SOETech Audio/Video Transcription Tool

Choose File

No file selected

Identify speakers — color-coded speaker turns, first run adds a one-time ~28 MB speaker model

First use downloads the AI model to your browser (~40–120 MB depending on device — Wi‑Fi recommended); it is cached for next time. Everything runs on your device.

Ready to transcribe

Model credits & licenses: transcription by Whisper tiny.en (MIT, OpenAI) · speaker identification by pyannote segmentation-3.0 (MIT, via the ungated onnx-community port) · runs in your browser with Transformers.js (Apache-2.0). Models are cached in your browser after the first run.

Privacy Notice: All processing happens directly in your browser. Your files are never uploaded to any server or stored anywhere.

This is an open source based project that runs on tools which may take time to process larger requests. Please feel free to leave a message in the contact form on the home page if you have any issues or requests for similar app development for you or your team.