I'm late to the party but I just figured out how interesting / powerful AI dialogue from ChatGPT can be. IMO, the technology is already there to make characters far more interesting than what the industry currently produces.
Are developers fast tracking this technology into upcoming games or is there something preventing this from being implemented? Single player gamers of NeoGAF, educate me.
Yes 100% already being done especially for side characters and, upon my looking into it as a reviewer who covers sound, apparently, a couple of games have come out with this already done for some side characters that so far no one has guessed.
Also, while I love the craft the sad fact is the bar isn't high for a large number of games for style or ability either and AI especially trained on world models can easily add authoritative audio to games that goes beyond what a stressed-out random writer can do for NPC dialogue even if it includes none weighted world suggestions, smaller language models or those even with low variance in their prompt results.
For example, the world-weighted examples I supplied to 20 devs/youtubers/streamers for a video coming out in a couple months, there was no difference in ratings between AI written and writer-written dialogue samples when it came to detection from professionals. They weren't massive paragraphs but many were well within the larger structures of NPC dialogue during quests and quest giving.
To give you a nice delta. MUSIC created by AI was over 90% detected. So that shows you how far one has come and the other has to go.
Voice is even more interesting. Even when working on this with audio leads or with composers, discussion of this I mean, one of the major issues is that PEOPLE are varied. Many sound stilted, robotic, actual real people can have limitless sounding voices and speech patterns. We have seen this in everything from shows to movies to YouTube videos to presenters. Compressions and audio touch-ups raise the bar of lower audio and can lower the bar of better audio as we have seen in games that tried to save space due to having a ton of voice work and wanted to shrink disk size(Valhalla's audio issue with its initial recordings as well as the touch-ups on Legion).
That also does not include stuff like Eleven Labs and others are working on that is already passed what is out now and coming soon.
Lastly, there is a very interesting situation that occurs where expert bias begins and people "THINK" they can tell. However, we are currently testing that in a pretty unique way and while the percent of detection will be higher for sure I have already collected samples of over 30 AI spoken samples, and quests, that got by professionals including a couple voice actors. The crazier part is they were "sure" as in higher than a 95% certainty that the voices were real.