Transcribe audio and video privately with Whisper Web.

Select an audio or video file to create editable text, timestamps or subtitles. Whisper processes the media in this browser. Files can be up to 300 MB and 20 minutes.

No account · No media upload · TXT, SRT, VTT and JSON

ModelAudio languageOutput languageCompute

Processing happens on this device. Direct URLs must allow browser access (CORS).

Transcribe locally in three steps

01ImportFile, direct link or microphone

02Recognize locallyWhisper runs on your device

03ExportTXT, JSON, SRT or VTT

Start with your recording

Choose a workflow that matches the recording and the next task

Choose a scenario for task-specific steps, or read a guide before deciding where the audio should be processed and how the result should be exported.

Use cases

View all use cases

Meeting transcription without botsWork from a saved call recording without adding a participant to the meeting.Voice memo to textTurn a supported phone recording into editable notes and timed text.Private interview transcriptionReview quotes and timestamps while the source audio stays on the device.Private podcast transcriptionCreate text and captions for a short episode, trailer or selected clip.

Guides

View all guides

Transcribe audio without uploading itSee what stays local and what the browser still downloads.Local vs cloud transcriptionCompare the data path, limits, compute and collaboration trade-offs.WebGPU vs WebAssemblyChoose a local compute backend based on the device and browser.

Why your media stays private

Whisper Web decodes media and runs the Whisper model in your browser. Completed transcripts stay in IndexedDB until you export or delete them.

No media upload

The selected media is processed on your device and is not sent to Whisper Web servers.

Two compute options

WebGPU can run faster; WebAssembly provides a broadly compatible fallback.

Edit and export locally

Correct the text, copy it or export TXT, JSON, SRT or VTT.

Questions about local transcription

Does my file leave this device?

No. In local mode, the browser reads and decodes it in memory.

Why is the first transcription slower?

The selected Whisper model downloads once and is then cached by your browser.

Which formats are supported?

Whisper Web accepts common MP3, WAV, M4A, MP4 and WebM files, subject to your browser's decoding support.

Which languages can I transcribe?

The Audio language menu includes 99 languages, including English, Spanish, Arabic, Chinese, Hindi, French, German, Japanese and Portuguese. Choose the language spoken in the recording before you start.

What if I need to transcribe a large file?

Use the large-file page for a local audio or video file up to 1 GB and 1 hour. Keep this tab open: a long job can warm the device and take time. A reference estimate appears after processing begins.

Transcribe a large file