Community

AirTypes Is Growing — How a Diverse Community of Voice Typists Is Changing Their Workflows

When we started AirTypes, we imagined it mostly for developers who didn't want to type. What's actually happening is more interesting — people from dozens of different fields, backgrounds, and countries are adopting offline voice dictation, and many of them are talking to each other about it. Here's what the growing community looks like, and what's driving it.

Who is actually using AirTypes?

Voice dictation used to mean one thing: you dictate a letter or a report, a human types it up. Then it meant: you speak into a cloud microphone and a server guesses what you said. Neither of those is what AirTypes is.

AirTypes is a desktop app that runs entirely on your machine, injects text at your cursor in any app, and — optionally — routes your speech through your own AI. That combination is surprisingly versatile. It turns out a lot of people have been quietly waiting for something like this.

Looking at where AirTypes users come from and how they use it, several distinct groups emerge. They didn't choose AirTypes for the same reasons. But they're all sticking around for the same one: it actually works, and it works offline.

Developers: voice for code comments and docs

Developers are the most visible slice of the AirTypes community, and their use case is not what you might expect. Almost none of them are dictating code directly.

What they are doing:

  • Dictating commit messages and PR descriptions while keeping both hands on the keyboard
  • Narrating code comments in plain English — speaking "this function takes a list of user IDs and returns only the ones with active subscriptions, sorted by created_at descending" is faster than typing it and usually results in a better comment
  • Writing inline documentation with the My Agent feature — speak your intent, let your own Claude or GPT-4 key rewrite it into the right format for your codebase's style
  • Drafting Slack messages and emails without breaking focus from the terminal

The common thread is that developers use AirTypes for the natural language parts of their work — the bits that are hardest to type efficiently. Code is still typed. Everything around the code is spoken.

"I didn't realise how much of my day was spent typing things that weren't code. Now most of that is voice. The actual coding hasn't changed at all." — Backend engineer, remote team

Writers: speaking is faster than the blank page

The second major group is writers — and "writer" here covers a wide range: journalists, novelists, content creators, technical writers, bloggers, and anyone whose job is to produce text at volume.

Writers don't have a typing speed problem. They have a blank page problem. Speaking out loud short-circuits it in a way that typing doesn't, for reasons that aren't entirely understood but seem to involve how the brain relates to speech versus writing. You don't edit yourself as aggressively when you're speaking. You get more down faster. Then you edit.

The offline angle matters here too. Many writers work in unusual places — libraries, trains, coffee shops. Being able to dictate without needing a connection, without worrying about who's processing your words, is a practical and creative relief.

A common workflow:

  1. Hold the hotkey, speak a rough paragraph or idea
  2. Let AirTypes inject the transcription directly into your editor (Obsidian, iA Writer, VS Code, Word, anything)
  3. Edit the transcript by typing as normal
  4. Repeat

Some writers use My Agent to run their spoken draft through a cleanup prompt before it hits the cursor. Others prefer the raw transcription so they can edit themselves. Both workflows work.

Non-native English speakers: a different kind of unlock

This is the group that surprised us most. A significant portion of AirTypes users work in English as a second, third, or fourth language.

For them, the bottleneck isn't typing speed — it's cognitive load. Thinking in one language while composing in another is exhausting. Typing slowly while searching for the right word creates a different kind of friction than speaking does.

The My Agent feature is particularly useful here. You can speak in your natural English — the way you'd say it in conversation, even imperfectly — and have your own AI clean it up into polished written English before it reaches your cursor. You're not sending your audio to anyone. You're using your own API key. The result is professional-quality written output from natural spoken input, entirely offline at the transcription stage.

Several users in this group have mentioned that using voice dictation helps them feel less self-conscious about their English. Speaking and then seeing the cleaned-up result is less nerve-wracking than staring at a partially-typed sentence that doesn't look right yet.

"My written English is fine but composing it is slow. With AirTypes I just speak how I think and the result is good. It changed how much I can write in a day." — Researcher, working in English as a third language

Remote workers and async-first teams

Remote work created a surge in written communication. Slack messages. Long async threads. Documentation. Feedback on documents and PRs. All of it typed, all day.

AirTypes fits naturally into async workflows. You hold your hotkey, speak your message, and it appears wherever your cursor is — in Slack, Notion, Linear, GitHub, email. There's no app to switch to. No clipboard to manage. You don't break your reading flow to start typing.

For teams spread across time zones, a lot of communication happens in long-form writing. Speaking is faster than typing for most people once it's a habit. That speed compounds across a full day of async communication.

What connects all of them

Different use cases, different reasons for trying AirTypes, but a few things are consistent across all of them:

  • They found dictation uncomfortable before. Cloud-based dictation requires trust. Sending audio to a server, not knowing who processes it or for how long, feels wrong for a lot of people. Knowing everything runs on their own machine removes that discomfort entirely.
  • They weren't looking for a dictation app. Most users came via word of mouth — a colleague mentioned it, a post mentioned privacy and offline, something in a workflow thread. The first impression is usually "I didn't think I'd use this" followed quickly by "I use this every day now."
  • The hotkey model works. Press and hold to speak, release to inject. No mode switching, no separate window, no clipboard. It fits into existing workflows without asking you to change how you work.

The tipping point: when dictation becomes a habit

The most common pattern we see is a slow start followed by a sharp uptick in usage around day 4–7.

The first day or two, people are self-conscious about speaking to their computer. Accuracy feels surprising ("it actually got that?") but the workflow feels unfamiliar. People hold the hotkey and then hesitate.

By day 4 or 5, they're no longer thinking about the tool. They're thinking about what they want to say, and the tool is getting out of the way. Usage statistics in the AirTypes dashboard start showing the shift: transcription count goes up, session duration goes down. People are getting more done in less time.

By the end of the first week, most users who stay past day 3 are still using AirTypes 60 days later. The retention isn't because we've locked them in — there's no lock-in, no data held hostage. It's because speaking faster than typing isn't a novelty once you've proven it to yourself. It's just a better way to write.

What's coming for the community

The AirTypes roadmap is shaped heavily by what the community asks for. A few things that keep coming up:

  • Windows support. It's the most-requested platform. It's on the roadmap. We're not giving a date we can't keep, but it's coming.
  • More agent profiles. Power users are building elaborate My Agent configurations — one profile for code documentation, another for formal email, another for casual Slack messages. Making profiles easier to create and share is a priority.
  • Better vocabulary customisation. Technical users want to add their own jargon — domain-specific terms, names, product nouns. On the roadmap.
  • Community spaces. People want to share their workflows, compare agent prompts, and ask questions. We're thinking about where this happens best without becoming noise.

If you've been watching AirTypes from the sidelines, now is a reasonable time to try it. The 7-day free trial is genuinely unrestricted — you get the full feature set, all accuracy tiers, My Agent, everything. No credit card to start.

Join the community

Download AirTypes free for 7 days — full features, no credit card. macOS and Linux builds available.

Download AirTypes