Coding with Voice — SpeakUp for Developers

Here is a number that might surprise you: developers spend roughly 60% of their working time on activities that are not writing code. Pull request descriptions, code review comments, documentation, Slack messages, email, issue descriptions, commit messages, RFC documents, onboarding guides, architecture decision records. The average developer writes thousands of words of prose every week — and types every one of them.

SpeakUp is not a tool for writing code by voice. It is a tool for everything else — the 60% of your day that is prose, not syntax.

The Developer Prose Problem

Think about your last workday. How much time did you spend on these tasks?

  • Writing a PR description that explains why you made the changes, not just what changed
  • Reviewing a colleague's PR with substantive, helpful comments
  • Explaining a technical decision in Slack or a design document
  • Drafting a bug report with clear reproduction steps
  • Writing or updating documentation
  • Composing an email to a stakeholder explaining a timeline

Every one of these tasks requires clear writing. None of them require a keyboard. They require your thoughts, organized and expressed — and speaking is roughly three times faster than typing for most people.

How It Works in Your Editor

SpeakUp types into whatever application has focus. This means it works in VS Code, Cursor, JetBrains IDEs, the terminal, your browser, Slack — anywhere you have a text cursor.

The workflow is simple: click into the text field where you want to write (a PR description on GitHub, a comment field in your code review tool, the commit message input in your terminal UI), press your SpeakUp hotkey, speak, press the hotkey again. Your words appear as typed text. No clipboard, no paste — it simulates keystrokes directly.

For commit messages, this is surprisingly effective. Instead of typing git commit -m "fix: resolve race condition in auth middleware", you press your hotkey and say it. For longer commit messages with body text, dictation is dramatically faster than typing.

For PR descriptions, the difference is even more pronounced. You know exactly what you changed and why — dictating that explanation takes 30 seconds instead of the 5 minutes of typing that often results in a terse "fixed the thing" description nobody finds useful.

RSI Prevention — A Practical Concern

Developers have among the highest rates of repetitive strain injury of any profession. Hours of daily keyboard use, combined with mouse movements and poor ergonomics, take a cumulative toll. Carpal tunnel syndrome, tendinitis, and other RSI conditions can become career-threatening if left unaddressed.

Dictation is not just a productivity tool — it is an ergonomic one. Every sentence you speak is a sentence your wrists do not have to type. Even using dictation for 30% of your prose output meaningfully reduces the repetitive load on your hands. Some developers start using dictation after an RSI diagnosis; the smarter ones start before.

Why Developers Appreciate the Architecture

SpeakUp is built on whisper.cpp — an open source C/C++ implementation of OpenAI's Whisper model. It runs on Apple's Metal GPU framework, entirely on your Mac. No data leaves your machine. No cloud API. No account.

For developers, this matters in a few specific ways:

No network dependency. SpeakUp works on an airplane, in a coffee shop with terrible WiFi, and behind corporate firewalls that block cloud services. If your Mac is on, dictation works.

No API keys, no tokens, no auth. You download it, open it, and it works. No OAuth flow, no account creation, no team admin panel. This is a tool, not a platform.

Inspectable technology. whisper.cpp is open source. The model weights are published. You can read the inference code, understand the architecture, and verify that the transcription is happening locally. There is no black box.

No telemetry. SpeakUp collects nothing. No usage analytics, no crash reports phoned home, no "anonymous" data collection. If you run Little Snitch or a network monitor, you will see zero outbound connections from SpeakUp.

What It Is Not

SpeakUp does not write code for you. It does not autocomplete function signatures, generate boilerplate, or translate natural language into Python. That is what Copilot and similar tools do, and they do it well.

SpeakUp does one thing: it turns your speech into text, accurately, in whatever application you are using. For the majority of a developer's communication work — the PRs, the docs, the reviews, the messages — that is exactly what is needed.

Getting Started

Download SpeakUp at getspeakup.app. It costs €29 once — no subscription, no team pricing, no "enterprise tier" where the useful features live. There is a 14-day free trial with no account required.

For more on how SpeakUp fits developer workflows, see the Developers page. For benchmarks on how whisper.cpp performs on different Mac hardware, see whisper.cpp Benchmark on Mac.

Try SpeakUp Free for 14 Days

No credit card. No account. No cloud. Just download and start dictating.

Download Free Trial