If you are a software developer, you probably think voice-to-text has nothing to do with your work. You write code. Code has brackets and semicolons and variable names that no dictation tool can handle. Fair enough. But here is the thing: you also write a lot of stuff that is not code.
Pull request descriptions. Code review comments. Documentation. README files. Slack messages to your team. Jira ticket updates. Commit messages. Emails to product managers. Confluence pages. Architecture decision records. Incident postmortems. Interview feedback. Performance review notes. Stand-up updates.
Studies of developer time allocation consistently show that 30-40% of a developer's day goes to communication and documentation, not coding. You type at maybe 60-80 words per minute. You speak at 130-150. That means every PR description, every Slack thread, every doc page takes roughly twice as long as it needs to.
Voice-to-text will not help you write code faster. But it can dramatically speed up everything around the code. This guide covers the tools that work for developers, where voice-to-text fits in a dev workflow, and where it does not.
Quick Comparison
| Tool | Price | Best For | Tech Vocab | System-wide | Offline | AI Cleanup |
|---|---|---|---|---|---|---|
| Steno | Free / $4.99/mo | All dev writing | Strong | Yes | Yes | Yes |
| Superwhisper | $8/mo | Quick dictation | Good | Yes | Yes | Yes |
| Voibe | Free / $5/mo | AI-first drafting | Good | Yes | No | Yes |
| Talon | Free | Hands-free coding | Excellent | Yes | Yes | No |
| macOS Dictation | Free | Basic text input | Poor | Yes | Partial | No |
The Non-Code Writing Problem
Let us break down where developers actually spend their writing time and why it matters.
Pull request descriptions
A good PR description explains what changed, why it changed, how to test it, and what reviewers should pay attention to. Writing a thorough one takes 3-10 minutes. Most developers write thin descriptions because the effort feels disproportionate to the value. But thin PR descriptions slow down reviews, create back-and-forth questions, and make the codebase harder to understand months later.
With voice-to-text, you can dictate a comprehensive PR description in 30-60 seconds. "This PR refactors the authentication middleware to use JWT refresh tokens instead of session cookies. The main change is in auth middleware dot ts where we now validate the access token first and fall back to the refresh token if it is expired. I also updated the user model to store refresh token hashes. To test, log in normally, wait for the access token to expire which is set to 15 minutes in dev, and verify that the next request triggers a token refresh without logging the user out."
That took about 25 seconds to speak. It would take 2-3 minutes to type. And because speaking is lower friction, you are more likely to include the context that makes the review go smoothly.
Code review comments
Good code review comments explain why something should change, not just what. "This could cause a race condition if two requests hit this endpoint simultaneously because the read and write are not atomic. Consider wrapping lines 42 through 48 in a transaction or using an optimistic locking pattern." That is a useful review comment. It takes 10 seconds to speak and 45 seconds to type.
Documentation
Nobody likes writing docs. Part of the reason is that writing docs on a keyboard is slow and tedious. Dictation makes it faster and more conversational, which often produces better documentation. Technical writing that sounds like someone explaining something to a colleague is usually clearer than writing that sounds like someone filling out a template.
Slack and async communication
Developers spend an average of 90 minutes per day on Slack. Most of that is typing. A thoughtful response to a technical question might take 2-3 minutes to type. With dictation, you speak your response in 30 seconds, clean it up in 10, and move on. Over a full day of Slack communication, that difference compounds significantly.
Detailed Tool Reviews
Steno
Steno is a dictation app for Mac and iPhone that works system-wide. Press a keyboard shortcut, speak, and your words appear wherever your cursor is: VS Code, Terminal, GitHub in the browser, Slack, Jira, anywhere. It uses OpenAI Whisper for transcription and adds profession-aware vocabulary on top.
For developers specifically, Steno handles technical terminology well. It knows the difference between "React" and "react," between "TypeScript" and "type script," between "Kubernetes" and any mangling a generic dictation tool might produce. When you say "refactor the gRPC endpoint to use protobuf serialization," it transcribes that correctly because it understands developer vocabulary.
Steno's AI actions are useful for developer writing. After dictating a rough explanation, you can clean up filler words, shift to a more professional tone (useful for documentation), or shorten the text (useful for commit messages). Dictate a paragraph explaining your changes, then tap "shorten" to get a one-line commit message.
Steno processes audio on-device. Your dictation never leaves your Mac. This matters for developers working on proprietary codebases who cannot risk having snippets of their work sent to third-party servers.
Best for: Developers who want a general-purpose dictation tool that handles technical vocabulary and works in every app. Especially useful if you also want dictation on iPhone. For a broader look at how Steno compares to other Mac dictation tools, see our best dictation app for Mac guide.
Superwhisper
Superwhisper is a Mac dictation app that runs Whisper models locally. It offers multiple model sizes so you can trade accuracy for speed. The smaller models transcribe almost instantly. The larger models take a beat longer but handle complex vocabulary better.
For developers, Superwhisper works well for quick bursts of dictation: Slack messages, commit messages, short PR descriptions. It is fast to invoke, fast to transcribe, and gets out of your way. The AI mode can rewrite your dictation in different styles, which is handy for turning casual speech into professional documentation.
Best for: Developers who prioritize speed and simplicity. Good for quick dictation bursts rather than long-form writing.
Honest limitation: More expensive than Steno at $8/month. Less specialized for developer vocabulary specifically. No iPhone companion app if you want cross-device dictation.
Voibe
Voibe takes an AI-first approach. You speak your thoughts and it uses a language model to reshape them into polished text. It is less of a transcription tool and more of an AI writing assistant that takes voice input. Say "write a PR description about migrating from REST to GraphQL for the user service" and it generates a full PR description, not a transcription of what you said.
This is powerful for certain tasks but unpredictable for others. When you want the AI to draft something from a rough idea, Voibe is fast. When you want your exact words transcribed accurately, it can be frustrating because it rewrites rather than transcribes.
Best for: Developers who want AI-assisted drafting more than faithful transcription. Good for generating boilerplate documentation and standard PR descriptions.
Honest limitation: Requires internet because it uses cloud AI models. You cannot control exactly what it outputs. Not suitable when you need your exact words preserved. Can hallucinate details if you give it a vague prompt.
Talon
Talon is fundamentally different from the other tools on this list. It is a hands-free computer control system, not a dictation app. You use a custom command language to control your entire computer by voice: navigate files, write code, edit text, control applications. Developers with RSI, repetitive strain injuries, or disabilities use Talon to continue coding when they cannot type.
Talon can handle actual code syntax through its command language. You say things like "word function" to type function, "op equals" to type =, "snake my variable name" to type my_variable_name. The learning curve is steep. It takes weeks to become proficient and months to reach the speed of keyboard coding. But for people who need it, Talon is transformative.
Best for: Developers who cannot use a keyboard due to RSI, carpal tunnel, or disability. People who want full hands-free computer control, not just dictation.
Honest limitation: Steep learning curve. Not practical as a speed optimization for developers who can type normally. The command language is a new skill you have to learn and practice. Community-supported, not a commercial product, so documentation and support depend on volunteers.
macOS Dictation
The free dictation built into every Mac. Press the microphone key (or Fn Fn) and start speaking. Works in every app. Handles basic English well.
Best for: Quick, casual text entry when you do not want to install anything.
Honest limitation: Terrible with technical vocabulary. "PostgreSQL" becomes "post gray sequel." "Kubernetes" becomes "Cooper Netties" or worse. No AI cleanup. No custom vocabulary. For developer use, you will spend more time correcting errors than you save by dictating.
Where Voice-to-Text Helps Developers (and Where It Does Not)
Where it helps
- PR descriptions: Dictate the what, why, and how in 30 seconds. Use AI cleanup to polish.
- Code review comments: Explain your reasoning by voice. Faster and often clearer than typing.
- Documentation: README files, architecture docs, API docs. Dictation produces more conversational, readable documentation.
- Slack messages: Especially longer technical explanations. Dictate, clean up, send.
- Commit messages: Dictate a full explanation, then use AI to shorten it to one line.
- Jira and project management: Ticket descriptions, acceptance criteria, bug reports. All prose, all faster by voice.
- Emails: To product managers, stakeholders, recruiters. Dictation makes email fast enough that you actually respond promptly.
- Stand-up notes: Dictate what you did, what you plan to do, and blockers in 15 seconds.
- Incident postmortems: Capture the timeline while it is fresh in your mind by speaking it out.
Where it does not help
- Writing actual code: Too much punctuation and special syntax.
const fetchUser = async (id: string): Promise<User> => {is not something you can dictate naturally. Stick to your keyboard for code. - Code with lots of variable names: Dictation tools do not know your codebase's variable names. You will spend more time correcting than typing.
- Quick one-line messages: If you are typing "lgtm" or "sounds good," dictation is slower due to activation overhead.
- Noisy environments: Open offices, coffee shops with loud music. Dictation accuracy drops and you disturb others.
- Pair programming: Your partner does not want to hear you dictating while they are trying to think.
Developer Workflow: Using Steno Day to Day
Here is what incorporating voice-to-text into a developer workflow looks like in practice. This uses Steno as the example, but the workflow applies to any system-wide dictation tool.
Morning: Slack catch-up
Open Slack. Read through overnight messages. When you need to respond to something that requires more than a few words, press the Steno shortcut, dictate your response, release. Clean up if needed, send. A response that would take 2 minutes to type takes 30 seconds.
Coding session: commit messages
You finish a chunk of work and need to commit. Switch to Terminal or VS Code's source control panel. Press the Steno shortcut. "Refactor user authentication to use JWT refresh tokens. Moved token validation into middleware and added refresh token rotation on each use. Updated tests to cover token expiry edge cases." Release. You have a detailed, useful commit message in 10 seconds. Use the AI shorten action if you want a one-liner instead.
Code review
You are reviewing a colleague's PR on GitHub. You spot something that needs explanation. Click in the comment box, press the Steno shortcut. "This approach works but it might cause issues under concurrent load. The read on line 34 and the write on line 38 are not in the same transaction, so two simultaneous requests could both read the old value and then both write, causing one update to be lost. Consider wrapping this in a serializable transaction or using compare and swap." Release. A thorough review comment in 15 seconds.
Documentation
You need to update the onboarding docs for a service you refactored. Open the Markdown file in VS Code. Press the Steno shortcut and start explaining. Dictate for 2-3 minutes. You now have a rough draft of the documentation. Edit to add code examples and formatting that dictation cannot handle. This is still much faster than typing the entire thing, especially because dictation produces a natural, conversational tone that makes documentation more readable.
End of day: Jira updates
Update your tickets with progress. Press the shortcut in the Jira comment box. "Finished the migration to the new auth system. All endpoints are updated and tests are passing. Still need to update the integration tests in the QA environment and verify the refresh token rotation works correctly in staging. Will finish that tomorrow morning." Ten seconds instead of two minutes.
For a more detailed walkthrough of how voice-to-text fits into developer workflows, see our developer workflow guide.
Technical Vocabulary: How Good Are These Tools?
The make-or-break for developer voice-to-text is technical vocabulary. Here are real examples of how different tools handle common developer terms:
| Spoken Phrase | macOS Dictation | Steno | Superwhisper |
|---|---|---|---|
| "Kubernetes pod" | Cooper Netties pod | Kubernetes pod | Kubernetes pod |
| "PostgreSQL migration" | Post gray sequel migration | PostgreSQL migration | Postgres migration |
| "gRPC endpoint" | GRP C end point | gRPC endpoint | gRPC endpoint |
| "React useState hook" | React use state hook | React useState hook | React use state hook |
| "nginx reverse proxy" | Engine X reverse proxy | nginx reverse proxy | Nginx reverse proxy |
| "WebSocket handshake" | Web socket handshake | WebSocket handshake | WebSocket handshake |
Steno's profession-aware vocabulary adapts to developer terminology, which makes a significant difference for terms that generic dictation tools consistently get wrong. The custom vocabulary feature lets you add project-specific terms, internal tool names, and acronyms that no pre-trained model would know.
Tips for Developer Dictation
- Dictate prose, type code. Do not try to dictate code syntax. Use voice for all the writing around your code and your keyboard for the code itself.
- Speak in complete thoughts. Dictation accuracy improves when you speak in full sentences rather than fragments. "The authentication middleware validates JWT tokens" transcribes better than "auth middleware... validates... JWT."
- Use AI cleanup for tone shifting. Dictate in your natural voice, then use cleanup to make it more formal for documentation or more concise for commit messages.
- Add custom vocabulary early. Add your project's specific terms, internal tool names, and unusual acronyms to your dictation tool's custom vocabulary. This prevents repeated corrections.
- Start with Slack. The lowest friction place to start using dictation is Slack messages. They are short, conversational, and frequent. Once dictation feels natural in Slack, expand to PR descriptions and documentation.
- Do not over-optimize. Some messages are faster to type. "LGTM," "done," "will look tomorrow." Do not force dictation where typing is already fast. Use voice for anything longer than about 10 words.
RSI and Ergonomic Considerations
A significant number of developers deal with repetitive strain injuries. Carpal tunnel, tendinitis, wrist pain, finger numbness. These issues are caused by thousands of hours of keyboard use, and they get worse over time. Voice-to-text is not just a productivity tool for these developers. It is a way to continue working without aggravating their condition.
Even developers without RSI benefit from reducing keyboard time. If you type for 8 hours a day, shifting 30-40% of that to voice input meaningfully reduces the strain on your hands and wrists. It is not a replacement for ergonomic keyboards, standing desks, and regular breaks. But it is an underused tool for long-term hand health.
If RSI is your primary concern and you need hands-free coding (not just dictation), look at Talon. It is designed specifically for this use case and covers actual code writing, not just prose.
Frequently Asked Questions
Can I use voice-to-text to write actual code?
You can, but it is slow and frustrating for most people. Code has too much punctuation, special characters, and precise formatting for natural dictation to be faster than typing. Voice-to-text is best for the non-code writing that surrounds coding: PR descriptions, documentation, commit messages, Slack messages, and Jira tickets. Talon is the exception, designed for hands-free coding with a custom command language, but it has a steep learning curve.
Does voice-to-text handle technical terminology correctly?
Generic dictation tools like macOS Dictation struggle badly with technical terms. Apps like Steno that adapt vocabulary to your profession handle developer terminology much better: Kubernetes, PostgreSQL, gRPC, WebSocket, and similar terms transcribe correctly. You can also add custom vocabulary for project-specific terms that no pre-trained model would know.
Can I dictate into VS Code or my terminal?
Yes, if your dictation tool works system-wide. Steno, Superwhisper, and macOS Dictation all work in any text field on your Mac, including VS Code, Terminal, iTerm, the GitHub web interface, Jira, Slack desktop, and everything else. You activate dictation with a keyboard shortcut and text appears wherever your cursor is.
Is voice-to-text practical in an open office?
It depends on your environment. In a quiet open office, speaking at normal volume can disturb others. Some developers use voice-to-text only when working from home or in a private room. Others use it for short bursts like Slack messages and keep the volume low. If you work remotely, it is a non-issue. If you are in an open office full-time, try it for short messages first and see how your team responds.
How much time can developers actually save with voice-to-text?
Most developers type at 60-80 WPM and speak at 130-150 WPM. For pure text output, that is roughly 2x faster. The real savings come from reduced friction: when writing a PR description takes 30 seconds instead of 3 minutes, you actually write one. Developers who adopt voice-to-text for non-code writing typically report saving 30-60 minutes per day, mostly because they communicate more efficiently rather than because they speak dramatically faster than they type.
What is the difference between Talon and dictation apps like Steno?
Talon is a voice control system for hands-free computer use, including coding. It uses a custom command language and requires weeks of practice. Dictation apps like Steno convert natural speech to text. Talon is for developers who cannot use a keyboard. Dictation apps are for developers who want to write prose faster. They solve different problems and can even be used together.
You did not become a developer to write Jira tickets. Let your voice handle the prose so your hands can focus on the code.
Try Steno for developer dictation: trystenofast.today. Works in VS Code, Terminal, GitHub, Slack, Jira, and every other app on your Mac.