the lines.
Does speech to text work in Firefox?
Transcribing files and recordings works in Firefox, Chrome, Edge and Safari on computers and recent phones. Live dictation does not work in Firefox, which does not offer speech recognition to websites. In Firefox, use Record from microphone in the first tab, which transcribes on your device instead.
Related tools
Rate Speech to Text & help shape it
This tool is free and still growing. Tell us what works, what you would change, and what is missing. Your feedback is what decides what we build next.
Built and tested by the ÆON Toolhub team. Last updated: October 19, 2026. How we test our tools