Voice Dictation in VS Code on Windows

May 2026 · 8 min read

The honest answer: you cannot dictate raw code. Brackets, semicolons, camelCase, and operator symbols do not come out of speech-to-text cleanly. Anyone who tells you otherwise is selling something.

But here is what most developers miss: raw code is not most of what you write.

That 80% is where voice dictation pays off immediately. No learning curve. No training. Just speak and the text appears.

What You Can Dictate in VS Code

dictate.app uses clipboard injection. It works at the OS level. Wherever your cursor sits and wherever Ctrl+V works, dictation works. In VS Code that means:

Code comments. Single-line comments, block comments, inline explanations. Speaking a comment is faster than typing it and you get more natural language out of it. "This function handles the edge case where the user disconnects before the response completes" comes out better spoken than typed.

JSDoc and docstrings. Documentation is the #1 thing developers skip because it is tedious to type. Speaking it takes a quarter of the time. Describe the parameters, the return value, the side effects. Done in 20 seconds.

Commit messages. The VS Code source control panel has a text field. Hold your hotkey, describe what you did, release. A real commit message in 5 seconds instead of "fix stuff."

README files. Markdown documents are ideal for dictation. No syntax to worry about. Just speak the content. You can dictate a full README in under 3 minutes.

GitHub issue descriptions. VS Code's GitHub extension shows issue text fields. Dictate bug reports, feature requests, and pull request descriptions directly.

Terminal commands you already know. If you type the same git commands, npm scripts, or shell commands repeatedly, you can dictate them. "git push origin main" comes out fine. Complex flags do not.

What You Cannot Dictate

Raw code syntax is hard. These specific things fail:

Variable names with underscores or camelCase. "user underscore id" usually comes out as "user underscore id" literally, not "user_id." Same problem with camelCase.

Operator-heavy lines. "const result equals arr dot filter open paren item arrow item dot active close paren" is not faster than typing it.

File paths. Slashes, dots, and directory structures are painful to dictate reliably.

The 80/20 Breakdown

How dictate.app Works in VS Code

No VS Code extension needed. dictate.app runs as a background Windows app.

The setup: install dictate.app, set your hotkey (default is middle mouse button, but any key combination works), and start using it. Hold the key while speaking, release when done, text pastes at your cursor in VS Code or any other app.

The transcription uses Groq's Whisper API. It comes back in about 200 milliseconds. That is fast enough that it feels instantaneous in normal use.

Push-to-talk matters especially for developers. You are not broadcasting to the office. You dictate one comment, release the key, keep coding. Nobody around you knows you are using it.

VS Code Voice Extension vs dictate.app

VS Code has an experimental voice extension. It requires a GitHub Copilot subscription and only works inside VS Code. It cannot be used in your browser, Slack, email, or any other app.

dictate.app works everywhere. One tool. One hotkey. Every application on Windows.

Real Workflow: A Developer's Day

Morning standup notes dictated into Slack: 20 seconds instead of 90.

JSDoc for a new function: speak the description while looking at the code, done in 15 seconds.

Commit message: "refactor user authentication to use the new token validation middleware" dictated in 4 seconds.

GitHub PR description: 200 words spoken in 90 seconds instead of 8 minutes of typing.

Bug report to your team: dictated into Slack while reproducing the bug, instead of stopping to type.

The typing you save is not the exciting typing. It is the tedious surrounding work that slows you down.