Transcribe Audio to Text with AI

How to Transcribe Audio to Text with AI

  1. 1

    Choose a model size

    Whisper comes in several sizes. Tiny is fastest, Small is the most accurate on accents, noise and crosstalk. Base is a good middle default.

  2. 2

    Download the model

    The AI model is cached in your browser after the first download, so later transcriptions start immediately and work offline.

  3. 3

    Add your recording

    Drop in an audio file, or an MP4 or MOV video — the audio track is read straight out of it. You can queue several and add translation outputs alongside the original language.

  4. 4

    Transcribe and export

    Run the queue, then download the transcript as .txt or as .srt with timestamps.

What this is

AI transcription usually means handing your recording to somebody else's server. This tool runs OpenAI's Whisper speech recognition model on your own machine instead — the same class of model behind the paid transcription services, executed locally through WebAssembly. The audio is never uploaded, and neither is the transcript. That makes it usable for material you are not permitted to send to a third party: legal recordings, medical or therapy sessions, HR interviews, and journalism where a source needs protecting. Translation runs the same way, through a Marian neural machine translation model. Export as plain text or as timestamped .srt subtitles.

How it works

Whisper is a transformer-based speech recognition model trained on a very large multilingual corpus, which is why it copes with accents, crosstalk and background noise far better than the rule-based dictation software that came before it. Here it runs as an ONNX model executed by WebAssembly in a Web Worker, so inference happens on your CPU and the page stays responsive. Traditional DSP still does the preparation: your file is decoded and resampled to the 16 kHz mono signal the model expects before a single frame reaches the network. Translation adds a second AI model, a Marian encoder-decoder, applied to the finished transcript.

Credits

The research this tool runs on, credited in full — with BibTeX and licence notices — on the credits page.

  • Whisper

    MIT
    Alec Radford, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. Robust Speech Recognition via Large-Scale Weak Supervision. arXiv:2212.04356.Read the paper

  • OPUS-MT

    Apache-2.0
    Jörg Tiedemann and Santhosh Thottingal. 2020. OPUS-MT – Building open translation services for the World. In Proceedings of the 22nd Annual Conference of the European Association for Machine Translation (EAMT), pages 479–480.Read the paper

  • Marian NMT

    MIT
    Marcin Junczys-Dowmunt, Roman Grundkiewicz, Tomasz Dwojak, Hieu Hoang, Kenneth Heafield, Tom Neckermann, Frank Seide, Ulrich Germann, Alham Fikri Aji, Nikolay Bogoychev, André F. T. Martins, and Alexandra Birch. 2018. Marian: Fast Neural Machine Translation in C++. In Proceedings of ACL 2018, System Demonstrations, pages 116–121.Read the paper

  • Transformers.js

    Apache-2.0
    Joshua Lochner and the Hugging Face team. Transformers.js: State-of-the-art machine learning for the web.

  • ONNX Runtime

    MIT
    ONNX Runtime developers. 2021. ONNX Runtime: Cross-platform accelerated machine learning.

Supported formats

WAV, MP3, M4A, AAC, OGG and FLAC. Files are decoded by your browser, so anything it can play will work here. There is no audio output: the transcript is exported as a plain .txt file, or as .srt subtitles with timestamps.

Frequently asked questions

Is my audio uploaded anywhere?
+Is my audio uploaded anywhere?
No. Everything runs inside your browser using the Web Audio API, so the file never leaves your device. You can prove it: load the page, turn off your Wi-Fi, and the tool still works.
Can I transcribe confidential recordings?
+Can I transcribe confidential recordings?
That is exactly what this is for. Because inference runs locally, no part of the recording or transcript is transmitted, which is what makes it usable where uploading to a cloud transcription service is not permitted.
Does it work in languages other than English?
+Does it work in languages other than English?
Yes. The Whisper models are multilingual, covering over 90 languages, and you can add AI-translated outputs to English or to other languages alongside the original transcript.
Can I transcribe a video?
+Can I transcribe a video?
Yes — drop in an MP4 or a MOV and the audio track is read out of it, the same as it would be from an MP3. Nothing is extracted to a separate file and nothing is uploaded; the browser demuxes the container on the spot and only the audio reaches the model. The .srt export is the useful pairing here, since it drops straight back onto the same video as subtitles.
Does this use AI?
+Does this use AI?
Yes. Transcription runs Whisper, a neural speech recognition model, and translation runs a Marian neural machine translation model. Both are real AI models — the difference from a cloud service is only where they run, which here is your own device rather than someone else's server.
Do I need an account?
+Do I need an account?
No. The free tools work straight away with no sign-up, up to 15 minutes of audio per tool per day. An account is only needed if you subscribe to remove the limit.
Is it free?
+Is it free?
Yes. Every tool on this page is free to use, with a daily limit on total audio processed. There is no watermark and no trial period.

Other free audio tools