[ home / rules / faq ] [ overboard / sfw / alt ] [ leftypol / siberia / edu / hobby / tech / games / anime / music / draw / AKM ] [ meta / roulette ] [ cytube / wiki / git ] [ GET / ref / marx / booru / zine ]

/games/ - Games

Name
Options
Subject
Comment
Flag
File
Embed
Password (For file deletion.)

Join our Matrix Chat <=> IRC: #leftypol on Rizon


 No.29039

AI will be good for games, actually.

It will allow games like Morrowind to have fully voiced characters without needing to have either the hours put into voicing or the many audio files. Voice work will no longer consist of doing hours and hours of reading lines. It will instead consist of creating a distinct voice characterization with the most important lines and for the AI to read the rest (including procedurally generated ones). That is good news for voice actors because it will make it easier to voice characters, especially the more demanding ones. The actor will only need to read a small percentage of lines where the character screams and shouts for example, and the AI voice model will be able to use that for a wide variety of lines in the game. Since voiced characters has become standard (and creates bottlenecks with games like Skyrim), meaning that with similar amounts of effort from actors, there can be much more content in the game. If you combine that with chatbot-like systems to flesh out the dialogue, it means the writers can create more content with the same effort. It will allow much bigger and richer games with a much more efficient use of people's labor.

The same will also apply to all other assets once dynamic generation is good enough. There will be better dynamic models and textures, lighting systems, etc.

 No.29040

If your primary concern is quantity over quality then sure, I suppose. Text to speech was already kinda there anyway.

 No.29041

>>29040
It means you can focus your attention on the parts you want to be quality. The less important parts, like incidental and filler dialogue, doesn't need to be carefully crafted. It's more relevant for that stuff to be unique, so lower quality improvised lines work better than a small number of stock phrases. It doesn't matter how high the quality is if every guard you meet says "I used to be an adventurer like you, then I took an arrow in the knee."

 No.29042

>>29040 (me) like It feels like you're just turning so many hours of VA work into that many sevenfold playtesting to see if the AI fucked up the line, when good old bidibidibidi, animalese or espeak can do just fine.

 No.29043

>>29042
>bidibidibidi, animalese or espeak can do just fine.
It's not at all the same as actually reading the line, especially for accessibility purposes.
>that many sevenfold playtesting to see if the AI fucked up the line
People are going to be playtesting those parts anyway, including to ensure the dialogue system works. Just another thing to make note of and doesn't require special testing since that system is relatively isolated from the rest of the game. The AI fucking up a line is a minor bug at worst, too. Where you'd have problems with procedural AI generation is things that affect other game mechanics, like dynamic level design making sense or meshes matching hitboxes.

 No.29047

>>29042
AI tts is already doing a much better job than basic tts, try using one. Plus VA requires many hours of paying money, recording, editing, fitting it in, and repeating this a couple of times.

 No.29050

>>29047
Lol no plenty of vas get paid barely anything

What matters is the quality of voice acting and if it even works to communicate the emotions wanted by the game developer or studio, something that requires so much to time and planning that you might as well hire an ordinary va anyways to do due to the consistency and diversity of results possible


Maybe ai in this recieves enough funding that it's valorized to the point where it can be conventionally used along with real actors but for now it's gonna remain a, although very powerful tool for modders

 No.29053

>>29050
>you might as well hire an ordinary va anyways to do due to the consistency and diversity of results possible
Like I said you would still want actual actors to build the model for a character and to do the most important lines, but you could extend the character beyond what is reasonable to have an actor perform. Especially for the parts that matter the least and that you'd want to avoid repeating. Any amount of audio you record is finite, but if you are having the AI do the voice in real time can be reading newly generated lines. You could record 10 greetings and 10 farewells and then have the AI improvise 100 more if the player keeps talking to the same character so you don't keep hearing repeats. You could even have it adjust the lines when they have to be repeated so it doesn't sound too samey.

>Maybe ai in this recieves enough funding that it's valorized to the point where it can be conventionally used along with real actors but for now it's gonna remain a, although very powerful tool for modders

Considering that people accept solutions that don't really sound like speech at all, extremely compressed audio, or a mix of voiced and non-voiced lines it's probably a pretty low threshold to acceptance. If anything bigger games would have more use for it, to create more content on less budget. Games like GTA for example are extremely expensive, and you could save a lot of time and money if you can flesh out the dialogue (especially the random, least important parts) you can get a lot more bang for your buck.

 No.29077

>>29050
Didn't you hear that it is possible to set different intonations for different parts of sentences now?


Unique IPs: 5

[Return][Go to top] [Catalog] | [Home][Post a Reply]
Delete Post [ ]
[ home / rules / faq ] [ overboard / sfw / alt ] [ leftypol / siberia / edu / hobby / tech / games / anime / music / draw / AKM ] [ meta / roulette ] [ cytube / wiki / git ] [ GET / ref / marx / booru / zine ]