Getting you to an estimate, is very, very difficult, because there are so many variables in your questions, that you’d need to answer for yourself first. 
Wake word detection is not yet ready. Paulus made a statement regarding this in the last release video (2023.09.0), that the developers try to get that ready within the year of voice, but not anytime soon. He also stated, that the devices they want to use for this are sold out everywhere (ESP S3). 
So I wouldn’t keep my hopes up for an eta next month.
That brings us to your question, what do you need. That’s hard to say, because you have so many things to decide upfront. As an alternative, I’d like to give you some details about my setup, that might give you some insight on what you want for yourself.
- I’m using the very nice and nifty m5Stack ATOM Echo eg. on my workbench. There I only need voice assist, but nothing else like playing music or any other thing like a temp sensor.
Easy to install and very small, costs around 10 to 15 Dollar or Euro in the northern hemisphere. It’s not bigger than a 2 euro coin.
=> Ideal device for only voice assist. Giving commands and getting a feedback. Get read an audiobook or music is a no-no!
- I got an ESPMuse lend from a friend. That is a cool toy. You’d use these, if you want to have some nice speaker, that does real good sound (I’m no sound expert, for me, knowing "MC"s from my childhood, mp3 is “very high quality”
), this is for you. 
=> Ideal device if you want mainly to listen to music with this device and want the added benefit of having Assist there, too.
- Rapsberry Pis up from Modell 2 have an audio jack. So you can run nearly every speaker you have or can get for a small buck. Squeezelite (player on the Pi) and LMS (LogitechMediaServer, available as Add-on for HA!) to control the music.
=> Ideal, if you have lying some old speakers around, or some old Pis, that need a new playing field. I can recommend the old “PS3 Eye Camera” as microphone. Great 4-mic-array and you have the added benefit, if you have enough power on your Pi, you can even run video feeds of of it.
- ESP devices can channel music
HA->Wifi->ESP->BT->speaker, but that is not fully developed and seem to be not running reliable at the moment. I don’t have this in my setup, so only can say what I read about.
-
Rhasspy satellites are a nice alternative to run on a Pi. Rhasspy already supports wake word detection and can, in version 3beta, work with HA via Assist (as opposed to
events).
=> at the moment it is more a testing field, but it shows some of the concepts, that will find their way into HA eventually, I mean the developer of Rhasspy is the developer of HA voice assistant… 
- Old phones or tablets are very good for Assist. Install the companion app and you can even run it without https in a local network (opposed to a browser on the phone, who needs https). Just great, as everything is already there: mic, mostly good speakers and a touch display.
=> Ideal to get a second or third use for tablets and phones. I’m running four tablets at the moment, works great. I even switched my HA handling to a 10’’ tablet. For configuring I still use my laptop, but for the daily stuff.
- Yeah, an old landline phone works quite well. I don’t know how these adapters are called, but you can insert your line from the phone and an ehernet cable and you are good to go. If you take a look at the HA blog, there is even a guide to revieve your old walkie-talkie to use with Assist.
And there are so many more alternatives, it would really need some examples from you, where and how you’d want to use it.
But now to the downside: Google or Amazon devices are closed source applications, and therefor extremly hard to hack. They both did a pretty good job holding DIY people out. I wouldn’t bother with these devices and sell them on eBay. If you want to use Google as your source for further questions, you can already do so I heard, but I can’t tell you how that works. I can tell you, that it works very intuitive with OpenAI. You can set that and OpenAI gives you a really nice feedback. For example in the morning, you can let OpenAI generate a daily briefing for you, even with news headlines.
I don’t know, what you have Google Assist do for you, but I’m quite sure it won’t take long, that HA Assist can replace it. I for one haven’t used Google Home before (neither Alexa), but HA Assist can do everything I need. With the companion app even from outside (aka sitting with friends and asking HA what the weather forecast for tomorrow is works).
Hope this helps a little further!
But that is not the “normal” release cycle. The wake word will be the next step, but nobody knows when. 
