You owe someone a real explanation — a friend you went quiet on for three weeks, a landlord, a coworker who needs a real update. You open the thread and start typing. Sentence one is fine. By sentence three your thumb is cramping, autocorrect ate a word, and you're rereading instead of finishing the thought. Four minutes in: three sentences, and a message that still doesn't say the thing.
So you cut it down to "sorry, been slammed, call you soon." Technically true. A fraction of what you meant — not because you had less to say, but because typing it all out on a phone was more friction than the moment could survive.
The actual speed gap
This isn't a willpower problem, it's mechanical. A large 2019 study of everyday mobile typing (Palin, Feit, Kim, Kristensson, and Oulasvirta, "How do People Type on Mobile Devices?", MobileHCI 2019, ~37,000 participants) found average touchscreen typing speed of around 36 words per minute under real-world conditions. Conversational speech is widely estimated to run several times faster — figures around 130 to 150 words per minute show up repeatedly in speech research. The exact numbers vary study to study; the shape of the gap doesn't.
| Thumb-typing | Raw dictation | Dictation + AI cleanup | |
|---|---|---|---|
| Speed for a long message | Slow — several minutes for ~300 words | Fast — under ~2.5 minutes for ~300 words | Fast, plus a few seconds of polish |
| Output quality | Clean, but often cut short to save effort | Messy — filler, false starts, run-ons | Clean and complete, still your own words |
| Best environment | Anywhere, including loud/public spaces | Quiet, private spaces | Quiet, private spaces |
| Precision (names, spelling, jargon) | High — you control every character | Lower — speech recognition can fumble it | Same as raw dictation for precision |
| Risk of sending less than you meant | High — length gets sacrificed for effort | Low | Low |
Where dictation genuinely shines
- Messages you're already re-drafting in your head
- Notes to self that otherwise never get written down
- Emails and long explanations with real context
- Anything past a couple of sentences — the break-even point isn't subtle
Where it doesn't — and that's fine
- Loud or public spaces — background noise degrades transcription, typing wins
- Sensitive conversations — if you don't want to be overheard, don't say it out loud
- Precise names, jargon, spelling — speech recognition still fumbles this more than typing does
None of this makes dictation unreliable. It's a tool for the ordinary bulk of messages, not a replacement for typing in every situation.
The workflow: talk first, clean up second
Not "let AI write your message." Dictate first, then clean up.
Say it the way you'd actually say it out loud — messy, with false starts and an occasional "um." Getting the true version out fast, without stopping to self-edit mid-sentence, is the entire point of talking instead of typing.
Vavus Keyboard's dictation includes a free AI text-polish pass that tightens what you actually said — trims filler, fixes run-ons, smooths stumbles — without inventing content or changing your meaning. It's not ghostwriting. It's the same relationship a rough voice memo has to a clean transcript, not the relationship a prompt has to an AI-generated essay.
It works the same way on desktop via a hotkey, not just on your phone.
What it costs
Vavus Keyboard is $14.97/month on web ($14.99 on Apple) for unlimited dictation and translation, or pay-as-you-go with tokens. The AI cleanup pass is free on top of dictation.
The bottom line: most people default to typing because it's the habit, not because it's faster. For a quick reply, keep typing. For anything longer, say it, then clean it up.
Originally published at vavusai.com.
Top comments (0)