[Discussion] What we wish Unity to do after Unreal 5 tech demo hype

TextMesh Pro and ProBuilder still feel like 3rd party tools years after their acquisition. They haven't seen the level of integration or ongoing development that a 1st party tool should ideally have.

TextMesh Pro, for instance, only has a single page of documentation in the Unity Manual so far as I can tell. That manual page includes links to a 3rd party website for forum support, tutorials and further documentation. From that page it looks like even the new 2.x version of TMP isn't integrated with the built-in text mesh, it's still it's own separate component. To me all of that still very much feels like it's being kept at arms length from the core of Unity, rather than properly embraced.

ProBuilder isn't even as well integrated as that. It's companion tools also went free but, last I checked, aren't available via the Package Manager. So if you want to use ProBuilder along with ProGrids, which is how it's designed to be used, you need to cludge Package Manager and Asset Store stuff together yourself. Why so inconsistent?

I was really happy when Unity acquired both of those assets (I'm not familiar with Bolt and didn't know about that one) because I was hoping for better integration. But all that really changes, since I'd purchased them myself anyway, is that they're now in the Package Manager rather than the Asset Store.

For what it's worth, if the intent was just to make commonly used tools available to everyone and more standardised then I genuinely appreciate that in and of itself. If that's the case then cool, but also clear messaging to that effect would have set appropriate expectations. Right now this is one of the things that makes Unity look disorganized, where it could have been "cool, free useful stuff!"

 

Can you explain what you mean about a variety of controllers? Is this just different gamepads or does that include other types of controllers (e.g. joysticks, wheels, etc.)?

 

For me it's different gamepads.

 

The biggest thing I wish for Unity staff is not to overreact and do some stupid things. Epic forces their rules of the game by doing almost nothing (their demo looks like any other current gen AAA project, to be honest, and we know neither limitations of the new systems they use, nor how reliant they are on specific PS5 hardware), and it is alarming how easily people are manipulated.
A more constructive suggestion: add Vector Displacement Mapping support as a default Unity feature.

This is tried and true technology, and in essence does the similar things to this new LOD system (quality similar to high poly sculpt up close, and adaptive detail levels when paired with DX11+ Dynamic Tesselation). Still requires UV maps, but allows sculpting decorative props from nothing but a cube or a sphere as a base, so it isnt a huge deal.

 

With UIElements on the way, does TextMeshPro even have a future? Its tight integration with UGUI and superiority over Text is 90% of the reason to use it.

 

I imagine TextMeshPro is still used under the hood in UIElements.

 

Yes. Quote below from an official staff member.

https://forum.unity.com/threads/will-we-have-text-mesh-pro-text-uielement.756314/#post-5038766

 

I would like Unity has small number of hotfixes like Unreal.
For example, 19.3LTS or 2020.1 could only have 3 patches. It makes version control much easier to handle.

 

Have you tried enabling preview packages?

 

Hey neat! I will take a bite from my proverbial hat.

But only a bite.

Completely agreed. I'm under the impression they're working on a new GI system anyway, and I'm sure that if they decide to compete with Nanite (which may or may not even be necessary) that they could get 95% of the results for 95% of developers by providing great mesh simplification tools.

Nanite is technically very impressive, but since I don't want to ship film-quality assets with my games regardless of performance I wouldn't necessarily choose it over a good, traditional style LOD system. The real issues to solve there are workflow and (maybe?) quality moreso than performance.

 

i just wish Unity to polish and finish HDRP (it's still in preview phase).

 

https://blogs.unity3d.com/2020/02/24/hdrp-out-of-preview-in-2019-3/

 

Oh. I didn't notice it's already out of preview. Guess the continous popping internal errors in my HDRP projects causes my confuse.

 

I’m pretty sure if we polled all HDRP users if HDRP is in preview or not, most would say yes.

 

It is still a long time until we will get a new runtime UI, TextMeshPro should be standard text in 2020 for runtime UI but I don't suppose it is...

 

Textmeshpro is fine. It's been production ready for literal years.

 

It is fine, I mean add a UI label and it isn't TextMeshPro by default it is? It is still a package in addition to the UI rather than part of it.

 

Only if you don't select the component that uses TMP.

Spoiler

 

I'd like Unity to ignore it. Focus on the irons already in the fire. Get the various preview packages stable and out of preview.

Yes the Unreal demo was really cool, but what it demo'd is actually something that most of Unity's users aren't after. Unreal has been doing really well in the "almost" AAA space, and Unity has been at a disadvantage there. Let Unreal hold that ground while Unity continues expanding elsewhere. That strategy has gotten Unity to take over 50% of the total games market, and I see no reason to change course now.

 

Wait. You're telling me most people don't actually want an Enlighten without the suck?

