Candy
Private Speech to Text
Experience the power of local AI. Convert your voice to text instantly without your data ever leaving your device. 100% private, offline-capable speech-to-text application for macOS, Windows, and Linux.
Key Features
All speech recognition runs on-device using a bundled Whisper model — no audio data, no transcription results, and no usage metadata ever leave your computer. Local processing also eliminates network-dependent latency, delivering transcription results with sub-second response times regardless of your internet connection.
Candy operates without any internet connection after installation. Transcription is fully functional in airplane mode, on air-gapped secure networks, in areas with unreliable connectivity, and in situations where sending audio to a cloud API is prohibited by policy — making it suitable for classified, legal, and medical environments.
Built with Tauri, Candy compiles to native desktop applications for all three major operating systems from a single codebase — delivering platform-native performance, OS-appropriate keyboard shortcuts, system tray integration, and notification support on each platform without a separate codebase or build pipeline.
There are no backend servers, no user accounts, no telemetry collection, no usage analytics, and no data collection of any kind. Candy is a fully local, offline application with zero network dependencies at runtime — what you dictate stays on your device, period.