TTLG|Thief|Bioshock|System Shock|Deus Ex|Mobile
Page 1 of 2 12 LastLast
Results 1 to 25 of 45

Thread: Unreal 5 engine is amazing

  1. #1
    Member
    Registered: Nov 2019
    Location: Restaurant at end of universe

    Unreal 5 engine is amazing



    Hopefully this comes to PC. One programmers for the game speaks and they show in-game play.

  2. #2
    Chakat sex pillow
    Registered: Sep 2006
    Location: not here
    There's no hopefully about it. It's a multiplatform engine from the get-go and backwards compatible with UE4. Word is they're planning to launch it in 2021.

  3. #3
    Moderator
    Registered: Jan 2003
    Location: NeoTokyo
    No gonna lie, I looked at that like Homer looks a donut, gargling noises and all.

  4. #4
    Member
    Registered: Mar 2001
    Location: Melbourne, Australia
    1:10 - "It just works"


  5. #5
    Level 10,000 achieved
    Registered: Mar 2001
    Location: Finland
    Very pretty! Tho it won't get me to switch to UE for my own gamedev. Y'know, just because you use UE5 doesn't mean your game is gonna look like that. It still requires a lot of work and artistry. And I know that with the amount of effort I put into my visuals, my games are gonna end up looking pretty much the same whether in Unity or UE, so I'll be sticking with the one I know.

  6. #6
    Member
    Registered: Apr 2001
    Location: Switzerland
    When I look at this, as someone who doesn't really know all that much about engines, I wonder to what extent this makes development easier and to what extent it requires production values that are insanely expensive. I can see how more flexible, nuanced lighting syste,s or self-generating lower detail levels for game geometry etc. could save time, but at the same time I wonder whether such engines wouldn't also require models and environments at massively higher detail levels to begin with for the results to look considerably better than games already do. How much of a new engine is dedicated to intelligent automation (such as the Inverse Kinematics stuff)? And how much is about more polygons, more shaders, more detail and the like?

  7. #7
    Chakat sex pillow
    Registered: Sep 2006
    Location: not here
    My understanding of the current production pipeline (guys in gamedev, correct me if I'm wrong) is that art assets are initially made at as high a quality as the artists can go, after which they're baked into low-poly versions for the game with 'hero' assets that are the focus of any given scene given more of the polygon budget. The Nanite tech seems to promise that you don't need that step any more, as the tech will take care of it for the artists so they can just go whole hog and not worry about optimising the end result.

    Now the bigger question of what kind of fidelity are artists expected to crunch at this point for AAA: given the amount of clutter you can add to a scene with the new generation, there's definitely going to be more asset creation required. But that's really one of the big reasons why asset libraries like Quixel's with its photogrammetry scans (the entire scene is based on their limestone quarry assets) are going to be more and more important.

  8. #8
    Moderator and Priest
    Registered: Mar 2002
    Location: Dinosaur Ladies of the Night
    Banging out high res assets is actually fairly quick, and texturing them isn't that overwrought of a process, since at the resolution we're talking about here, you're really more concerned about coating your various objects with the appropriate material, rather than painting in specific details.

    Not having to retopologize your highpoly meshes for a low res bake, and banging out various LOD models would save TONS of time. Those are usually the most time consuming parts of the entire asset building process, easily taking up 50% of your time at least.

  9. #9
    Moderator
    Registered: Jan 2003
    Location: NeoTokyo
    What would really make this engine shine is if somebody cooked up a system for procedural asset generation and animations that are lifelike.

    Alright, so this is my vision of the, or an, ideal future of indie game-making.

    For players and AI ... if you've seen Fuse, you just take a stock human model, then you can pop it into a model program and tweak it (there's the in-game parameter-based tweaking, like cheekbone, nose shape, etc, etc, but it's easy to do just in Blender from the base), then you click a button and it rigs it and attaches it to a skeleton that already has a stock of like 300 animations that automatically work for that model. Then you make that open source so people continually grow the base of animations, clothes, etc, kind of like Second Life. And then you have Euphoria, which has a bunch of procedural human reflexes and this context-sensitive procedural gesture stuff. It isn't hard to imagine extending it to major categories of animals and creatures, quadrapeds, birds, fish, etc.

    For static objects, what'd be great is AI-generated objects from photo sources. So you just grab of photo of an object (maybe a set of photos from 6 directions, but a good AI system could predicatively fill gaps), click a button, and then AI generates a textured object out of it, and it could predict the material type on it, or you just paint it on, etc.

