A Couple VoiceOver Performance Questions From a (Potential) Switcher

My pleasure to answer the first question.
"First, is there a way to disable the speaking of web elements like landmarks, articles, banners, etc.? These all present as individual elements when navigating, and for me, they just slow me down."
This had me mad for about 10 months. This is the main reason why I ditched JAWS on Windows in the first place. NVDA was just so much more intuitive for this.
So yes there is a solution to that exact question. I personally use this everywhere except for select apps, but bear in mind that you must understand a lots of the ins and outs of voiceover to be fully productive with that so don't jump into it first. Basically you want to go in VO Utility, and under Navigation, set Grooping behaviour to anything Ignore groups, or Announce groups (like I do). Why I prefer announce groups over ignore is that with ignore html landmarks for example are not announced at all, and maybe because of habit and liking to have a clear picture of the webpage, even if it's not written properly, I prefer to know when I am in a landmark. You will see yourself more efficient in pwa especially such as microsoft teams, discord and etc, as there will no longer be
Element, group
Element
Element, end, group (or whatever it actually says)
.
There is just a little catch, for whatever reason voiceover won't announce the landmark when you jump into a specific element within that landmark with the rotor or the other voiceover navigation commands for html. For example on applevis:
If you go to the previous heading before the heading level one titled "A Couple VoiceOver Performance Questions From a (Potential) Switcher", NVDA on windows will announce the landmarks, in this case breadcrumb navigation, and then the heading level 2 Breadcrumb. On the mac, when you use voiceover find commands, only that element is announced, and to hear the full hierarchy (if any) you must go by previous/next element.
Also bear in mind that this will change slightly how controls are displaied in safari for example and will give you a preview of some of the inherent little glitches these modes have and why I would not recommend you to go full on into these modes unless you have enough experience with voiceover and perfectly know what to expect.
As a side note, unless I am wrong don't even try to use any of the iwork or notes app in non-standard, and same for garageband and xcode, as as far as I know things are not very usable unless there are specific tips for those. That's why I use announce groups everywhere (and now also when I have mathml content, thank you apple), but have some exceptions and activity called Standard view for iwork apps and some others.
Also, note that except for tables, you can't skip to the end of a group of element by using vo shift left/right like you can do in standard mode on the web, just so you know. I find this quite irritating although I rarely used it in standard anyway but it's one of these little undocumented nice things you are happy to know.
Hoep this helps!
I don't think I was very clear so don't hesitate to ask follow up questions, and other mac users to elaborate on this if necessary.

For your second question, I have an m2pro myself and tried an m2 air, I don't hink that at this small level there will be a significant performaance difference, as far as I know. Something for sure and I can definitely confirm that, but nothing that major. This is a problem in voiceover itself and why the snr is especially infuriating.

Another tip from a recent (slightly happier) windows user, with sequoia you wanna do like me and set vo-e/shift-e for next/previous text field, it's very very useful and now we don't have this mental berdun to get used to the various commanders while this could have been an option earlier. That way you don't have to get use to have a whole other set of keys with another modifier.
As a bonus, litterally the first thing I did on the first beta of sequoia was to go in these new command settings, and set vo-shift-b to battery level, vo-f12 for time, vo-cmd-f12 for wifi status and vo-shift-f12 for date. :)

i've prettymuch always used ignore groops, except in logic/other complex apps, think loopback, pages, izotope rx, etc

Maybe if you can tell what willbe your main and most important use case for this new computer we might help even better for mac VO specifics here?

VoiceOver's verbosity can be tamed in the Verbosity category of VoiceOver Utility. There are a few tabs right at the top of that category that are easy to miss if you're not expecting them. Personally I always set speech verbosity to Low, and enable tones or pitch changes instead of spoken descriptions wherever possible. A couple of years ago, Apple introduced a bug in some of the speech synthesizers that prevented speech changes so I also had to make it convey capital letters and deleted text by producing tones, but fortunately they fixed it in Sequoia so I reverted back to my preferred setup.

Using the trackpad to navigate with VoiceOver on MacOS is generally considered suboptimal, especially for websites, since you have options like QuickNav and Single Key QuickNav that make the navigation experience a lot simpler and closer to NVDA's Browse Mode. In Sequoia they have also added the ability to remap pretty much all VoiceOver controls, so you can customize everything to your taste, and with the full-size Apple keyboards with 19 function keys, dedicated text edition keys, and a keypad, you have plenty of customization options at your disposal to make navigation as efficient and comfortable as possible.

