/jp/ - Otaku Culture

>>47019641
Most of those hundreds of roles aren't all that useful. They're also writing their own engine, but I'm not expecting an internet collab to do that. 3-5 competent individuals is enough, but even those 3-5 is really hard to gather and keep together. At least in the early stages of pre-production/prototyping, IMO there should be clear division of labor with little to no reason to delegate anything. In the context of "an open world action rpg touhou game" would the minimum roles need to be filled are:

>The systems director
They need to design, implement, integrate, and test all the major gameplay systems like movement, combat, interaction, quest log, dialogue, stats, items, inventory, etc. They should also define any level design standards with a few example greybox levels.

>The art director
This is a tech-artist who can standardize and prototype the character models, rigs, materials, and animations. They should also prototype some VFX.

>The narrative director
This is someone who details out the worldbuilding and all the major plotlines and produces an initial draft of the dialogue. Ideally this is also someone who can kitbash the world map together in engine.

>>47019736
The most practical use of AI right now is to generate pro-AI propaganda to lure in more investment money. There are a few limited practical applications, but they're all janky and unreliable, and are still far from being a viable replacement for competent artists. At most, they can be used by a competent artist to accelerate their workflow. But, don't take mine or any internet grifter's word for it. You should test it out for yourself instead of guessing. The most practical applications I've found so far are:
>Voice changer
https://github.com/w-okada/voice-changer
https://github.com/Mangio621/Mangio-RVC-Fork
>Animation
https://cascadeur.com/
>Texturing
https://stableprojectorz.com

>Some other technology that have limited applications outside of this project:
Character image generators can appear to be really good, but are extremely janky and mostly a waste of time. You need to give the model a hint of the desired pose/gesture as a rendered 3D scene if you want any semblance of reliability (see picrel). The tracing idea you came up with is also fairly pointless because skilled artists have always been able to just trace 3D models. You can use this tech to grift with porn CG packs, or graphics for something like a card game, which can use arbitrarily stylized and inconsistent static images. I've yet to see anyone make any game sprites that are of higher quality than what you can just buy out of asset stores.
https://civitai.com/
https://tensor.art/

Story generation is riddled with nonsense and is of lower quality than impromptu ERP, but you can maybe roll for inspiration with it if you're feeling blocked.
https://novelai.net/