Context
Dictation needed to fit into the desktop applications people already use, with a small interface available from the tray and keyboard.
← Back to archive
Verba
About the project
A Windows-first Electron desktop app that brings AI dictation into existing writing workflows. Global shortcuts start Quick or Polish capture, recorded audio is transcribed with the user's own OpenAI API key, and an optional cleanup pass improves casing and punctuation. Verba copies the result to the clipboard and attempts to paste it into the previously focused application. The tray app coordinates recording, transcription, cancellation, error feedback, and configurable compact or extended overlays.
Use case breakdown
Goal
Make voice-to-text available through a shortcut, with a choice between direct transcription and an additional text-cleanup pass.
Challenge
The capture, transcription, clipboard, and paste steps had to work together while keeping the user informed about recording, progress, cancellation, and failures.
Solution
Built an Electron main/renderer workflow with global shortcuts, microphone capture, validated IPC, background transcription and cleanup requests, clipboard delivery, and a configurable status overlay.
Technical highlights
- Separate global shortcuts for Quick and Polish capture
- Speech transcription with a user-provided OpenAI API key
- Optional AI cleanup for punctuation and casing
- Clipboard output with best-effort paste into the focused app
- Compact and extended recording/status overlays
- Configurable language, shortcuts, and output formatting
Challenges solved
- Coordinating microphone capture, IPC, and background requests
- Handling shortcut conflicts and restoring the desktop writing flow
- Keeping recording, transcription, and cancellation states synchronized
- Providing useful feedback for permissions, connection, and API errors
Results & impact
- Delivered shortcut-driven dictation with distinct Quick and Polish modes.
- Connected transcription output to the clipboard and best-effort desktop paste.
- Added settings and status feedback for the complete capture-to-text workflow.
Product screens