Speech vs typing
What the research actually says
"Dictation is 3× faster" gets thrown around by every voice product, including ours. Here's where the number comes from, what it does and doesn't mean, and the honest caveats.
Three findings, three sources
Speech input: 3.0× faster
A 2016 study by Stanford, Baidu, and University of Washington researchers (Ruan et al.) had 32 participants enter text by voice and by keyboard. English speech input was 3.0× faster than the smartphone keyboard — with 20.4% fewer errors. Published as "Comparing Speech and Keyboard Text Entry" (arXiv:1608.07323); summary at Stanford News.
Speaking: ~150 words/minute
The National Center for Voice and Speech puts average conversational speaking rate at roughly 150 words per minute for American English. It's the most-cited baseline in speech research and the source of the "150 wpm" figure across the dictation industry (and on this site).
Typing: ~40 words/minute
Large aggregated typing-test datasets consistently put the average adult at about 40 wpm on a full keyboard (e.g. typing-test statistics, 2026). Professional typists reach 65–75; most knowledge workers don't. 150 ÷ 40 is where "roughly 3×" comes from — and it matches what the Stanford team measured.
One disclosure. Blabb's own "time saved" tile is an extrapolation from this same 40 wpm assumption — it multiplies your dictated words by an assumed typing speed. It is a marketing figure, not a measurement of you. Nothing times how fast you actually type, so treat that number as an illustration and the research above as the evidence.
When 3× is real, and when it isn't
Corrections eat the gain
The 3× figure assumes recognition is accurate. Every misheard word costs a round-trip of re-reading and fixing — which is why raw transcription quality and vocabulary learning matter more than headline speed. If you correct every third sentence, you've typed with extra steps.
Composing ≠ transcribing
Studies measure text entry of known phrases. Composing original thought is slower in any medium — but speaking still wins, because you can ramble at thinking speed and let cleanup handle the mess. That's the case for AI cleanup rather than verbatim transcription.
Where Blabb sits
Blabb's bet is the full loop: on-device speech recognition (Cohere Transcribe, which topped the Open ASR Leaderboard on release at 5.42% word error rate, March 2026), optional AI cleanup that removes filler and fixes punctuation without inventing content, and a glossary that learns your names and jargon from corrections — so the correction tax shrinks the longer you use it.
Test the 3× claim on your own work.
Two weeks free, and every word is recognised and cleaned up on your own PC.
FREE TIER 2,000 WORDS/DAY · 14-DAY UNLIMITED TRIAL IN-APP · CANCEL ANY TIME