Dictation for Developers: Voice Coding, AI Prompts, and Terminal Input

Developers spend a surprising amount of time writing things that aren't code. Pull request descriptions. Code review comments. Slack messages explaining architectural decisions. Documentation. Commit messages. AI prompts for ChatGPT, Claude, or Copilot Chat.

All of these are natural language. All of them interrupt your coding flow when you have to context-switch from thinking in code to composing prose. And all of them are faster to speak than to type.

The context-switch problem

You're deep in a debugging session. You find the root cause and want to leave a code comment explaining why the fix works — a comment your future self will thank you for. But switching from "reading assembly" mode to "writing English" mode takes effort. So you write a terse comment, or skip it entirely.

Voice input eliminates that friction. Hold a key, say what you're thinking, release. The comment appears. You never left the editor, never shifted your hands off the keyboard for more than a second, never broke your train of thought.

Where voice typing fits in a developer workflow

Code comments and documentation

The best code comments explain why, not what. But "why" comments require you to articulate reasoning — and articulating reasoning is what speech is for. Dictating a comment like "we retry here because the upstream API returns 503 during deployments, which happen every Tuesday at 3am UTC" is faster to say than to type and produces better documentation than "retry on failure."

AI prompts

If you use ChatGPT, Claude, Cursor, or similar tools, you're writing prompts constantly. Good prompts are verbose — they include context, constraints, examples, and specific instructions. Typing a 200-word prompt takes two to three minutes. Speaking it takes forty-five seconds.

OnType's Compose mode is particularly useful here. Speak your prompt naturally — including corrections and clarifications — and the AI rewrite engine cleans it up into a structured, clear prompt. Scene detection recognizes that you're in an AI chat interface and optimizes the output for prompt quality: separating context from instructions, making constraints explicit, removing conversational artifacts.

Git commit messages and PR descriptions

"fix bug" is the commit message of someone who didn't want to context-switch. Voice typing makes it trivial to dictate a proper message: "fix race condition in the connection pool where two goroutines could acquire the same connection if the health check timed out during a resize event." That takes four seconds to say.

Slack and async communication

Explaining a technical decision over Slack often requires three paragraphs. Voice typing turns that into a 30-second dictation. OnType's Compose mode is useful here too — speak your raw explanation with all its tangents and corrections, and the rewrite engine produces a clean, structured message.

Terminal and CLI input

OnType works in terminal emulators — iTerm2, the built-in Terminal, Warp, and others. This means you can dictate long command arguments, heredoc content, or even interactive prompts. The text appears at the cursor position just like it does in any other app.

Why it needs to be system-wide

Some voice tools only work in their own window or in specific apps. For developers, this defeats the purpose. The whole point is not having to leave your current context — your editor, your terminal, your browser with the GitHub PR open.

OnType works system-wide. Hold the hotkey in VS Code, text appears in VS Code. Hold it in iTerm2, text appears in iTerm2. Hold it in a GitHub comment box in Firefox, text appears there. No copy-pasting from a separate dictation window. No app-switching.

Why it needs to be on-device

Developers are often working with proprietary code, internal documentation, and confidential project details. Sending audio of you discussing implementation details to a cloud service creates an unnecessary data exposure surface.

OnType's default engine runs entirely on your Mac's Apple Silicon chip. Your audio — including every mention of internal APIs, architecture decisions, and unreleased feature names — stays on your device.

Getting started

If you spend more than an hour a day writing things that aren't code, download OnType and try it for a week. The on-device engine is free. The workflow change is immediate: hold key, speak, release, keep coding.

Read more about voice typing for developers or check our getting started guide for setup details.