Question 1

What is DysVoxa?

Accepted Answer

DysVoxa is an experimental Windows app that turns one spoken phrase at a time into reviewable text and, after the user chooses or edits it, reads it aloud with a synthesized voice.

Question 2

Does DysVoxa automatically speak whatever it recognizes?

Accepted Answer

No. The user must review, choose, or edit the text and then explicitly press Speak.

Question 3

Is internet access required?

Accepted Answer

Not for the intended Local workflow after the offline app and models are installed. Internet access is required for optional Cloud correction and may be needed for development-time model downloads.

Question 4

Is audio sent to the cloud?

Accepted Answer

Microphone audio is not uploaded for local speech recognition or by the cloud text-correction path. If the user deliberately enables Cloud mode, recognized text and relevant saved text context may be sent to the configured provider.

Question 5

Does a voice profile retrain the speech recognizer?

Accepted Answer

Not in the current beta - the bundled Parakeet and Whisper models are fixed, and a profile provides phrases, vocabulary, saved corrections, and other local evidence. The planned release direction is enrollment that fine-tunes the recognizer on the user's own recordings, on the user's computer.

Question 6

What speech data was used to develop DysVoxa?

Accepted Answer

Development research used the public TORGO corpus of dysarthric speech and the developer's own recordings, only to test whether recognition improves when a model is adapted to one speaker. Those recordings are research inputs: they are not collected from users and are not needed to run the app. The intended product adapts recognition to each user's own enrollment recordings, on that user's computer.

Question 7

Is it suitable for severe dysarthria?

Accepted Answer

The project is intended to investigate a broad range of dysarthric speech, including serious recognition errors. Current evidence is too limited to promise reliable performance at any severity level.

Question 8

Can it be used for emergencies?

Accepted Answer

DysVoxa should not be relied on as the only way to communicate in an emergency or another safety-critical situation. Users should keep an established fallback method available.

Question 9

What stage is development in?

Accepted Answer

A functioning Windows beta exists. Current work adapts the recognizer to each speaker's own enrollment recordings on that user's computer, followed by validation and a packaged offline release. The development status page lists each stage and what is still unproven.

Speech recognition built for dysarthria.

Three steps from phrase to chosen speech. You stay in control.

Record one phrase

Review and choose

Speak chosen text

Experimental beta

Questions, answered.

Join testing when available.