arma2 ai voice implementation

[sandzibar 0]

Didnt see this mentioned in the requests/hopes list. Presumably this may get a bit of a tweak to. make. it. less.

robot.

ic. and disjointed.

But apart from that, Id really love for the opfor to go back to speaking in russian (just as the Red Hammer pack brought us in OFP I believe). So much more atmospheric.

[maddogx 11]

The AI voices are choppy because of the way they are implemented:

Pre-recorded words or phrases are simply grouped to form a sentence.

The advantage of this is low resource usage but the main disadvantage is, like you said, that the AI sound like robots, because they can't change the way they say words (inflection, whisper/talk/shout etc.).

Solution 1: Create a vastly increased number of pre-recorded words with different variants for each word, depending on the situation and position of the word in the sentence.

Pro - still fairly low resource usage.

Con - huge effort to create all different words and variants.

Solution 2: Implement a text to speech (TTS) engine that reads any given sentence in different styles. Depending on the quality of the TTS engine, this could make AI radio extremely immersive. BUT this would require a lot more performance.

Pro - ideally, robotic AI voice would be eliminated and replaced by natural sounding voice.

Con - Extreme resource usage, since each soldier would need its own instance of the TTS engine. Otherwise all soldiers would sound the same.

You see, this problem is not easily solved in any case. Just making the AI talk faster won't help a lot.

[NeMeSiS 11]

Solution 2: Implement a text to speech (TTS) engine that reads any given sentence in different styles. Depending on the quality of the TTS engine, this could make AI radio extremely immersive. BUT this would require a lot more performance.

Pro - ideally, robotic AI voice would be eliminated and replaced by natural sounding voice.

Con - Extreme resource usage, since each soldier would need its own instance of the TTS engine. Otherwise all soldiers would sound the same.

You see, this problem is not easily solved in any case. Just making the AI talk faster won't help a lot.

Have you ever heard those? In the best case they are as immersive as prerecorded sounds, in the worst case they cant even properly pronounce all the words.

[Lou Montana 33]

third solution : record vanilla sentence and make the sound engine apply (accordingly to the AI behaviour) some pitch or other sound modification to allow better ambience sentences.

Not CPU requesting, not HDD demanding

[maddogx 11]

Solution 2: Implement a text to speech (TTS) engine that reads any given sentence in different styles. Depending on the quality of the TTS engine, this could make AI radio extremely immersive. BUT this would require a lot more performance.

Pro - ideally, robotic AI voice would be eliminated and replaced by natural sounding voice.

Con - Extreme resource usage, since each soldier would need its own instance of the TTS engine. Otherwise all soldiers would sound the same.

You see, this problem is not easily solved in any case. Just making the AI talk faster won't help a lot.

Have you ever heard those? In the best case they are as immersive as prerecorded sounds, in the worst case they cant even properly pronounce all the words.

Yes, I have heard quite a few TTS engines and, believe it or not, there are indeed some quite good ones out there. Some even provide good inflection, syllable emphasis and even some forms of emotional conveyance (although admittedly, I have never heard a TTS engine that could shout or whisper).

[CarlGustaffa 3]

I kind of like todays implementation. But I'd like more voice "states". I.e. one state in combat mode (screaming loudly), one in stealth (whispering), and one for the rest (normal talk).

There's nothing like laying in ambush in the forest in stealth mode, awaiting enemy patrol, where one of the guys screams out "Enemy spotted two o'clock"... Kinda ruins the feeling of trying to be stealthy.

But most other games fail in this respect too. Walk in stealth mode, but upon entering water you produce VERY loud splashes

[bravo 6 0]

There's nothing like laying in ambush in the forest in stealth mode, awaiting enemy patrol, where one of the guys screams out "Enemy spotted two o'clock"... Kinda ruins the feeling of trying to be stealthy.

lol

would be nice if BIS could simulate different tone of voices for the different group behaviours.

Meaning it would be huge amount of data.

Edit: Is it possible for an AI to detect (a player or enemy AI sound) in determinated distance?

Using the example above would it be possible for an ai to change behaviour by hearing hostile sounds? Such as a reload

[paragraphic l 2]

It should be possible if not yet implemented.

I can remember that Rainbow Six 3 for Xbox had this feature, where playing multiplayer co-op resulted in a whisper fest so you where not spotted

I think there should also come an civilian/at ease voice state

But for making mission making easier I really think some sort of TTS should be an outcome

AND something I really LIKE would be different languages

[mr.g-c 6]

Ohh yes whisper voices would be great.... imagine a recon night-mission with a smal 3 Man squad....

This is something really needed, however i doubt there will implement anything like this as the list is full unfortunately....