Gonna preface all this by saying I am by no means an expert in Unreal Engine and I'm still very much learning, so if I'm wrong about some of this stuff on the technical level - I'm wrong. But I'm going to reply to the best of my ability about the tech *as I understand it*.
It saves time if you're using their nanite system, which is bleeding edge tech that is completely untested on consoles and basically requires an SSD for PC. It allows you to import assets directly from ZBrush (or other 3D sculpting program), and that's if you're creating every asset from scratch, which they're not necessarily doing. Furthermore, that means that a much higher standard is being set for the assets and considering the "stylized" look of many of the NPCs in S4, that means that most of those will need to be thrown out. Also, nanite does not work with deformed meshes, so now you're talking about having two separate pipelines, legacy for the character models (which still need to be LODed which Unreal has tools to automate anyway), and high quality for all the static meshes. That's a ton of work. UE5's high end features are for ultra budget games, developers who can figure out a clever way to use nanite for everything (ie: games about robots or other non skeletal meshes), or film production.
UE5 isn't production ready yet so neither of us can say *definitively* that Nanite runs totally fine on consoles with no drawbacks, hiccups, or tweaks; but both the engine reveal and recent livestream had footage of demos utilising Nanite running on PS5 and Xbox Series X. There is no way the engine is shipping if one of its flagship features doesn't work on consoles, and I'm not sure why SSDs are particularly relevant when both new consoles come with one by default...?
Not sure why you're going on about stylisation here either tbh. The pros of using Nanite are removing the need for authored LODs and instead using multi-million triangle high quality art assets at runtime without a major performance cost. Shenmue III's character art wasn't low poly nor was it intended to look so (Ryo's Shenmue III model is 100,000+ triangles), so I don't see where the concern is regarding an artistic clash? It would be no greater difference than much of the environmental art in Shenmue III, except maybe looking much nicer closer up and rigid meshes (which make up quite a lot of practically any game world) having a far lower performance impact than they would otherwise.
Nanite doesn't work on deformed meshes *yet, as per the documentation stating that rigid meshes were their priority in development because of the overall prevalence of them in projects that would use UE5. They may have a solution for deformed meshes in place by the time the engine is production ready, or they may not, but it can still be used for rigid meshes all the same.
Again, not sure why you're talking about two pipelines here. Whether they stick with UE4, or they move to UE5 and it doesn't support deformed meshes by the time the engine launches proper, YsNet would still have to do that "ton of work" on characters that require LODs regardless, but at least with UE5 the environment team could have some of their workload reduced using Nanite for rigid meshes. UE4 does have automatic LOD authoring tools but they're not magic and do require someone to actually manually adjust the distances at which LODs load in based on what the player may or may not be able to see, and then do that for *every* static mesh. Nanite is pretty much just "set it and forget it".
I think you might be under the impression that Nanite is all or nothing in UE5? It's actually not, and you can use a mix of objects that do or do not utilise Nanite in a project. The Valley of the Ancient project just released is a mix of Nanite and non-Nanite objects.
Here's what that project looks like when looking at the Nanite triangles in the environment (the coloured splotches) and non-Nanite objects (the black masked out parts; the player character, sky sphere, weeds, etc) via the viewport:
No the real time saver will be using everything that's been set up from S3 to create content (cutscenes, dialogue, scripting etc.) and not having to port those tools over. I've ported projects to different engine versions of UE4 and had things break.
I'm not sure if YsNet made any custom tools in the development of Shenmue III, but things like cinematics are handled in-engine through sequencer, etc. I'm pretty sure practically all those things you've listed were handled in-engine with the toolset UE provides by default, but don't know for definite.
Things do sometimes break from version to version in UE4 if you are porting your project forward and are using a beta plugin that got updated to a 1.0 version, or they've depreciated some tech, or they're requiring a tickbox to turn something on now, or some values got changed, or a node works differently... but you're still broadly working with the same framework you started with. It's not like porting from one game engine to another and it is generally beneficial to keep on top of the latest developments with the engine.
Shenmue III ported forward to newer versions of UE4 several times during development before they decided to lock it off for the version they were going to ship on, which from memory was 4.20 which released in mid-2018, so the game had been in development for approximately three years before they decided to stop updating the engine.
There are countless reasons. A game on a budget as tight as S4 with as much content as it needs will need to reuse as much from S3 as possible and this mentality of pursuing the latest shiny new thing will no doubt lead to budget/time issues, delays, cut content and other horrible things. Particularly since this is a rabbit hole that Suzuki has shown himself the type to fall down.
Well we don't actually know what kind of budget Shenmue IV may have, if it's even been greenlit at all, but that's kind of irrelevant to moving development over to UE5 once it's production ready.
It's not chasing a "nice new shiny thing", UE5 is very much an extension on UE4. UE4.27 will release later this year as a step forward from UE4.26, then UE5 will get a production ready release earlyish next year as a step forward from that. Even stuff Epic plan to depreciate out of UE5 will still be in the first few versions of the engine, if for whatever reason YsNet are relying on something in UE4 that has been replaced in UE5.
Improvements to UE aren't necessarily about giving developers more and more tech to create cutting edge visuals with, a lot of additions are about making it quicker and easier to make games as well as giving small teams the kind of production quality they wouldn't be able to achieve without the engine. Shenmue III is a perfect example of this, as there's no way YsNet would have been able to stretch what resources they had as far as they did without an engine like UE4 - and even then it has developed considerably since the game came out with things such as the control rig (to enable rigging and animating in-engine and saving time on roundtripping), and the sky, cloud, and atmosphere system for creating real-time "time of day" lighting changes on the fly as seen here:
Even outside of Nanite and Lumen, features in UE5 will allow developers of all sizes to take a step forward as plugins exit beta, and features added in the last few versions of UE4 mature further.
Off the top of my head:
- New full body IK solver
- Data layers (for making it easier to include persistent changes to the game world or day/night cycle map changes)
- World partition system & hierarchical level of detail (for better loading and unloading of the environment)
- Motion warping system (to warp a character's root motion to align with targets instead of creating bespoke animation, huge implications for dynamic gameplay instances like combat if used)
- Modular gameplay and gameplay abilities systems (hugely beneficial to a game like Shenmue developed out of several individualised pieces that form a whole)
- Enhanced input (making it easier to have control methods switch on the fly, or enabling context sensitive presses)
- MetaSounds (more control for audio designers in-game)
None of these have very much to do with graphics and are all new or maturing features for UE5 that would benefit YsNet.
Prototyping what exactly? You run around talking to people, picking up objects and getting into fights. They made Shenmue 3; how different do you think S4 is going to be?
If YsNet are going to add anything to Shenmue IV, or improve upon what they've made, or make alterations to the core gameplay loop, it'll be prototyped. All games get prototyped, even sequels.
Want Ryo to be able to step over a fence in Shenmue IV? Then it's got to be prototyped as the developers ask themselves questions like "how high are the fences and are they all going to be this height?", "how are we going to animate this?", "is this activated with a button press or does Ryo automatically step over the fence if the player is pressing in that direction on the thumb stick?" and find the answers by building test cases.