Skip to content
DevWizard productmacOS 14+Developer tools

EchoType

System-wide voice dictation for macOS that drives ChatGPT's own web dictation, with no API keys and no per-use cost.

View source on GitHub
Latest release
v1.0.76 Sept 2026
Language
Swift
License
MIT
Last updated
6 Sept 2026
Illustrative interface drawn from the project's features, not a screenshot.

The problem

Local speech models take gigabytes of disk and memory, and cloud speech APIs bill per use. For people who already use ChatGPT, dictation was already available in the browser but not inside every other app.

What we built

What we built

EchoType keeps a hidden, logged-in chatgpt.com session in a WKWebView. A global hotkey drives ChatGPT’s dictation buttons, reads the transcript from the composer without ever sending the message, clears it and pastes the text into the focused app.

  • Hold-to-talk with a configurable key, a hands-free mode and a non-focus-stealing waveform pill.
  • Works in any app with a text field: editors, browsers and chat apps.
  • Local history with pin, copy and multi-delete. Audio goes only to chatgpt.com through the user’s own session.
  • Self-updating from GitHub Releases, checked against a SHA-256 digest.
  • Keep-warm settings to trade instant dictation for lower RAM: about 150 to 250 MB while the page is loaded and about 30 MB idle.

How it was built

A single Swift binary with zero external dependencies, packaged as a DMG. The README documents the first-run permissions, since the app asks for Microphone and Accessibility access, and notes that it is not notarized.

Source and releases: DevWizardHQ/EchoType.

Stack

  • Swift
  • WKWebView
  • macOS 14+

Outcome

Versioned releases on GitHub, with in-place updates verified by SHA-256.

Services behind this work

The same people and skills are available for your project.

Start a project

Tell us what you want to build. We reply with questions, then a written scope and quote.

Start a projectSee all work