Upload an audio file (up to 10 minutes), choose the spoken language (the model doesn't detect it automatically), and get a timestamped transcript using the multilingual Whisper tiny model running in your own browser through Transformers.js: with WebGPU when your device supports it, and with WebAssembly otherwise. The first time, the browser downloads the model from Hugging Face (the exact size is shown before you start) and may keep it in its cache; your audio is never sent to any server. Export the result as plain text, SRT, or WebVTT. It's an automatic transcript from a small model, so it's worth reviewing, especially with noise, music, or several speakers.
Checking your browser's compatibility...
Your audio is never uploaded: it is decoded and transcribed on your device. The only external service it connects to is Hugging Face, to download the Whisper speech model (your browser may keep it in its cache); the WebAssembly engine is served by CreaNexo itself.