Howdy!
I’m back with a bang after taking some time off for serious bug blasting at my home base (and getting my smart meter upgraded may have been a big part of that). While hunting down some a11y bugs with Windows Speech Recognition over there, I found out that inSided is affected as well. This one is more complicated to set up than most other bugs however, so I would strongly recommend reading through my other thread first to get some important pre-requisites out of the way before you attempt to test any of these.
You’ll soon find out why I’ve been away recently. Because yup… This one takes a fair amount of time!
You can probably find similar issues using Windows Narrator, JAWS or NVDA as well, but I tend to use Windows Speech Recognition more during testing because it doesn’t interfere with keyboard/mouse controls as much, while still getting mostly the same results for things where all of them are able to do things. It also sometimes shows a few more visual clues as well about what it’s trying to do (easier for testing!).
I will look at copy/pasting some of the important info over here later as well. But in the meantime, my original thread at my home base will hopefully help you get the show on the road. And yes, this is totally a new way for me to completely blast both
I will probably tidy this OP up later today with my usual template style and formatting as well. Any further bugs related to WSR that I spot will be added to this thread as I’d prefer not to flood this place with loads of threads related to the same thing.
I should also mention that I’ve spent several years training Windows Speech Recognition across two laptops in order to build up a solid training profile and recognise my voiceprint with pretty good accuracy. When you first start out, please bear in mind accuracy will not be perfect. It’ll take at least a few hours of training and use before WSR becomes reasonably reliable (bearing in mind that it doesn’t have any kind of cloud based magic going on and is totally offline!), but you’ll get there. It’s not like Cortana so you can force it not to phone home at all and still get the same results, but you’ll ideally want to use the same device each time because there’s no way to automatically transfer the voice training profile to another device!
It’s going to take me some time to pick up on all the cases where WSR chokes, so I will update this thread whenever I find anything new.
The first ones that I’ve noticed however, are that the user menu in the top right corner is locking onto my username rather than anything more obvious - so it’s a good thing I’ve trained my voice recognition profile to recognise Blastoise and Blastoise186! - meaning I can kinda get to the log out button with my voice, but if someone has a more complicated username and hasn’t trained WSR, it may struggle.
The other really big one right now however, is that it seems to be completely impossible to use WSR for dictation within the New Editor here. I’ve only tried it on the CKEditor 4 based editor that’s out here and not (yet) on the legacy Froala Editor. But for reasons that I can’t work out, not only does WSR seem to be completely unable to automatically lock onto the editor, but even when I manually enter into it, WSR doesn’t realise that I’m focused on a text editor and doesn’t attempt dictation at all. I would probably argue that this is the one that causes the most trouble right now out of the two.
In my case, I’ve been using my laptop Swampert for these tests, which currently runs Windows 10 Home 20H2 64-bit, Build 19042.964 but my voice training profile spans maybe 10 years or so having been originally initialised on an old laptop back in 2011 that ran Windows 7 at the time and carried over since. I’ve tested these bugs using Google Chrome Stable 64-bit, most recently with version 90.0.4430.93 (Official Build).
If you need any more details, please let me know. :)