I understand that it is difficult to have voice actors record all the different things that ATC may say, but using AI voices for the passengers in career mode flights just feels cheap and out of place in a game that puts so much focus on realism.
Also the pronunciation of German place names is bad to terrible. Can’t comment on the rest of the non-english speaking world, but it really drags you out of the game when places are pronounced incorrectly, especially by “passengers” that would know the area.
7 Likes
AI voices are not perfect, but on the other side it give MS a lot of opportunities to add new texts, to change phrases and etc and etc. I think, this is overall right decision. MSFS has AI-generated world which is also not perfect. But it’s all about possibilities, that are much wider with AI. This is the future. And I think it would be better to improve AI speech, then to exploit old technologies. Just IMHO)
5 Likes
There are over 10,537 lines of dialog in the career alone, and there are many variables in those scripts (591, to be exact) which would also mean any variable would have to have all possible values voice recorded, or separate recordings with each value included which could explode the number of total unique lines dramatically. While I agree the current scripts are really bad (the AI voices aren’t even the main issue, the scripts themselves are just poorly written, full of grammatical errors and just goofy at times, not things real people would never say) it’s definitely cheaper than voice acting.
Some of the Azure voice models are better than others. There’s one in particular that is so significantly better than others that it kind of surprised me, was during a Certification mission, but I forget which one. But it was extremely natural, like shockingly good compared to every other voice. They need more like whatever that one was.
But yes, they also need to either A) Have AI help rewrite the entire script because it’s really not good, or B) Have AI generate unique dialog on the fly (my preference, but I’m sure they’d worry it would go rogue and start promoting violence or something 
I am a voice actor, so part of me is upset they didn’t hire actors, it would have gone a LONG way toward improving immersion.
But the AI approach is to make sure dialogue specific to your flight can be used, and there are just too many variables to properly record live voices for such a wide open script.
1 Like
Not to mention, they’d ideally need to record voice lines in many languages! Quite an undertaking. I don’t blame them for the AI. That said, do the AI voices only ever read English or if I set my sim to German (or my OS, maybe?) would it read them in German? I know there are tons of locPak files with different languages so I assume it should work.
1 Like
Sure, they work in German. And they are hilariously bad.
2 Likes
I you want to have your home town hand crafted and a couple of dozen professionally acted characters, you can have that.
You want to do the same with the whole world? Then this is the only practical way (sadly).
I also hope it doesn’t stop hand crafted and professionally acted roles where there is still a need.
There are some obvious times when 2024 would have benefited from voice acting, and others where AI helps.
My main issue is the AI voices MS/Asobo went with are hilariously bad at intonation, and so much of the script is written like the characters are having an emotional response. “Wow! That is BEAUTIFUL!!!” instead sounds like, “wow that is beautiful I am jennifer, jennifer’s sister. CAPTAIN!”
I have definitely heard better AI voicing algorithms for following the intended musicality of a manuscript.
2 Likes
They could do it with red dead redemption (although it took them a few years) I’m sure Microsoft can do it if they wanted.
The current career AI voices sound so bland and uninspired.
There are better AI voices available but I guess Microsoft has to use their MS Sam voices from the nineties.
Intonation is hilariously wild. Every single time end a sentence with “Captain” or someone’s name, it reads as if there’s no pause. “Did you enjoy, Micheal?” is “Did you enjoy Michael?” and I’m like “In what way was I meant to enjoy him??”
But I guess in the words of the First Flight dispatcher… “No so easy to impress the locals.”
3 Likes
RDR2 had a branching script, not a dynamic script, so it was possible to record all dialogue paths. In 2024 that is just not possible.
Also, there’s AAA game budgets and then there is Rockstar’s unlimited budget and timelines.
2 Likes