NexaWebStudio
RUDiscuss a project

VoteXt — voice typing into any app

A finished desktop app for Windows and Linux: press a hotkey, speak, and the text lands in the active window

Category
Desktop · Windows & Linux
Platforms
Windows · Linux
Core stack
Electron · electron-vite · React 18
1.5 s

The problem

Typing long messages and emails is slow. VoteXt turns speech into text and pastes it wherever the cursor is — a messenger, a browser or an IDE.

What we built

  1. Global hotkey and a floating widget with a live waveform that never steals focus
  2. Whisper large-v3-turbo speech recognition: Russian, English, Ukrainian, Kazakh, auto-detect
  3. AI modes: fix grammar, format as an email, translate — with editable prompts
  4. Custom replacement dictionary and spoken punctuation commands
  5. Transcription history, auto-stop, paste via clipboard or simulated typing
  6. Builds for Windows (NSIS installer) and Linux (AppImage, RPM)

Engineering

≈1.5 s per phrase

  • About 1.5 s per phrase with no native modules or ffmpeg — webm/opus audio goes straight to speech recognition
  • Global hotkeys on Wayland via GNOME and KDE system shortcuts, where Electron's built-in mechanism doesn't work
  • Electron
  • electron-vite
  • React 18
  • TypeScript
  • MediaRecorder / Web Audio
  • Whisper (STT)
  • LLM API
  • electron-builder
Next case A CRM/ERP that replaced Bitrix24