    And then for geometry, something like Voxel Farm, or other procedural systems, where a system generates a real world geometry based on geology principles (age of the planet, weathering, climate, etc.), and then you can edit it from there, like Minecraft except at the voxel level, which then coverts it to pixels.

    All this kind of tech has already been made. It's just a matter of it all coming together in a unified system. So the idea is, you click a button and it generates a realistic world geometry that you modify from there, you drag in photo sources that generate textured static objects, and you have the human/animal generator that you tweak and add animations, place everything... The point is, if procedural generation of realistic assets catches up with this, it's going to be great for people to make realistic maps and focus just on the placement, art design, etc. All the pieces have already been made; it's just a matter of bringing them all together. And in an engine like this, it'd bring a lot of power to individuals who otherwise wouldn't be able to do anything with an engine like this because of the prohibitive difficulty-curve of making realistic assets by hand.

  10. #10
    Member
    Registered: Apr 2001
    Location: Switzerland
    To what extent might it make things more difficult for games with a non-realist aesthetic if the tools are all geared towards one form of realism or another?

  11. #11
    Moderator
    Registered: Jan 2003
    Location: NeoTokyo
    I saw a tutorial for Fuse that gave me thoughts about that issue. So the point of Fuse is that the algorithm procedurally rigs a bipedal model to a skeleton that already has 100s of animations attached to it. In the tutorial, the guy took the base model of a human, and to make a long story short, he had a method that best-fit the vertex-arrangement to this troll monster he'd made that was wildly non-realistic and nothing at all like the original model. But the beauty of it was that, because it was still vaguely bipedal, the vertex grid still lined up for the rigging algorithm to work, and his massive and cartoonish troll could automatically use the skeleton and all the animations. And then IIRC he could even pop it into Blender's animation shop and tweak all the pre-made animations to just wiggle a little more here and there to fit the model's aesthetic.

    The basic punchline is, if you have a really good general procedural system, it's actually going to (or they should make it be able to) make it easier to go for a non-realistic aesthetic because you get all the groundwork taken care for you, and all you have to do is add the bits that make it distinctive, vastly cutting down the overhead to making such an asset. I understand the risk of what you're saying, and my response is the tools should be made in a general format that works with even really alternative visions, and I didn't really emphasize it earlier, but I was thinking from the start about categories like "vaguely bipedal", "vaguely quadrupedal", "vaguely fish-like", etc, etc.

    Edit: Sorry, I just realized you're talking about UE5's tools not what I said. But in that case ... the thing is a billion pixels is a billion pixels. I have an idea people are going to come up with some really otherworldly stuff if they're really unbounded, anything that can be modeled can be put to screen.

    I'm much more worried about the organizational culture of the studios that have the resources to take advantage of this tech, i.e., financially risk adverse, meaning low-risk, hypersensitive to acceptable ROI, focus-group approved "safe" aesthetics, than the tech itself gearing art direction one way or another. I think any good artist can develop their own style with any decent tech if left to their own devices, but financial and organizational pressures are the real limiting factor IMO.
    Last edited by demagogue; 14th May 2020 at 08:40.

  12. #12
    Member
    Registered: Mar 2001
    Location: Melbourne, Australia
    So if the average decent AAA production game present day takes 3 - 5+ years to make, games under this new engine will take 10?

  13. #13
    Quote Originally Posted by icemann View Post
    So if the average decent AAA production game present day takes 3 - 5+ years to make, games under this new engine will take 10?
    They're aiming for the opposite, a lot of the time spent in development of those large titles is spent making and optimizing the various graphics. If UE5 makes a lot of the currently (semi-)manual graphics optimization redundant, there's the potential for a noticable amount of time to be saved.

  14. #14
    Still Subjective
    Registered: Dec 1999
    Location: Idiocy will never die
    Does look good. Feels like an ad for the PS5, which I'm probably going to get anyway.

  15. #15
    Member
    Registered: Dec 2001
    Quote Originally Posted by WingedKagouti View Post
    They're aiming for the opposite, a lot of the time spent in development of those large titles is spent making and optimizing the various graphics. If UE5 makes a lot of the currently (semi-)manual graphics optimization redundant, there's the potential for a noticable amount of time to be saved.
    It seems like people sculpting stuff in Zbrush and making megascans would be able to skip the optimisation phase, I wonder how will it bump memory and disk space / download requirements for games. We already have >50 gig games, which is a nuisance to download on average connection.

    The realtime GI / "infinite bounces" with Lumen is exciting though. No more hours wasted on baking lightmaps (which can take hours for a fairly finished level). Lighting artists will have a blast if it's as good as it looks like.

