A tool for working with your words.

Select, review and download: a simple path from audio to text.

The project

VoxTexto is an independent audio transcription project focused on a simple interface and on-device processing. The public brand is VoxTexto at voxtexto.com. No account is required.

Application and open components

The interface, editing flow, language detection integration, file validation and document export belong to the application. Recognition uses OpenAI's open Whisper Base or Small models, converted to ONNX by onnx-community and run with Transformers.js. Silero VAD detects speech locally.

VoxTexto does not claim to have trained these models and is not an official OpenAI or Hugging Face product.

A result you can review

Automatic transcription does not replace listening when accuracy matters. Use the segments, audio playback and editor to check your words before export. Current limits are 60 minutes and 100 MB, subject to device resources.

Availability and credits

The web app can be installed from supported browsers. Windows system recording is a desktop feature; this version has no meeting bot. Whisper and Silero VAD use MIT licenses; Transformers.js uses Apache-2.0. Additional dependencies retain their own licenses. See licenses and attributions.