Transcription often means recording in a separate tool, waiting for the result and copying it into another application. These steps interrupt work, and the text may still need corrections or rewriting before it is usable.
02
The solution
A desktop app for dictating into the active window through a global shortcut and a discreet floating overlay. Vox transcribes and cleans up dictation, accepts voice instructions for writing or rewriting, imports audio files and produces speaker-labelled meeting transcripts. A library makes recordings searchable, playable and exportable. Available on Windows, macOS and Linux, with capabilities adapted to each system.
03
The challenge
Making voice-to-text insertion reliable across shortcuts, processing delays and window changes. Faithful dictation must remain distinct from writing instructions: a spoken question should be transcribed, not answered by the AI. Permissions, audio capture and insertion methods also differ across operating systems.
04
The result
Speech becomes reusable text with fewer window changes and copy-paste steps. If insertion cannot be confirmed, the text remains available to copy or insert again. The library keeps transcripts and recordings easy to find and reuse.
Insertion methods and available features depend on the operating system.
Key features
Dictation in the active window
Global keyboard shortcut
Discreet floating overlay
AI cleanup and rewriting
Audio file imports
Speaker-labelled transcripts
Library and audio playback
Exports and usage tracking
Technical stack
Python
PySide6
AI APIs
Audio processing
Inside the project
Explore the application’s interfaces and user journeys.
Dictation in the active applicationVoice instructions for writing and rewritingSpeaker-labelled meeting transcriptsSaved dictations, ready to reuseCustom cleanup and tone settings