  16. #16
    Moderator and Priest
    Registered: Mar 2002
    Location: Dinosaur Ladies of the Night
    Quote Originally Posted by Judith View Post
    It seems like people sculpting stuff in Zbrush and making megascans would be able to skip the optimisation phase, I wonder how will it bump memory and disk space / download requirements for games. We already have >50 gig games, which is a nuisance to download on average connection.
    Here's a question for you: if UE5 is able to use high res models straight out of Zbrush en masse, would you even need normal and heightmaps anymore?

  17. #17
    Member
    Registered: Dec 2001
    If you wouldn't need normalmaps, that could lower the disk space cost, provided the high res models wouldn't take much more. I'd have to make a comparison, but something tells me that e.g. a million-polygon .fbx model needs much more space than a 4k normalmap.

  18. #18
    Member
    Registered: May 2017
    Location: USA
    Karras, hm? I don't trust it...

  19. #19
    Moderator and Priest
    Registered: Mar 2002
    Location: Dinosaur Ladies of the Night
    Quote Originally Posted by Judith View Post
    If you wouldn't need normalmaps, that could lower the disk space cost, provided the high res models wouldn't take much more. I'd have to make a comparison, but something tells me that e.g. a million-polygon .fbx model needs much more space than a 4k normalmap.
    Right. You'd only need diffuse and spec maps, the former of which can be compressed somewhat. Without high definition normals, heightmap (does UE4 use 16BPP for those?), and multiple LOD models, you'd probably be looking at a slight surplus in relative asset size.

  20. #20
    Moderator
    Registered: Jan 2003
    Location: NeoTokyo
    The guy in the video said directly there's no need for a normal map and LOD models.

    I could probably look this up, but I'd imagine the file size of a model is proportional to the number of vertexes, so that a million poly model shouldn't be smaller than 3MB or whatever.

    If you watch distant geometry in some parts of that video, you'll see it morphing, which look like dynamic LOD, which I guess it would have to be if it's allowing an unbounded line of sight.

    He also said something about specular when the flashlight was reflecting off the ceramics, which makes it sound like it's more than just a vanilla specular map, but I didn't really catch the significance. I don't think it was procedurally computed specular and diffuse, but who knows what they're capable of anymore.

  21. #21
    Member
    Registered: Aug 2003
    Location: Jafaville New Zealand
    Well looks like the System Shock remake is going to be pushed back to 2022.

  22. #22
    Moderator and Priest
    Registered: Mar 2002
    Location: Dinosaur Ladies of the Night
    I'm just wondering what the real bottleneck is for this. We're all making it sound like it's just WHATEVER DUDE! FUCK THE POLYGONS! JUST GO FOR IT! But comeon, there has to be some limitation here.

    Even with some ultra snazzy tessellation at work, nigh unlimited polygon budgets seems too good to be true.

    I could probably look this up, but I'd imagine the file size of a model is proportional to the number of vertexes, so that a million poly model shouldn't be smaller than 3MB or whatever.
    They can get surprisingly large. I'll try and do an experiment right fast, see how big I can make a mesh with my piddly machine, and look at the .fbx filesize.

    edit: A 3.14 million tri sphere with a smiley face sculpted on it gave me an 84 meg .fbx file. I was also surprised at how well Blender handled a polycount that high.
    Last edited by Renzatic; 14th May 2020 at 23:34.

  23. #23
    Member
    Registered: Dec 2001
    Quote Originally Posted by Renzatic View Post
    edit: A 3.14 million tri sphere with a smiley face sculpted on it gave me an 84 meg .fbx file. I was also surprised at how well Blender handled a polycount that high.
    A 4k normalmap compressed with DXT5nm is slightly below 22 megs. 1M poly .fbx mesh is 30 mb. 4M poly mesh is 115 mb.
    Last edited by Judith; 15th May 2020 at 01:38.

  24. #24
    Chakat sex pillow
    Registered: Sep 2006
    Location: not here
    Well, I think we can foresee those 1 TB SSD sizes getting maxed out by around 4 titles, give or take, if we're talking source models and 4/8K textures and lightly compressed audio going directly into a game.

  25. #25
    Moderator
    Registered: Jan 2003
    Location: NeoTokyo
    Next in line is 4+TB SSD becoming standard then. (I want to imagine. I don't really know the technicalities.) But anyway HD size is going to be the next obvious bottleneck if every other game is suddenly shooting for photo realistic fidelity. Maybe cloud gaming could take off to address it. Anyway, somebody will have to do something about it.

Page 1 of 2 12 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •