Sona / Personal Windows application
Sona: a thought, made into text.
Local dictation for Windows, with on-device speech recognition, editable drafts and guarded insertion into compatible text fields.
Public source / Early preview / Build from source
The problem
Dictation needs more than recognition. A useful desktop workflow also needs an editable draft, clear control over where text goes, and an honest fallback when insertion is unavailable.
What I did
I define the everyday workflow and visual direction, direct AI-assisted implementation, and review the app against practical dictation tasks.
How it works
- Recognize speech on the device and keep the draft available for review.
- Offer optional local cleanup while preserving an editable result.
- Guard insertion into compatible text fields and provide a copy fallback when direct insertion is unavailable.
What you can verify
- The public source and build instructions describe the native Windows app and its early-preview status.
- The documented interface includes CLI and MCP access alongside the desktop workflow.
Project scope
An early preview distributed as source, with no published binary installer. Compatible-field behavior varies and universal insertion is not claimed. Local models require appropriate hardware and setup. This is a personal tool, not a client delivery or a production-readiness claim.