Transcribe on your own device
Whisper speech recognition runs inside your browser. Your recording is never uploaded — only the model is downloaded, once.
Checking what this device can run…
How on-device mode works
Your recording stays here
The file is decoded and transcribed inside this browser tab. It is not uploaded to VoqText or to any provider, and nothing is saved unless you download it.
What does use the network
The first time, the model files are downloaded from Hugging Face (huggingface.co) and the speech engine from jsDelivr (cdn.jsdelivr.net). Those services see an ordinary download request, never your media.
Cached for next time
Your browser keeps the model so later runs start quickly. Clear this site's storage in your browser settings to remove it.
Honest about accuracy
On-device models are smaller than the cloud service. Accuracy varies with model size, language and audio quality, and there are no speaker labels.