Spoke is a macOS application that converts spoken words into text directly within any application interface. It functions as a push-to-talk dictation tool where users hold a keyboard shortcut, speak naturally, and release to have their speech transcribed and inserted at the cursor location.

The app offers on-device transcription using a local speech recognition model through CoreML, requiring no internet connection for basic functionality. It includes AI Skills that can process transcriptions with external providers for tasks like translation, grammar correction, and custom prompts. Users can configure push-to-talk shortcuts, access transcription history with search capabilities, and choose from 25 supported European languages with auto-detection.

Spoke operates by running a speech recognition model entirely on the user's Mac, processing audio locally without uploading it to servers. When AI Skills are used, only the transcribed text is sent to the chosen provider while the original audio remains on the device. The app leverages an industry-leading open-source model optimized for Apple Silicon with fast processing speeds.

The primary benefit is enabling faster text input without switching applications, particularly useful for people who think faster than they type. It maintains complete privacy by keeping audio processing local and allowing users to bring their own API keys for AI services. The one-time purchase model eliminates recurring subscription costs.

Spoke targets macOS users who need efficient dictation capabilities, particularly developers and professionals working with text input. It integrates with various AI providers including OpenAI, Anthropic, Google Gemini, and Ollama, and requires macOS 14+ on Apple Silicon devices.