Why We Never Upload Your Voice

Most dictation tools quietly stream your microphone to a server, transcribe it there, and send the text back. BlaBlaType does not. Your voice is transcribed on your own Mac and never leaves the device. Here is exactly how that works, and why we built it that way on purpose.

What "we never upload your voice" actually means

It is an easy phrase to print on a landing page, so let us be precise. When you dictate with BlaBlaType, the microphone audio is captured, filtered for silence, and fed straight into a speech-to-text model that is already sitting on your Mac. The transcript is produced locally, cleaned up locally, and pasted into whatever app your cursor is in. At no point is the raw audio, or the resulting text, sent to a remote machine. If you are curious about the general question of where your voice goes when you dictate, the honest answer for cloud tools is "to a data center you cannot see." For us, the answer is "nowhere."

This is the opposite of how most polished dictation apps work. Cloud dictation needs your audio on its servers to run the heavy model, which is also why those tools stop working the moment you lose signal. Because our processing is local, none of that applies.

Why we designed it this way

Privacy was not an afterthought bolted onto a cloud product; it was the starting constraint. Once you decide that audio must never leave the Mac, a lot of good things follow for free. There is no upload latency, no per-minute server bill, and no outage that can take your dictation offline. It also makes the app honest by construction: we cannot leak what we never receive. For work under an NDA, client notes, medical or legal drafts, or just private thoughts you are talking through, that guarantee is the whole point. If you want the broader argument, we covered whether Mac dictation is actually private in a separate piece.

There is a productivity angle too. Most people speak around three to four times faster than they type, so a dictation tool you can trust with sensitive text removes the last reason to fall back to the keyboard.

On-device versus cloud dictation

Cloud tools are not evil, and some are genuinely polished. But the trade is structural: to get their model, your audio has to travel. If you are weighing this up, it helps to know what actually needs internet in a dictation app, because the answer for a well-built local tool is "only the one-time model download."

The honest trade-offs of staying local

We are not going to pretend on-device processing is free of compromise. It is the right choice for privacy, but it asks a little of your hardware, and we would rather you hear that from us.

For most people on a modern Mac, the "asks" column is a non-event: you download once and forget it. The gains, by contrast, apply every single time you speak. That is why the tool is built around matching the app you are in and staying out of the way, so much so that we wrote about dictation that adapts to the app you are in separately.

Where local dictation fits in real workflows

Keeping voice on the device is not just a privacy stance; it changes how the tool feels day to day. Because it runs system-wide, you dictate the same way into email, Slack, Notion, and a code editor. Developers often pair it with an AI IDE such as Cursor to talk through a prompt instead of typing it, and the transcript still never leaves the machine. It is also a genuinely kinder input method for people who find typing draining: we looked at voice to text for ADHD in its own guide. Whatever the workflow, the privacy model is the same, which is the point. You can read more about the overall approach on the BlaBlaType home page or compare tiers on the pricing page.