Voice Dictation for Blind Users: How a VoiceOver Review Reshaped Our Keyboard

A few months ago, a review landed on our App Store page that changed how we build Quill Flow. It came from a blind user in Germany. He was happy — he called the app "quite accessible" with VoiceOver, and said he had been looking for a way to improve on native iOS dictation for months. Then he described, precisely and patiently, the two things that made it slower than it needed to be.

First: when he started a dictation, VoiceOver's focus stayed wherever it was. The button to finish recording existed, but he had to search for it by touch, key by key.

Second — and this is the part most sighted developers never think about — while he searched, the microphone was already recording. Every key name VoiceOver read aloud went straight into his transcript. "Button. Speak. Selected." — transcribed ahead of the words he actually wanted to say.

This post is about what we shipped in response, how to use Quill Flow with VoiceOver today, and why voice dictation is such a strong fit for blind and low-vision iPhone users in the first place.

Why dictation matters more when you can't see the keyboard

Typing on glass is slow for everyone. With VoiceOver it's slower still: standard typing means touch-exploring to a key, then double-tapping it — for every letter. Experienced users get fast with touch typing and braille screen input, but even then, composing a long message character by character takes real effort.

Speech flips that equation. Speaking is three to four times faster than sighted thumb-typing — compared to character-by-character screen reader input, it's not even close. For blind users, a dictation tool that works properly isn't a convenience feature. It's often the fastest way to write, full stop.

The catch is that "works properly" has a higher bar with a screen reader running. The tool has to be operable without sight, confirm what it did out loud, and — crucially — not transcribe the screen reader itself.

What we changed after that review

1. The magic tap: start and stop dictation from anywhere

VoiceOver has a system-wide gesture for "do the main thing": the two-finger double-tap, or magic tap. It answers calls. It plays and pauses music. In Voice Memos, it starts and stops recording.

In Quill Flow, the magic tap now starts a dictation, and another one finishes it — no matter where VoiceOver's focus happens to be. There is nothing to find. The gesture also produces no spoken announcement of its own, which means nothing extra ends up in your transcript.

2. Focus moves to the finish button on its own

If you prefer the buttons, the original complaint is fixed at the root: the moment recording starts, VoiceOver focus automatically moves to the finish control, which announces itself as "Finish dictation." Ending a recording is one double-tap. The cancel button is labeled "Cancel dictation" so the two are impossible to confuse — one keeps your words, one discards them, and VoiceOver tells you which is which.

3. Your screen reader stays out of your transcript

Two layers handle the "VoiceOver gets transcribed" problem. While VoiceOver is running, Quill Flow enables echo cancellation on the microphone, so speech played by your own device is suppressed before it ever reaches the recognizer. And whatever still leaks through is caught by a transcript filter that recognizes VoiceOver's announcements — our own control labels in all 34 languages we ship, plus VoiceOver's role words like "button" and "selected" in every language — and strips them from the start of your text.

The filter only runs when VoiceOver is on, and it's deliberately conservative: real sentences that happen to start with words like "speak" or "selected" pass through untouched.

4. Spoken confirmations, both directions

When your dictation lands, VoiceOver reads back exactly what was inserted — the same confirmation Apple's built-in dictation gives, so you never wonder whether your words made it. Recording started, paused, resumed, and stopped are all announced. If a burst contained no speech, you hear "No speech detected" instead of silence.

5. A keyboard where everything has a name

Every key and control in the keyboard carries a proper VoiceOver label and trait: letters, Shift (including whether it's on or caps-locked), Delete, Return in all its forms, the globe key, emoji search, the settings drawer. Labels are localized in 34 languages — a German VoiceOver user hears "Diktat beenden," not English filler.

Using Quill Flow with VoiceOver: quick start

1. Download Quill Flow from the App Store and open it once — the onboarding walks you through adding the keyboard in Settings → General → Keyboard → Keyboards, and enabling Allow Full Access (that permission is what lets any keyboard extension record audio).
2. In any app, move to the text field and switch keyboards with the globe key — announced as "Next keyboard."
3. Two-finger double-tap. Speak. Two-finger double-tap again. Your text is formatted — filler words removed, punctuation added — then inserted and read back.
4. To type instead, use standard or touch typing exactly as on the system keyboard; every key is labeled and swipe typing simply stays out of your way.

How this compares to Apple's built-in dictation

Apple's dictation is genuinely accessible to start and stop — but what it produces is raw. No automatic punctuation cleanup, filler words left in, no formatting for the context you're writing in, and editing mistakes afterwards is exactly the character-by-character work dictation was supposed to avoid. That editing cost hits screen reader users hardest.

Quill Flow's pitch for blind users is that the text that lands is already clean: punctuation placed, false starts removed, formatted for a message or an email, in 100+ languages. Fewer corrections means the speed advantage of speaking actually survives contact with reality. For the fuller comparison, see Quill Flow vs Apple Dictation.

What's still imperfect

Honesty is part of accessibility. iOS keyboard extensions live inside real platform constraints, and a few decorative corners of Quill Flow — the mini-games, some visual flourishes — aren't meaningfully usable with VoiceOver yet. Our accessibility page tracks what ships and what's next: quillflow.app/accessibility.

And if something in your setup doesn't work — a braille display quirk, a language where an announcement reads wrong — email hello@quillflow.app. The features in this post exist because one person took the time to describe what was slowing him down.