 

Not sure what everyone is saying about nanite being only good for uber-realistic games. I thought the main benefit is to not care about baking LODs (even LOD0). Baking models is probably the single biggest time sink in my experience, and the iteration cost is huge. If Unreal lets me drop in my highpoly model and be done with it, that's an incredible benefit.

Realtime no bake GI is a no-brainer.

As it is, I don't think it makes sense for Unity to focus on competing with nanite. But it would certainly help if the engine got a spring clean, and stable features were brought in that focus on biting big chunks out of the game development workflow. Realtime GI has got to be at the top of that list.

 

Yeah. Something where a lightmap that needs ambient occlusion in the corners and underneath.. example.. couches.. doesn't a 1024x1024 lightmap that is a huge white square with little gradients here and there.need for definition. I looked into this one language learning application that uses hotels,restaurants and governments buildings etc. to teach you in an environment where you learn the phrases appropriate to it. They wanted to cut the download size. The lightmaps took up 200 out of a 400 mb download after reviewing the build..

 

My impression was that people typically want GI or Lightmaps, but not necessarily Enlighten.

 

Be careful what you wish for with new tools...


A little barrier to entry can be a good thing in some way. "When everyone is special, no one is special". Back in the day we'd get all riled up over a few polygons, now anyone can download free tools and have the most amazing free 3d assets at their fingertips. Good graphics are increasingly becoming meaningless. Neary anyone with a small bank account can hire competent artists to make a good looking game.

Meanwhile in the pursuit of this, everything else has suffered. Why have retro games made such a comeback? Why are S***ty old games with 15 frames per second, S*** hardware, and S*** optimizations still so fun to play? Why are so many companies in free fall? Why has it been so long since we saw a new genre or must have design tool?

We have become so fixated on graphics that we have created a multitude of tertiary pipelines and considerations that prevent us from refactoring the core to be better. Guilty of it myself these days being knee deep maintaining 3d animation pipelines, particles, AI systems, and trying to "find the fun" is a lot harder than if I just had some cubes flying around on the screen.

It's insanity how little we have done to improve the core pipeline to lead to more fun both in development and in final product. Burnout is the end of forward progress. Quality of life tools DIRECTLY funnels to better games.

We have done everything we can to make this industry a backhanded nightmare: more budget into marketing (sweet lies), a lowering of standards across the board as teams become bigger and bigger, cutscenes and pipelines get longer, graphics go through the roof, but the core experience is little more than what you experienced 15 years ago. Players are tired of your gameplay loop and you haven't even put the game out yet.

We have buried gameplay under a multitude of now standard "fluff". We don't even flinch when trailers don't match the play. Games are just S*** these days and no one really cares, it's just how it is.We're all complicite in this mind game of mediocrity.

Underneath all the billions of polygons are just a few raycasts and rudimentary collision volumes lumbering around. But no one gives a damn about that because it's not as easy to flex on a stat sheet, or to push hardware. For a long time new hardware meant actual advancements. Going from mario brothers, to mario world, to mario 64 was an amazing journey over time, and a triumph of gamedev. Back then hardware meant something. But somewhere along the line the framework required to push the graphics overtook new methodologies for fostering fun. We've stagnated and regressed for about 15 years now. Kids just getting into the industry have no idea what it could have been.

Gamedev has really gone to S*** as of late. Everything's been bought up and turned into a hellish exercize in pushing more tris than the next guy, in out-marketing other studios, in dressing up the dead corpse of your IP's and whoring it out with more microtransactions.

Minecraft, a game where you as a extra chromosome looking mother f*cker comprised of 50 polygons punching trees for lego blocks to build a mud fort comprised of 5 kb of texture memory routines outperforms 99.9999999% games made in fancy editors.

Think about that.

How miscalibrated is the world of gamedev? That fact we've turned the entire process into a top heavy slave-driven effort to sell people snake oil says a lot about us as a species.

You know what would get me excited for the future of Unity?

If i could type the following code

public ParticleSystem ps;
public Transform turret;

ps.allEmission = false; //Disables all emission from current and child particle objects

ps.allSpeed *= -1f; //Easily reverse time of all particles in the group

turret.rotation = Mathf.Lerp(turret.rotation, Vector3.left, rotationRate); //make the turret point to the left over time.

I'm not asking for a lot here, I'm not asking for some savant to dedicate 30 years of their life to make a "make awesome button". Just some key standardizations and ease of use injected into some of the key elements of the game engine that will persist and make it better over time. THAT would excite the F*** out of me. Put all these improvements in one sample document with a simple use case that explores the core functionality of a singular pipeline.

Spend a little time coming up with the absolute best and communicative naming conventions you can, the best functions that get at the heart of what a designer would be wanting to do with the tools in your engine. Make something built to last, an end goal that you will either meet or exceed with your future updates.