Finally, while you didn't task for this, I suggest actually trying virtualized Windows on your M1 Mac before investing in a new Mac, because interacting with the screen-reader in the Windows guest might be problematic, not to mention potential accessibility issues with virtualization software. I run virtual MacOS instances for development and testing on my systems using UTM, which is accessible, however the only option available to virtualize Windows in UTM is through Qemu, which is a hardware emulator, so while it can run Windows, either by emulating an x86 or ARM PC, you can expect an extremely low guest performance even on an M4 Mac. UTM also leverages Apple's Virtualization framework to run native guests, however the Virtualization framework only supports Linux and MacOS guests so it's not an option for Windows. VMWare might be accessible and be able to run Windows natively on ARM-based Macs, but I have never tried it myself.

Yeah vmware is quite decent for windows. I ditched utm cause it has too many random problems, like permanent sound cut from the vm (nvda or windows sounds) until reboot of the vm which made it completely unreliable for real work.

I use the trackpad a lot to go from headings to links and other navigation things, and when I swipe around on the screen, like swiping left or right, or navigating by headings, I don't notice it having a slower response time at all. I disabled sounds a while ago, so that might be it. I dont remember what it was like before I had sounds on.

As far as I know, there isn't a way to disable or change the verbosity of specific HTML elements like landmarks. This is one thing I really enjoy about NVDA's Document Formatting settings. I encourage you and anyone else to report this to Apple Accessibility. It doesn't seem like it would be very difficult to implement into the existing custom verbosity framework. I'm going to report this right now.

I don't disable grouping behavior. I like how VoiceOver works by default, and there are strange issues that pop up when you change the setting which has already been discussed.
As for disabling sounds, I believe it does make the experience more responsive, but I like the ability to play tones for different actions such as navigating to links, so I'll keep it enabled. I thought the sounds were supposed to play at the same time as the speech, but this doesn't seem to work all the time. I'd love a fully customizable set of sound cues for all kinds of actions. Sadly, this is only partially implemented with certain elements and actions as I said.

If you can copy and paste the html element into the "from" field and then press the spacebar once in the "to" field in the pronunciation dictionary, I'm willing to bet VoiceOver will treat/speak the html element just like it treatsspeaks a space in a text field. Has anyone tried this? I use the pronunciation dictionary quite a bit for changing words and phrases into silence or vastly shortened abbreviations.

Based on the suggestion about grouping behavior, I set up an activity to make Firefox ignore groups. It still sometimes says group, e.g. for comments here. Now I need to figure out how to get it to stop saying things like definition and term, see first message here.

https://www.thegearpage.net/board/index.php?threads/assinging-midi-controls-in-garageband.1709222/

Michael,
For a snappier response in Mac VO, I;
Load the VO Utility.
go to the Visuals category,
Click the Panels and Menus tab.
The first item is, Show caption panel,
and uncheck the checkbox.

This always makes my Mac VO usage a bit snappier. Of course, I cannot see the screen at all anyway and have no need for a panel that shows everything that VO speaks.

Also, to change the Verbosity level on the fly, Press Control-Option (VO) and "v", continue holding the VO keys and use right/left arrows to move through the Voice options. Use up/down arrows to adjust each as you see fit. Press Esc key to put away the "on the fly" Voice Options adjuster.

Reducing the Verbosity Level to Medium or Low speeds up your own usage since in some cases less is more.
. Likewise, 'Raising' the level can give more info about web elements and construction types, when needed.
I have reported things including all the info VO can give, to webmasters before, where specific details were important.
Maybe this helps?
Nicholas

You may want to try changing the setting in VoiceOver Utility under the navigation section of the web category to group items instead of navigating the DOM on webpages. This gets rid of the start of and end of messages, but forces you to interact with a lot more which I find rather cumbersome. I still recommend everyone send a suggestion to Apple for a setting that disables the messages when you enter or exit groups such as lists, landmarks, articles, etc without disabling interaction completely.

100% agree with this suggestion. When we have one of the 2 announce/ignore groups we lose the vo shift left/right arrow key commands to go at the beginning or end of a container like list, except a little bit for tables. Otherwise we just get the interactable sound but it's actually flat? Those who understand understand :)