AI has existed in various forms a lot longer than 18 months. It’s been far more heavily marketed in the last 18 months, however AI has existed in some basic form since spell checkers.
Text to speech has been able to dynamically, and usually correctly say people’s names since it was introduced in Windows 2000, hardly a recent development, nor is a massive stretch to say that games like Forza used this 23 year old api to read my name to me as part of it’s announcements.
Windows 2000 is also hardly the earliest example of this, Apple had this technology in the 90s, and there are DOS applications that can read text from a file aloud from the mid 80s, again with largely correct pronunciation.
Do not confuse Chat GPT with the generic idea of AI, there are examples on wikipedia of AI from as far back as 1956.
AI has been in existence for quite some time now, but it is a broad term that encapsulates numerous technologies and countless applications. However, the technology we are discussing today is not yet common or widely used, and therefore, it may not be worth the effort to dive into it unnecessarily.
Regarding your previous examples, it is essential to note that text-to-speech and Fortza use different back-end technologies, which makes a significant difference in the level of quality they offer. While they may have similar outcomes, the difference between them is enormous, and it directly affects the quality of the results.
All you did was provide a false equivalency. Generative AI applications have only been around for about 18 months in an affordable state for both consumers and B2B Clients.
AI would never be able to pronouce my name correctly and I think Im not the only one with a strange name. Being mispronounced all the time is way worse than champion (which is also cringe).
It’s important to understand that text-to-speech and the feature you’re referring to are not the same thing. They are two systems that are functionally different and cannot be compared under the same umbrella of “AI” that uses audio, the feature you’re talking about takes information from your Microsoft account, not the in-game files.
Implementing a voice-synthesised feature for unique character names would require a complete overhaul of the system, including training a voice synthesizer model for every VA. This is not a simple task and would take months of work, overhauls to entire databases, and legal concerns to consider. Why would you hire a VA again when you already have an algorithm trained to mimic their likeness?
Therefore, it’s not as easy as reading a file, and the feature you’re suggesting is not possible. It’s essential to understand the limitations of different AI systems and not to compare them under the same umbrella. Text to speech isn’t an issue but you aren’t going to get the same quality as what you are asking for.
PS text-to-speech is an early form of generative AI. You are generating audio from text and yes a machine that needs to learn how it sounds across multiple languages.
Then you accept that AI has not only existed for 18 months. Further, there is little programmatic difference between fetching a string “name” from a database online (like say the database containing your character), and fetching a string a game file or from any other location. You don’t need a full AI suite to read one word fetched from a database using a voice api.
Would it sound identical to the VAs? No, would it be better than either nothing or champion as we get now? Yes
I would also have liked to be called by my character name, but I am sure most names would not be pronounced correctly. I am glad text often use my character names at least.
I didn’t say it existed for 18 months. I said generative AI in its current form hadn’t existed in the market for 18 months, which is true. Stop trying to play semantics. It just shows you don’t know what you are talking about. Yes, there is a vast difference between accessing a completely different database on a separate platform and accessing a string in a script. It wouldn’t even work with that implementation because they would call you by your battle net tag and not your character name.
If you train a model on the voice lines of the characters, it will sound like the voice actors, and it’s a massive issue in voice acting right now, nor would it be on par with what you receive right now. There is no point in implementing this system as it’s a complete waste of resources when you don’t even have to mention the name because it would be grammatically correct either way.
Honestly, I get you want this feature but you just waste everyone’s time with your nonsense because instead of talking about the merits of this system you’re too busy playing a SQL programmer.
So if it’s such a vast difference as you suggest. What’s to stop the developer from doing it sensibly, ie using the character name from RAM where the game has already trivially fetched it the moment you logged in.
Modern generative ai is very complex however fetching a string the game already fetches for other purposes and feeding it to a voice api that already exists really isn’t.
Since blizzard already has routines for fetching characters details and storing them locally, as without them the game wouldn’t work. It seems trivial to reuse that same string. After all addons capture this data without issue all the time. Almost as if the blizzard api makes character names accessible to developers regardless of where they’re from.
I am not fighting for or against this feature, there is challenge in making the voices sound right, not all of them technical but certainly some of them are technical . Merely demonstrating that getting strings out of a database regardless of where that database is, is something that already happens in the game and is fairly trivial. Indeed it’s something I did many years ago in my first year of study.
I will store some strings in a database on my own webserver and write some code to have them read out. If it would help demonstrate the point.
This is what I’m talking about your too busy trying too poke holes in a made up strawman. You compared a text-to-speech algorithm, to a completely different one because they’re under a single term. Now your trying to find solutions to a made-up problem that has nothing to do with the benefits of the system you propose. Stop trying to play IT engineer you’re wasting everyone’s time.
In terms of Forzta you are not fetching a string, you are receiving data from a third-party platform SQL database and then having it placed into every quest text so you can then have a tailored made VAs voice seamlessly state what is a poorly written name or a workaround that would be pronounced different than what the user intended because the naming scope of WoW is limited due to a lack of secondary names. Not to mention the legal issues with using said VA’s voice in such a manner to begin with and how the company would try to use it in the future.
There’s nothing wrong with the concept there’s everything wrong with how the game is set up in order to even acheive it not just internally but externally with the perceived negative use of AI especially as it pertains to VAs. Implementing this would be a complete waste of time and a huge amount of unnecessary legal drama for what would be an NPC to state at a poorly written and pronounced name.
So instead of wasting a ton of resources and adding in a completely usless system just remove the voice line “Champion”. It’s saving people from having to write, say it and even think about it.
No it won’t be, unless you are absolutely devoid of artistic taste and think that character suddenly switching from proper voice acting to text to speech robo voice is okay.
It’s not, and it is not worth it just for the sake of getting rid of characters calling you champion
It seemed that at the same time you got Commander in GW2 you got it in WoW and SWTOR, then Outlander (reeee) but just being “Commander” or hell even “Boss” was so much better in GW2 I wish it stuck around.
It fetches your name from the account that’s used to make the purchase (typically the MS account you sign in with) it doesn’t ask anything about pronunciation.
My IRL name is pretty mainstream so it’s never got it wrong for me, of course other people might be less lucky.