That is near the success rate I have as well. And when I dictate a text on my phone voice works great because I can retype the one in twenty words it gets wrong. There's no keyboard on a watch. Worse, when you are misunderstood Wear is likely to launch the wrong app or pull up the wrong information. So you have the problem of canceling one thing and trying, awkardly, to get to what voice didn't understand.

If you look on the wear store, apps that create rows of icons on your watch face are very popular. I think that shows that this is a common concern.

Your post has 47 words. A 95% success rate means 1 in 20 failure, so you're going to have to stop, swear, fix it, and restart dictation about two or three times in your post. Even the somewhat optimistic 99% still gives only a systemic fifty fity odds of your post being error free.

Its just too annoying for most people to use.

What voice recognition reminds me of, is a disobedient / distracted little kid, the kind you have to tell everything twice.

> A 95% success rate means 1 in 20 failure

I'm not sure my success rate is any better than that typing on my iPhone.

But the correction is simpler and less distracting. I can backspace the word I misspelled and fix it without breaking my thought flow, whereas when I try that with dictation... well, it doesn't work that way. Especially that the whole point of dictation should be that you're not actually looking at the screen and double-checking if what's recognized is what you said.

"Especially that the whole point of dictation should be that you're not actually looking at the screen"

I've noticed UIs always get "needier" over time. Call it VLM's Law. In the old days a dumbphone could be operated tactile in total darkness without looking at it at all. Then smartphones came along which needed occasional glances at the screen to verify the screen and touch screen position/orientation (portrait vs auto-rotated landscape). Now you need to stare continuously at the screen to watch mistakes in the speech to text dictation system as you slowly, tediously, agonizingly argue with a machine.

How clumsy of a UI, like going thru life wearing oven mitts. And VLMs Law is it always gets worse over time. It'll take a lot of work, but we'll find a way to make UIs even more painful.

Yeah. What I loved about dumbphones and feature phones was not just keys - it was that they had firmware instead of a full-blown OS, and thus predictable timing. I could operate my K800i in total darkness without looking at it, because I knew that, say, "joy-click, wait for around half a second, left, joy-click, down, joy-click" would take me to my messages. If I tried something like that with my S4, I'd end up pretty much wherever, because the UI likes to hang for half a sec every now and then.

#nostalgia

Voice recognition has improved tremendously for me as well. But what I can't seem to get over is that it's almost always slower than using my thumbs. Talking is slower than typing, plus the process time needed to interpret the voice commands. -- from a keyboard shortcut junkie