Spoke is a macOS app that transcribes your voice into any text field using a local speech model for privacy. Hold a keyboard shortcut, speak, and text appears wherever your cursor is.

Spoke is a voice dictation application for macOS and iOS that allows users to transcribe their speech directly into any text field across their system. It is designed for anyone who types frequently, including writers, bloggers, developers, and professionals, with the primary purpose of enabling faster, hands-free text input through a simple push-to-talk mechanism. The app solves the problem of slow typing and cumbersome voice input tools by providing a seamless, private, and integrated dictation experience that works universally without requiring app-specific integrations or constant internet connectivity.

Spoke operates using a 600M parameter neural network that runs entirely on the user's device, ensuring audio never leaves the computer and eliminating privacy concerns associated with cloud-based transcription services. It offers a visual workflow editor called Flows, which allows users to chain together steps like AI post-processing, text transforms, conditions, and automated actions to create custom voice-powered automations. The app supports 24 languages for on-device transcription and provides features like per-flow customizable shortcuts, mouse button hotkeys, hands-free voice activity detection mode, and biometric speaker verification.

Key capabilities include AI Skills that let users integrate their preferred language models—such as Apple Intelligence, OpenAI, Anthropic, Gemini, or local models via Ollama—for tasks like grammar correction, smart formatting, translation, or code prompt generation. The Flow Builder includes 16 node types for building multi-step workflows, enabling actions like taking timeline-aware screenshots, capturing clipboard content inline, switching between applications, simulating key presses, running AppleScripts, triggering webhooks, and saving output to files. Spoke also offers a unique Claude Channel feature for developers, creating an encrypted pipeline to send voice, screenshots, and copied content directly to Claude Code with full context and multi-session routing.

The app works by holding a customizable keyboard shortcut, speaking, and releasing, which triggers the selected Flow to process the voice input locally and output the result precisely where the cursor is located. Its native macOS architecture, built with Apple's frameworks and optimized for Apple Silicon, ensures low resource usage, with less than 50MB RAM and 1-2% CPU at idle, and smart text insertion that functions in any application. Users benefit from significantly faster text input—up to 4x quicker than typing—complete privacy with no cloud dependency, a one-time payment model with no subscriptions, and the flexibility to automate complex tasks across different apps and contexts.

Spoke

Spoke

Key Features

Use Cases

Who is this for?

Comments