Transcribe

Transcribe Audio to Text

Transcribe meetings, interviews, lectures and videos into text with the Whisper speech recognition model. Choose a file or record with your microphone, pick speed or accuracy, and get text with timestamps that you can edit, copy or save as SRT/VTT subtitles. The audio is processed on your device and never uploaded.

1. Choose audio

2. Settings

Model

Your audio is processed only in your browser. It is not sent to our server or to Hugging Face.

  • This is automatic transcription by AI, so mishearings, missing words and wrong punctuation will occur. Always check names of people, companies and places, technical terms and numbers against the audio.
  • Processing runs on your device. On older computers and smartphones it can take longer than the audio itself. For long recordings, try “Speed first” first.
  • Music, background noise and people talking at the same time are hard to recognise. Speakers are not identified.
  • Microphone recordings can be up to 60 minutes. Ask everyone’s permission before recording a meeting or interview.
  • AI model: OpenAI Whisper tiny / base / small (Apache-2.0, ONNX versions Xenova/whisper-*). Runtime: Transformers.js (Apache-2.0) and ONNX Runtime Web (MIT).

About this tool

Transcribe audio files (MP3, M4A, WAV), video files (MP4, MOV) or microphone recordings to text with the Whisper speech recognition model. It supports Japanese and English, with three models to choose from: speed first, balanced and accuracy first. The result comes with timestamps, can be edited on the page and saved as plain text or SRT/VTT subtitles. Audio is processed on your device and never uploaded.

How to use

  1. Choose a file under “Audio / video file” (or drop it), or open “Record with microphone” and press “Start recording”.
  2. Pick a model (Speed first, Balanced or Accuracy first) and a language (Japanese, English or Detect automatically).
  3. Press “Start transcribing”. The first run takes a little longer while the AI model downloads.
  4. Press a time in the result to play from there, and correct the text as you listen.
  5. Use “Copy”, “Text (.txt)”, “Subtitles (.srt)” or “Subtitles (.vtt)” to copy or save the result.

Useful for

  • Drafting meeting minutes from a recording
  • Transcribing interviews and podcasts
  • Creating SRT or VTT subtitles for YouTube and other videos
  • Reviewing recorded lectures and seminars as text
  • Turning voice memos into notes

FAQ

Is this transcription tool really free?

Yes, with no sign-up and no usage limits. Because processing runs on your device, long recordings take longer. Microphone recordings can be up to 60 minutes each.

Is my audio uploaded to a server?

No. Audio is transcribed inside your browser and never sent to our server or any other service. Only the AI model (about 44–325 MB) is downloaded from Hugging Face the first time and kept in your browser.

How accurate is the transcription?

Clear speech is mostly transcribed correctly, but mistakes and missing words do happen, especially with names, technical terms and numbers. Always check against the audio, and choose “Accuracy first” for better results.

Can I make subtitles from a video?

Yes. Choose an MP4 or MOV file and save the result as SRT or VTT subtitles with timestamps. The AI decides where lines break, so you may want to adjust them in your video editor.

Can it tell different speakers apart?

No, speaker identification is not supported. You get text and timestamps for each section, and you can add speaker names to the saved text yourself.

Contact