We could have an evolving doc with a wish list. Or with a grievances list. "It's a little wonky trying to access particle scale, how can we improve this?". All the pluses and minuses of a pipeline in one place and the potential workarounds with sample code to get you started that likely does 90% of what you want it to do.

But there is no central hub for this stuff, and that is just insane, both from a documentation standpoint and educating your users, and from a self improvement standpoint improving your tools.

Have a vetting system before you add the new tool. Does it improve the pipeline and ease of use? If not then it's not good enough. Rename some naming conventions, make it maleable and solid. Does it have that ergonomic feeling that's so joyful to utilize like the old prefab system used to have? I wouldn't give up the new prefab system for the old one. But I feel it was released too early and too many concessions were made for parity. It should have been released only when it was perfect, backward compatibility be damned. We are now stuck with sub par tools forever.

There are a billion core use cases in Unity and every engine that are a pain. And no one seems to give a damn. We have all the tools and power and tech to figure out better solutions to these pain in the arse problems, but it's not glitzy or glamorous to solve them.

Both unreal and unity make it a pain in the arse to make games, it could be easier, but everyone is running around trying to squeeze out more blast processing, rather than trying to help us make games.

 

Half Life Alyx takes abut 50 gigs. I bet at least 50% if not more is lightmaps. VR devs would love a performant realtime GI

 

This is true in a way, though not literally. More likely, innovation in graphics field yields less and less results. We are getting marginal improvements that nonetheless require new hardware. Would you notice that environment in UE5 demo uses raw megascans instead of pre-generated LODs if a commenter wouldnt tell you? Me for sure wouldnt. Same with a raytrace video cards. From a consumer perspective, there is little difference since industry learned to fake all those effects perfectly before raytracing was available. So yes, I'm pro improving usability and stability of Unity before chasing nebulous "AAA" standards (which are first and foremost determined by budget of the project). If someone is excited about Nanite because it'll let them dump millions of Megascan rocks all over the scene and that will somehow elevate every indie dev to "AAA" level - they have to remember that the rest of their assets need to look on par with those scanned rocks, so characters, weapons, pickups, etc. will be their bottleneck when it comes to graphics. The demo trailer proves my point - the girl's model looks so basic compared to everything else and sticks out like a sore thumb.

 

I want the workflow mostly. If you have tried to cretate a large performant scene in unity you know the workflow is a pain. Almost everything is manual.

 

I think we'll see a rise in "games" similar to "dear esther". Games fit the strengths of the engines that supports them. So i'd imagine we'll see lots of hyper realistic looking environments with lots of narrative makes sense. Maybe some realistic styled survival games. We're going to see the Unreal engine further combine with other mediums as it has already done with TV (the mandalorian).

I was trying to say that graphics no longer matter in the sense that phenomenal graphics are increasingly nothing special, they are becoming standardized. This tech doesn't really empower game dev teams in a way that it makes the games more fun. It simply normalizes hyper real looking graphics, and frees up monetary resources that can be used elsewhere. This doesn't necessarily help studios produce games that are more fun, it just forces them into unreal to get this level of fidelity.

This is a narrow view however. The potential for new interactive VR/ Shows/ Movies/ media/ communication/ time wasting outside of the gaming sphere is MASSIVE, but that's not what i'm focussing on here.

I'm trying to say there is still an untapped market for an engine that wants to focus on the moving parts that makes games fun and functional.

The problem with improving the pipeline of making games, is there is a disconnect usually between the back-end coders and tool creators and the designers. A sort of a resentment between the two. "Why should I bother polishing this tool for some idiot that doesn't even know how it works?". "Do these coders even attempt to use their own tools? This is damned near unusable".

Nothing is ever good enough to placate the designers who want a "make everything awesome button", and both sides become increasingly burned out over the disconnect and compromises that arise over time. What Unity and Unreal need to do if they want to solve usability issues, is breach this divide as best as possible, by finding individuals who are still passionate about the entire pipeline who can mediate between the extremes of the pure designer and pure coder.

I'm just blathering and stating the obvious at this point. It's obviously a very complicated situation where game engines like unity must strike the right chord between innovation, backwards compatibility, and following market trends. But it seems like amidst all the madness the actual practice of making games has been stagnating for some time.

 

This would be very welcomed in Unity. To bring assets and animated characters from Blender without having to change rotation for them or to export them as .fbx

 

You can already drag and drop a blend file directly into Unity. I just verified it with Blender 2.8 and Unity 2019.3.

 

You've been able to do this since like... 3.x too, I'm pretty sure.

 

fbx format is considered optimal and best practice, so always better to export where possible. my friend uses something called '3dsmax' and this is her workflow. File size is smaller and animation and rigs more reliable, along with textures too.

