Audio to text
Meetings, lectures and recorder files turned into readable text on this device, with no upload.
- Nothing is uploaded
- Works offline after first use
- Open source
- No watermark · limit is your device’s memory; phones hold less
Turn a WAV, OGG, FLAC, M4A or MP3 recording into text in your browser. It is free, needs no account and adds no watermark, and your files are not uploaded: everything runs in your browser on your device. The first run downloads the speech model (about 76 MB) and the WebAssembly engine that runs it to your browser, and after that the tool works offline.
Done
Open the network tab. Your files never appear in it.
Add files and run the tool to see every request this tab makes.
Your files are processed on this device and never sent to a server. Reload the page to clear everything from memory.
Audio recorders, meeting apps and phones all save sound in different formats: WAV from a handheld recorder, OGG or Opus from a chat app, FLAC from a lossless capture, M4A from a phone. This page reads all of them and writes out what is said as paragraphs of plain text you can search, quote and paste into your notes. For a single MP3 or a voice memo, MP3 to text is the same tool with that case in mind.
A worked example: a one-hour team meeting saved as a WAV by a recorder on the table. Drop it here with the language left on Detect. The sound is decoded in the tab, cut at pauses into pieces, and each piece is transcribed by Whisper on your device. A paragraph break falls wherever there was a pause of two seconds or more, which usually lands where someone else starts talking. It does not label who is speaking, so add names when you read it through.
The first run downloads the speech model (76 MB) and the engine that runs it from this site, once; after that it works offline. Because the model runs in your browser, the recording is not uploaded: a client call, a patient dictation or a source interview stays on your device. Long recordings take proportionally longer and need memory, since the whole recording is decoded first.
If the recording is a video, such as a recorded lecture or a Zoom MP4, use Video to text. For timed captions instead of plain text, choose Subtitles (.srt) on the general transcribe tool.
How it works
- Drop one or more audio files: WAV, OGG, FLAC, M4A, MP3 and more.
- Leave the language on Detect, or pick it for short or noisy recordings.
- Press Transcribe, then copy the text or download it as a .txt file.
In your browser, not on a server
Your browser’s own decoders read the sound, from an audio file or a video’s soundtrack, and it is resampled to 16 kHz mono in the tab. The recording is cut at pauses into pieces of up to 28 seconds, and each piece goes to OpenAI’s Whisper base model (MIT), 8-bit ONNX weights run by transformers.js (Apache-2.0) and onnxruntime-web (MIT) in a background worker. The model, 76 MB, is downloaded from this site on first use and cached. Its timed segments become paragraphs, SRT or WebVTT.
Compared with upload-based tools: Online transcription services upload the recording to their servers, cap free minutes, and some keep your audio to train their models. Here the model runs on your device: nothing is uploaded, there is no minute limit beyond your device’s memory (phones hold less), and it works offline after the first run.
Verify it yourself. Give the page a short while after your first run to save itself for offline use, then turn off Wi-Fi or switch to airplane mode and run the tool again. Or open your browser's developer tools (F12, or Cmd-Option-I on a Mac), pick the Network tab, and watch it while you convert. The panel above does the same count for you.
Questions
Which audio formats does it read?
MP3, WAV, M4A, AAC, OGG, Opus and FLAC, plus WebM audio. A video file works too, on Video to text.
Is there a length limit?
No fixed limit. Long recordings take proportionally longer, and very long ones (several hours) need a device with plenty of memory, since the whole recording is decoded first.
Does it work offline?
After the first run, yes: the page and the speech model (76 MB) are stored by your browser, so you can transcribe without a connection.
Can it translate the recording?
Into English only. Choose Translate to English for a foreign-language recording; for text in the spoken language, leave it off. It does not translate into other languages.
Does it say who is speaking?
No. It writes what is said, not who said it. Long pauses start a new paragraph, which usually falls where the speaker changes.
Is the file uploaded anywhere?
No. There is no server in this process. The page is a static file and the work is done by code running inside your browser tab. Give the page a short while after your first run to save itself for offline use; then you can turn off Wi-Fi and it still works.
More video and audio tools
Transcribe audio to text
Speech to text and subtitles made by an AI model running in this tab. Interviews, lectures and voice notes are never uploaded.
Compress video
Make a video a fraction of the size, or fit it under a limit like 10 MB for Discord or 25 MB for email. It is compressed on this device and never uploaded.
Video to MP4
Turn MOV, MKV and WebM files into MP4s that play everywhere. Converted on this device, never uploaded.
Trim video
Cut a clip out of a longer video without uploading it. Most cuts take seconds and keep the original quality.
Other conversions
Guides
- A free Otter.ai alternative with no monthly minute capOtter.ai's free tier meters your minutes each month and caps how long a single recording can run. Here is what that looks like, and an alternative with nothing to meter.
- How to transcribe audio to textEvery platform has a way to turn speech into text now, at very different prices. Here is what each one offers, and a free option that never uploads the recording.