Loccut

Speech to text

Open a recording. Whisper writes the text on this device.

How it works

  1. The audio is decoded in this browser and resampled to 16 kHz.
  2. Whisper tiny runs here with WASM.
  3. The model is downloaded once and cached. The audio is not uploaded.

Questions

Where is the model stored?

In this browser, after the first download from the model host or Hugging Face.

Will audio be sent out for recognition?

No. Only the model files are downloaded. The recording stays here.

Does the audio leave this device?

No. Decoding and export happen in this browser.

More image tools