Skip to content
Work data is processed in your browser, not stored on our servers.
tools
Whisper Audio & Video Transcription | SRT Generator

Whisper Audio & Video Transcription | SRT Generator

Transcribe audio or video locally with Whisper, edit timestamped text, and export TXT, SRT, or WebVTT subtitles in your browser.

Focus mode
Audio or video
Transcribe the primary audio track locally in overlapping 30-second windows.

The first run asks before downloading the roughly 44 MB Whisper model and shared runtime files.

The file and prepared audio never leave your browser.

Editable subtitles
Review the timing and text, then export TXT, SRT, or WebVTT.
Choose a file and start transcription to see editable cues here.

Key features

  • Transcribes the primary audio track from audio and video with multilingual Whisper Tiny.
  • Supports automatic language detection or explicit English, Japanese, and Korean selection.
  • Lets you edit recognized timing and text before exporting TXT, SRT, or WebVTT.
  • Processes files up to 20 minutes sequentially in overlapping 30-second windows.
  • Passes generated cues once to the subtitle editor without server persistence.

How to use

  1. Choose one audio or video file up to 500 MB and 20 minutes.
  2. Select automatic language detection, English, Japanese, or Korean.
  3. Start transcription and approve the model download on the first run.
  4. Review timing and text, then export TXT, SRT, or VTT or continue in the subtitle editor.

Examples

  • Meeting and interview drafts

    Turn a recording into editable timestamped text for a meeting note or interview draft.

  • SRT captions for video

    Transcribe video speech, refine the timing and wording, and save SRT or WebVTT captions.

  • Private local review

    Inspect speech as text without uploading the source media to a transcription server.

Privacy

Media decoding and Whisper inference run in your browser. The file, prepared audio, and transcript are not uploaded to our server. After explicit consent, pinned model files may be downloaded from Hugging Face and cached in browser storage.

Requirements

A current browser with WebCodecs or the bounded Web Audio fallback is required. Files are limited to 500 MB and 20 minutes, with a 96 MiB prepared-PCM ceiling. The roughly 44 MB model plus shared runtime files are required; speed and format support vary by browser and device.

Common questions

Is my file or audio uploaded?
No. Decoding and AI inference run in your browser. Only the first model download contacts the external model repository.
How is this different from live microphone dictation?
This tool handles saved audio or video and creates SRT/VTT files. Use the separate speech-to-text tool for live microphone input.
Will every file transcribe accurately?
No. Codec support, recording quality, noise, speech, browser support, and device performance all affect availability and results.

Related tools