No sound coming through+
Check the microphone permission in your browser or system and confirm the input device. Moving bars indicate audio input; recognized text indicates the model is producing results.
Why download a model first+
Browser speech recognition uses a local model. Once the page and model are cached, they can work offline. Clearing site data can remove the model and require a new download.
Why does the next sentence light up early+
At about 90% text match, the reading focus moves ahead while recognition still tracks the end of the original sentence. This gives your eyes time to prepare.
How do I resize the script+
Pinch in the script area, or use the text-size control on desktop. Available interactions depend on the test version being used.
Are Android and Windows officially released+
They are in browser validation. Target-system, battery, installation, and update acceptance are not complete. Official download links open when each version is ready.
Can I record video with the camera+
The camera is an optional local preview, off by default. Video recording is not part of the current product scope.
Updated November 9, 2026