 

I think unity must continue what they are doing!! The roadmap of this year is to stabilize all the features and tools and unifying all the graph based tools which is great!! I read on forums and posts about the limitations of UE5 tech which cannot be ignored ... First of all the tech demo was running on just a slight better resolution than full HD(where nowadays where people expect to play 8k resolution games on 120fps) and also it was only running at a very low fps of 30fps and the scene was really very small.... As per epic this small scene rquires RTX 2070 level of graphics card and a powerfull SSD was a must to run such a SMALL scene....the nanite tech does not support things like trees, grass, leaves, skinned meshes etc... The only thing it supports is rocks!!! The game called hellblade 2 using ue5 does not have a single TREE in trailer!!! The new voxel based GI will be good to be used with nanite as nanite uses voxel to render this much high polygons ... But if u want to use voxel GI with normal mesh the performance will suck even more!!! The tech cannot never run on mobiles!!!

 

Nice necro.

Also, yeah, they should continue doing what they're doing. Their response so far to Unreal showing a high tech new GI (among other things) solution was...

To bring back Enlighten.

So if they continue what they're doing, which is to go backwards in time, eventually they'll bring back Beast and I'll be finally happy.

 

For me nanite is more impressing than lumen, if it can live up to the hype.

 

Their response to that was to buy Pixyz and charge thousands of dollars per year for something that is a core part of a game engine.

 

I pointed out nanites limitations :-

Are u really hyped now??! Even I was hyped but these limitations cannot be ignored

 

If it can autolod lower fidelity games like ours it's a game changer for indie workflow. I don't need the fidelity of that demo I want a better workflow for lod and batching

 

I think auto LOD was already there in UE 4?!

 

UE4's LOD pipeline is already light-years beyond Unity as is. You can create LODs for both static and skinned meshes automatically directly in the engine, LOD is handled at the asset level (so no need to add components and bloat your scene hierarchy with LOD objects), and there's built-in tools for mesh baking.

 

Nanite is not just autolod it streams the meshes. So it can adjust the quality changes much more smooth and avoid micro triangles (triangles smaller than a pixel) all together.

 

I don't find the tech that useful because it only runs on SSD which are costly for PC' s which means lower people playing your game and I don't think naniIte's mesh can be smoothed so the triangles have to be of pixel level

 

What rock have you been living under? While most components have increased in price thanks to the current world conditions SSDs have only become cheaper with time and larger capacity.

 

He might mean gen4 NVMe though, which is the only disk available on pc that can match the ps5

 

Speaking of which if he wants to play any modern games he had better buy that SSD because I'm very confident that few if any games that include a console port will be playable with HDDs. I know you know but what do people think their HDD will be able to do when the game is expecting a 5.5GB/sec SSD?!

 

Yeah and they are not that expensive compared to other components. I bought the samsung 980 pro and I'm super happy with it

 

You must not find raytracing very useful either, but here we are.

Remember that unlike Unity, UE4 is used by top studios who want their games to be look on par with the most technically advanced games on the latest console generation.

 

What are you actually talking about? Like, nothing you have said is accurate. For a start nanite does not just handle rocks, if you are saying that just because that is all you have seen in screenshots then you are jumping to conclusions as you have done throughout your posts in this thread.

Secondly, SSDs are in almost all PCs now, they are one of the cheapest components and will be standarized in consoles as the new gen goes more widespread.

So yeah, not really understanding where your "insights" are being taken from, but you are using wildly inaccurate sources clearly.

 

He might use some information on the demo and state of the Nanite released like a year ago.
The last time Epic talked extensively on UE (Unreal Fest in summer?) it mentioned that currently, this new renderer lacks few things. Obviously, this brings a parading shift in the world of rendering.



@EagleG
But that's why they "only" presented the demo, not production-ready UE5. Technical demos like this one are meant to validate new technology. Check if it useful for artists? What are the benefits and new challenges? What's needed to make it run and be good enough to show it to the public?
Such a demo is essential to finish the new tech, it wouldn't be possible to test it out otherwise. Still, just an early tech demo.

AFAIK in UE5, you can mix Nanite meshes with traditional ones. Nanite doesn't need to support everything.
And engine gonna provide a fallback path, so the game could run there just fine.

Of course, they supported rigid geometry first. The simplest type of geometry and this what's used to create environments. Enough to make a demo and impress the entire industry. Soon after the presenting this demo (and new licensing terms on the same day) many non-Unreal studios evaluated their strategy for coming years.
I already helping out some "all-Unity studio" with their first Unreal game. I bet we gonna there is more studio like this, partially encouraged to switch engines thanks to this demo. And these studios might already release their first Unreal games before Nanite would become production-ready.

As you can showing a demo of the brave new technology is very important. It doesn't matter that much if that wasn't using the production-ready version on Nanite yet.

 

Since Nanite is basically generating one triangle per pixel I'm betting it works fantastically with DLSS.