Omikuji Tools Image & Video

Audio & Video Transcription

Turn speech in recordings and videos into text with AI (Whisper). Everything runs on this device; your audio never leaves it.

Features

  • Transcribe meetings, interviews and video audio with Whisper – no upload
  • English, Japanese or automatic detection; choose a model size with download size, speed and accuracy shown
  • Export timestamped text, plain text or SRT/VTT subtitles; click a time to play from there
  • Long recordings are split at pauses with progress and time remaining; microphone recording included

Use cases

Confidential meetings

Get minutes without handing recordings to a cloud service.

Interviews

Timestamped transcripts make it quick to find quotes.

Video subtitles

Export SRT/VTT for your video editor.

Lectures and talks

A first draft of notes from a recording.

Privacy

Audio and video are processed only inside your browser and are never uploaded. On first use the AI model (Whisper, 41–586 MB) and the inference engine are downloaded from this site (falling back to huggingface.co and cdn.jsdelivr.net if needed); your audio is not sent. The model is stored in your browser for next time. The microphone is used only on this page and recordings are not saved.

FAQ

Is my audio uploaded?

No. The transcription AI (transformers.js with Whisper) runs inside your browser. The only network access is the one-time download of the model and engine.

Which model should I choose?

For English, the small (tiny) model is fast and usually good enough; the larger ones are more accurate on difficult audio. For Japanese, choose Large (small).

Can it handle long recordings?

Up to 3 hours. The audio is split at pauses into parts under 30 seconds and processed in order. On a CPU it can take as long as the recording itself or longer, so keep the tab open.

Does it identify speakers?

No, speaker diarisation is not supported. Export with timestamps and add names where needed.

Related tools

Merge PDF Split PDF & Extract Pages Organize PDF (Reorder, Compress, Password) PDF to Image Image to PDF Audio Converter Image Converter WebP Converter SVG Converter Video Converter Markdown Preview

Add audio or video

Drop an audio or video file here, or click to choose

MP3, M4A, WAV, OGG, WebM, MP4, MOV and other formats your browser can play

Settings

Model size
Model Download Speed Japanese accuracy

The model is downloaded once and kept in your browser. Measured on our test PC (CPU, model already downloaded), a 27-second Japanese meeting clip took 5 s with tiny, 11 s with base and 20 s with small.

Spoken language

How to use

  1. Drop an audio or video file, or record with your microphone.
  2. Choose a model size and the spoken language.
  3. Press Start transcribing. The first run downloads the model. Keep the tab open while it works.
  4. Check the result and save it as timestamped text, plain text, or SRT/VTT subtitles.

Privacy

Audio and video are processed only inside your browser and are never uploaded. On first use the AI model (Whisper, 41–586 MB) and the inference engine are downloaded from this site (falling back to huggingface.co and cdn.jsdelivr.net if needed); your audio is not sent. The model is stored in your browser for next time. The microphone is used only on this page and recordings are not saved.