Audio to Text Converter

To convert audio to text for free, drop your file into the tool above, choose a model and click Transcribe. The text appears line by line with timestamps, and you can copy it or download it as TXT, Word, SRT or VTT. It works with MP3, WAV, M4A, OGG, FLAC and the audio inside MP4, MOV and WEBM video, in 99 languages, without uploading your recording anywhere.

Under the hood it runs Whisper, the open-source speech recognition model from OpenAI, directly in your browser. The model downloads once, gets cached, and from then on everything happens locally. That's why there's no per-minute pricing, no queue and no account. It also means you can safely transcribe things you'd never paste into a random website, like a client call or a doctor's appointment recording.

Which Model Should You Pick?

All three options are versions of Whisper. Bigger models make fewer mistakes but take longer to download the first time and longer to run. If you're not sure, start with Balanced.

The download size depends on how your browser runs the model. Chrome and Edge on most laptops can use WebGPU, which uses your graphics chip and is several times faster than running on the processor alone. The tool picks the fastest option your device supports and tells you which one it's using.

How Accurate Is AI Transcription?

On clean speech, Whisper gets the vast majority of words right, and it adds punctuation and capital letters on its own, so the output reads like normal text rather than one long run-on line. Accuracy drops with overlapping speakers, music under the voice, heavy echo, or a phone held far from the person talking.

A few things help more than anything else:

• Set the language instead of using Auto-detect on short clips. Whisper guesses the language from the first 30 seconds, and a wrong guess hurts everything after it.
• Use the Accurate model for names, jargon and accents. It's the single biggest quality jump.
• Trim silence and music from the start with the audio trimmer. Long silent stretches are where speech models are most likely to invent text.
• Record closer to the mic. A phone on the table in the middle of a meeting picks up the room more than the people.

Supported Audio and Video Formats

The tool reads whatever your browser can play, which in practice covers almost everything people record with:

If a file won't open, your browser doesn't have a decoder for it (WMA and some AMR files on Chrome are the usual culprits). Run it through the free audio converter to get an MP3, then transcribe that.

Making Subtitles From a Video

Drop the video file straight in. There's no need to extract the audio first. Once it's transcribed, click SRT for YouTube, Facebook, Premiere Pro, DaVinci Resolve and VLC, or VTT for HTML5 video players and most web platforms. Each subtitle line keeps the timing Whisper detected, and any corrections you make in the transcript are carried into the file. If you only need part of a long video, cut it down with the video trimmer first.

What People Use It For

• Students: turning recorded lectures into searchable notes.
• Journalists and researchers: transcribing interviews without sending sources' voices to a third-party server.
• Podcasters and YouTubers: show notes, blog posts and captions from each episode.
• Teams: meeting minutes from a Zoom or Teams recording.
• Everyone else: reading a long voice note instead of listening to it, or getting the English meaning of a message in another language.

Once you have the text, you can check its length with the word counter. For text trapped in screenshots or scanned pages rather than audio, use image to text or PDF to text.