I have speech to text working that way on my mac. Some days, I just can't get my typing game on; typing is a struggle for me at best. Bogdanw's tips are on point. Also look for sites dedicated to ADA Section 508 compliance for tips, tricks and apps.
Probably wouldn't want modern Dragon anyway, Mac or Windows, unless you have a critical dependency and enterprise support. It's very expensive now @ $700 a seat, and their operational model is enterprise-only. Partly, Nuance must have tired of dealing with all the unprofitable peasants, who are always revolting. Mostly, speech recognition and to-text implementation was matched or exceeded in the OS, Mac and Windows (Never tried this in Linux).
MacOS Speech-to-Text tools have a learning curve, but the implementation is pretty good. Frustratingly, Apple split functionality among a couple Accessibility modules, probably to avoid law suits. These days, there's no telling how much intellectual property is original, licensed and stolen; they ALL invent, license and steal.
[EDIT: Oops! This is actually conditional = [SIRI is NOT required for the Speech to Text cycle. Despite how useful it can be, if you're going hard at Speech-to-Text, consider disabling Siri entirely. ] Sonoma does it different. You can only keep on-device using Apple Silicon. If you're stuck with intel it WILL transmit to Apple servers to process. MacOS is 100% capable of collecting and transmitting all that spoken business to Apple's servers. Yeah, yeah, you can click "Don't help Siri" or "Don't Listen for 'Hey, Siri'", but there are switches and prompts embedded throughout MacOS to trick you into unintentionally re-enabling Siri. There is no reason to trust Apple any more than Amazon, Google or Samsung. Furthermore, iPhones and Apple Watches run Siri independently of the Mac where you dictate your manifesto. To say nothing of "Smart Appliances" and IoT devices that have microphones for no apparent reason. Except I just said something.
