Skip to content

Speech to Text (Transcription)

Convert speech from an audio or video file, or a recording from your microphone, into text and subtitles. The open Whisper model runs in your browser, so your recording is never uploaded to a transcription service.

  • Free
  • No sign-up
  • Runs in your browser

All processing happens in your browser — files never leave your device. Your audio is decoded and transcribed on this device and is not uploaded. The Whisper model downloads from Hugging Face on first use.

How it works

  1. Choose an audio or video file (up to 30 minutes), or record from your microphone.
  2. Select the spoken language or leave it on automatic detection. Optionally translate the speech into English.
  3. Choose Transcribe, then edit the text and download it as .txt, .srt or .vtt subtitles.

Good to know: On first use the Whisper base model (about 80 MB) downloads from Hugging Face. Transcription speed depends on your device and can take longer than the recording itself on older phones and laptops.

How the transcription works

Your browser decodes the file, mixes it to mono and resamples it to 16 kHz, the format the model expects. The audio is then processed in 30-second windows with a small overlap, and Whisper predicts the words and their approximate start and end times. Whisper was published by OpenAI under the MIT licence and was trained on many languages, so it can also detect the language and translate speech into English.

This tool uses the base-size model, which balances download size and accuracy for in-browser use. Clear speech with little background noise gives the best results; overlapping speakers, music, strong accents and specialist vocabulary make errors more likely.

Subtitles and editing

Timestamps are used to create SRT and WebVTT subtitle files, which most video players and platforms accept. Timings are approximate, so review them in a subtitle editor before publishing. The tool does not label different speakers.

Check names, numbers and technical terms, and add punctuation where needed. For interviews or legal and medical recordings, treat the transcript as a draft that needs human review.

Frequently asked questions

Is my recording uploaded?

No. Decoding and transcription happen in your browser. Only the model files are downloaded from Hugging Face.

Which file formats work?

Any audio or video format your browser can play, such as MP3, M4A, WAV, OGG, WebM and most MP4 files.

Why is transcription slow?

The model runs on your own processor. A modern laptop usually transcribes faster than real time; older phones can be several times slower.

Which languages are supported?

Whisper was trained on many languages, including English, Spanish, German, French, Czech, Portuguese, Hindi and Bengali. Accuracy is highest for widely spoken languages with clear audio.

Can I transcribe a YouTube video?

Only files you have on your device. If you have the right to use the video, download or export it first and then choose the file.

Related guides