Voice dictation sounds like a gimmick until you realize how much of a developer's day isn't code.
Emails. Slack. Documentation. Code reviews. Meeting notes. Pull request descriptions. A lot of what you type every day is prose, not syntax.
That's where voice dictation earns its keep.
What works well
Prose writing is where dictation shines. Emails, Slack messages, documentation, blog posts, README files. If you're writing in a natural language, speaking is almost always faster than typing.
Code reviews are a good fit. You're describing problems, suggesting approaches, explaining trade-offs. That's all natural language. Dictate your review comments instead of typing them.
PR descriptions take forever to type and most developers write terrible ones because of it. Speaking a detailed PR description takes two minutes. Typing one feels like work. The result is better descriptions and better code review.
Meeting notes are obvious. You can't type and listen at the same time. You can speak a summary into a push-to-talk app right after the meeting ends.
What doesn't work
Code itself is where voice dictation fails. Variable names, function signatures, brackets, semicolons. These don't translate from speech to text reliably.
There are voice coding tools like Talon and Cursorless that are designed specifically for this. They're impressive but have a steep learning curve. Worth researching if RSI is a serious concern.
Terminal commands have the same problem. git rebase -i HEAD~3 doesn't dictate cleanly.
Anything requiring precision is safer to type. Configuration files, SQL queries, regex patterns.
The push-to-talk workflow
The best dictation apps use push-to-talk mode. Hold a key, speak, release. Text appears at your cursor.
This is important because it solves the pacing problem. You can pause and think without capturing silence or filler words. You can stop speaking, move your cursor, and start again.
dictate.app does this well on Windows. It works in any app because it types the text using simulated keystrokes. No browser tab required. No switching windows.
Tips for dictating technical content
Spell out the unusual stuff. For a variable name like getUserById, say "get user by ID" and then manually type the camelCase version. Dictate the surrounding prose, type the precise bits.
Use short sentences. Voice dictation accuracy improves when you speak in clear, complete sentences. Long run-on sentences with lots of clauses create more errors.
Trust the edit pass. Don't stop and correct every mistake as you go. Speak the whole thing, then do one editing pass at the end. You'll move faster.
Speak punctuation when needed. Most modern dictation tools handle punctuation automatically based on context. When they don't, you can say "comma", "period", "new paragraph" explicitly.
Getting started
Pick a week where you have a lot of email and documentation work. Try dictating everything non-code for five days. Pay attention to where it saves time and where it slows you down.
The first two days will feel awkward. That's normal. By day five you'll have a clear picture of whether it fits your workflow.
dictate.app has a 7-day free trial if you want to experiment on Windows without committing.
