Why I stopped using Apple Dictation

I used Apple Dictation for about two years. It worked, mostly. Double-tap the Fn key, speak, text appears. Simple enough. But somewhere along the way I started noticing things that bothered me. Small frictions that don't show up in reviews but stack up over months of daily use.

The biggest one: Apple Dictation sends your audio to Apple's servers for processing by default. There's an offline mode, but it requires a separate model download, it's not on by default, and it still ties you to Apple's speech recognition pipeline. You get no control over quality, accuracy, or which model runs under the hood.

For most people that's fine. But I work on things where it's not.

The tradeoff nobody mentions upfront

When you use cloud-based voice to text on Mac, you're not just converting speech to text. You're sending audio to a third-party server where it gets processed, logged in some form, and retained under a privacy policy you agreed to without reading. For casual use like texting a friend or writing a grocery list, that's a reasonable tradeoff.

But try using cloud dictation where data actually matters:

• You're a doctor dictating patient notes that fall under HIPAA
• You're a lawyer summarizing a client call
• You're a developer prompting an AI coding agent about a codebase under NDA
• You're working on a product that hasn't launched yet

Suddenly the convenience of cloud processing starts to look different. You're not just offloading computation. You're offloading custody of the words coming out of your mouth.

Why voice and local-first belong together

There's something specific about voice that makes the privacy question more urgent than it is for, say, a notes app. When you type, you can edit before anything leaves your machine. You think, write, revise. Offline dictation on Mac doesn't work that way. The whole point is that it's fast and unfiltered. You speak the way you think.

That means your dictated text is rawer, more personal, and often more sensitive than what you'd type. The stream of thought you speak into a dictation tool contains things you'd never publish. Half-formed ideas. Names, numbers, context that makes no sense to anyone else. Details you'd never consciously decide to send to a server.

Sending that audio to the cloud feels qualitatively different from uploading a finished document. It's the difference between handing someone your published article and handing them your notebook.

What on-device transcription actually means

I built VOOIS around one constraint: the audio never leaves your Mac. Not during processing, not ever. When you hold the push-to-talk hotkey and speak, transcription runs locally using WhisperKit, Apple's on-device ML framework, on your own Apple Silicon chip. When you release the key, the text lands directly in whatever app is in focus.

No network request happens. You can verify this yourself: turn off WiFi and use it. Works identically. That's not a feature you can fake.

Running locally also means it's fast. No round-trip to a server, no queue, no latency spike when an API is under load. On Apple Silicon, a typical sentence transcribes in well under a second. The speed isn't a selling point I added on top of privacy. It's just what happens when the processing happens three inches from where you're sitting.

The subscription problem is upstream of the pricing

Cloud processing creates economic pressure for subscriptions. If a company runs your audio through their servers at scale, they have real infrastructure costs. Those costs get passed to you monthly, forever, regardless of how much you actually use it.

I find this exhausting. Not because the dollar amount is always significant, but because renting a tool means you never actually own it. If the company changes pricing, gets acquired, or shuts down, your workflow disappears with it. You've spent years building a habit around something you don't control.

On-device dictation flips this. The compute runs on hardware you already own. There's no marginal cost per transcription. VOOIS is a one-time purchase. Not because I'm philosophically opposed to subscriptions, but because the economics of local-first software make it the honest choice.

Who this is actually for

VOOIS is not for everyone. If Apple Dictation works for you, it works. It's free, built in, and for everyday use it's fine.

But if any of this resonates, if you work with sensitive information, if you want an Apple Dictation alternative that keeps audio on your machine, if you want to own your tools rather than rent them, then local-first voice to text on Mac is worth taking seriously.

The people who get the most out of it tend to fall into a few groups: healthcare and legal professionals who need HIPAA-compliant dictation and can't be casual about where audio goes; developers who dictate long, context-heavy prompts to AI coding agents and want that context to stay local; and writers and researchers who use voice as a thinking tool and treat their dictated notes as personal as a journal.

For all of them the question isn't really which dictation app is best. It's whether they want their voice on a server or on their machine. Once you frame it that way, the answer tends to be obvious.

---

VOOIS is an offline dictation app for Mac that runs 100% on-device on Apple Silicon. No account required, no internet connection needed, no subscription.