George Oreo is still around. And he is French.

[WarmWombat]

I was hoping that the English voices were going to be redone for the main release on 4th August, but it remains unchanged. I still don't know who 'George Oreo' is. I am also assuming that 'Artilay' is French for 'Artillery'? IL2 Korea is one of the first sims in a long time that really draws you into the simulated world of 1950 around you, until the radio comes alive...

[DetCord]

5 hours ago, WarmWombat said:

I was hoping that the English voices were going to be redone for the main release on 4th August, but it remains unchanged. I still don't know who 'George Oreo' is. I am also assuming that 'Artilay' is French for 'Artillery'? IL2 Korea is one of the first sims in a long time that really draws you into the simulated world of 1950 around you, until the radio comes alive...

On the other end, the Russian comms traffic is excellent. Hopefully we'll get proper English V/O's down the road.

[TheNappingMan]

And why is Red "3" pronounced Red "Berry"?

[Valis]

1 minute ago, TheNappingMan said:

And why is Red "3" pronounced Red "Berry"?

I always thought it was being said 'theree' - i.e. someone had slipped a rogue 'e' into 'three' during the AI training.

 

[TheNappingMan]

Oh, good point.  🤷‍♂️

[FuriousMeow]

Aren't the UN comms South Koreans speaking English? That's my impression based on the accents.

Edited August 5 by FuriousMeow

[Banzai]

Some of the English voice lines sound like seriosly depressed individuals are speaking them. I am really worried for my wingmen when I hear them. Maybe that's why they crash into each other so much in my Mustang career. Maybe they are fed up with war. 😉

[actrade]

The accents are really bad to the point where it is sometimes hard to understand what they're saying. Also, when the AI says the number 3, I swear they are saying "berry".

Are these just placeholders or will this be improved in the future. Thanks. 

[DetCord]

On 8/5/2026 at 9:42 PM, FuriousMeow said:

Aren't the UN comms South Koreans speaking English? That's my impression based on the accents.

They used AI/text to speech to create the comms traffic so they could do it in a wide array of languages, eleven specifically. While it is true that they used real voice actors for the TtS system, the English comm chatter is jarring and overtly difficult to listen to. 

[FuriousMeow]

5 hours ago, DetCord said:

They used AI/text to speech to create the comms traffic so they could do it in a wide array of languages, eleven specifically. While it is true that they used real voice actors for the TtS system, the English comm chatter is jarring and overtly difficult to listen to. 

I understand, but I think that English comms is a "Korean" speaking English and not someone more "Western." There's a definite non-"Western" accent to those comms.

[Bumfluff]

UN controllers would have been predominantly American 

[DetCord]

15 hours ago, FuriousMeow said:

I understand, but I think that English comms is a "Korean" speaking English and not someone more "Western." There's a definite non-"Western" accent to those comms.

And that is likely due to them using Russian VA's for the comms traffic. Also, TtS typically sucks, even with VA's you'll have issues with the way the system attempts to enunciate, or rather fails to properly enunciate words and phrases.

14 hours ago, Bumfluff said:

UN controllers would have been predominantly American 

Yep. A small handful were British and Australian, but that's it.

[Aapje]

It would be interesting to know what the technology is that they are currently using for this and what the requirements are with regard to allow RAM and CPU use.

I got a bit into a rabbit hole and noticed that Pocket TTS seems to be a very good open source model that is still very efficient. You can try it out here yourself. I would suggest trying the Jean voice:

https://kyutai.org/blog/2026-01-13-pocket-tts/

It's under the MIT license which allows commercial use and many default voices seem to be under attribution licenses only. It's also possible to train your own voice, but not your own language.

The Kyutai people also plan to train more languages than the currently supported ones, although it doesn't yet have the languages that IL-2 Korea needs. But they started with just English in Januari and added 5 more languages in March already, so apparently they can add languages quite fast.

In general, this technology seems to still have quite a bit of room for improvement, for example, one person was able to half the memory usage and get a 25% increase in speed, with 'just this one trick.' I also saw people tuning the model to be higher quality.

The big upside of using a Text-To-Speech generator is that it is relatively easy to replace the system with something different, without having to alter any content. Given the amount of progress, I think that they shouldn't put too much effort into improving it now and then later make a big improvement.

 

Edited August 13 by Aapje

[FuriousMeow]

I just heard, perhaps the most egregious of, the comms text to audio "AI" blurt out... "spotted 'triple uh'" 

Hilarious, not triple A but "triple uh"

The "dot" instead of "point" when it calls out range irks me but the "triple uh" actually had my chortle.