Hi guys! WOW! I wasn't expecting so many views so quickly. Thank you! ❤ I would love to hear from you, what topics would you like me to cover in the future? Any problems you have encountered or you want to do something but don't know how? Leave a comment below and I might make a video about it! 👇
Thanks to the TH-cam algorithm for putting an amazing person like you on my suggestion list❤. Keep going bro any tricks you will post will be popular if you put an attractive thumbnail like this one ✨👌
Big, gigantic thanks. I was looking for ways to reduce some of the level bloat as it was starting to get out of hand, even if Nanite makes it easy to render on GPUs the sheer storage requirements were starting to hurt, this is great on multiple levels. Much appreciated.
I mean, yes, I can see it very clearly, there is quite some difference in detail. But then again, I pay a lot of attention to detail. But the low poly version has enough detail for most people.
Exactly. There is definatelly a difference, but an average viewer will probably not notice it. My girlfriend made 2 guesses on the 5 rocks example and was wrong twice 😅 I wouldn't use AUTOLOD if you need a high quality optimized asset. But if you just need a quick result - its handy to have it right there in Unreal.
You gotta keep in mind that most people won't shove their face into models, and very small details are sometimes actually impossible for people to see because the camera simply doesn't allow them to get close enough.
Hey man, thanks for this cool tip! I was always optimizing within static mesh window but that only reduces the amount of triangles. This one bakes geo maps as well. Neat!
If you have packaged actors, is there a way to cull the geometry that can't be seen due to the actors' collision is hiding polygons inside each other's meshes?
Good tip but if you're using Nanite, don't you get the same result with changing the "Keep Triangle Percent" that you can see on the right in the mesh editor window at 1:13 ? You have it set to 100%, change it to 0.1% and compare the result. The information on the top left will adjust too, like the disk usage and such.
Hey Sweenus. I have set Triangle percent to 0.5% (to come to 5k triangles) and the size is 0.26 MB (0.1 nanite) vs 0.13 MB (0.06 nanite) from autolod. see screenshot below. The mesh looks pretty good too. The downside is that I have to go into every texture and lower its resolution too, while autolod bakes it for me automatically. Also the original uasset is still there. (which in my case is 67MB and 750 MB from 8K textures). I would prefer the ability to just delete the original asset and save drive space. In the end I work on projects for VR on Meta Quest and cant use nanite, so Im stuck to autolod for now 😆 But you are right! Its a viable option too! Good to know! Thank you! drive.google.com/file/d/1T_PvP-yzERU8kn3cHXWpf42-wGxP67tg/view?usp=sharing
You can still see the difference in the video when I switched on nanite triangle view. I personally use autolod when I need the same assets for the close ups and for the background. There is no reason for the background assets to use 1 million polygons each, even if its nanite. Besides there are cases where you cant use nanite, not all devices support it, DX11 or consoles (eg switch) dont.
Nanite itself is expensive to run, not supported on mobile, VR, etc. On a lot of projects you won’t be able to use it. Also getting rid of the high poly assets saves disk space. In my experience using optimized assets with LODs is way cheaper than using Nanite. Nanite is good for next-gen systems and quickly designing incredibly detailed scenes, but not for games where maximum optimization is required
yes there is more, the more polygons the more nanite has to work with, and 4K textures are 4 times bigger than 2K textures, if you have a game with thousands of objects that have million polygons and 4K textures it will run more poorly than with optimized meshes, even with nanite and take way more space on disk
You can definitely tell the difference between 1M Tris and 1,000 Tris at that close of distance, Jagged Edges and a difference in how shadows are received makes it obvious Still everyone should optimize their Geometry and you can easily get better results for close objects by lowering less to maybe 3000 Tris and more for long distance objects
Hi Unreal ArchWizard!!! I love this tip. How would you go about using this with something like the city sample assets in the marketplace? They are all in blueprints made up of many different static meshes. Pull each section out and replace it with a AUTOLOD'd version?
Now, Idon´t know sh*t about unreal engine other than it´s a graphics engine and looks great. But is it possible to change the triangle count realtime during live run, so that the triangle count can change according to distance to camera?
Yes, thats exactly what games do. Usually, the meshes change between like 5 different Levels of Details (LODs). But in Unreal there is now Nanite which does that automatically and saves a bunch of time for developers.
@@unreal_archviz So cool. I look forward to play games with these kind of graphics. I'm old enough that my first computer was a Sinclair ZX Spectrum 48K with really shitty graphics. So I really enjoy the development. Even then my friends and I was talking about how realistic the graphics looked when the Amiga came.... 😀
We need this kind of video, quick, simple, and important information for development! Thank you for creating this video, I hope I can see more video like this from you again.
This is simply moving size on disk to size in memory, by adding an extra 2k texture to the streaming pool. If you are using Nanite, this is simply a bad idea, as Nanite compression is fantastic.
If you keep both of the models, yes. But it you delete the original and keep the new one - you are saving memory. Also you cant always use nanite. I do many VR projects and I cant use it there. Its still useful to know that Autolod is right there.
It works with all static meshes. I wouldn't use it for the weapons and enemies because these are some of the most important objects in the game and the player's eye often focuses on them. It's better to optimize them properly. AutoLod is more for situations when you just want a quick result for some mid to background meshes.
You are not the first person to think I'm Serbian. But close, I'm Ukranian. I used to hang out with Serbians at the Uni. You could tell I wasn't from there because I was the shortest in the group 😅
@@unreal_archviz A hahaha, yeah guessed cause of English accent, Serbs sound the same speaking it, and I feel you, I'm only around 180 which is short for a Serb
Honestly, I was barely able to spot the difference (when you only showed 2, when you showed 5 it was not possible due to the video), its the blacks and an educated experienced unreal user will tell the difference, but thats you going from 1 000 000 down to 5 000, like, if you go down to 20K which is still relatively low, I do not think I'll be still able to spot the difference anymore...
Then they wonder why games have such poor optimization, because idiots make models of stones with 1 million polygons. There should be 1 million polygons for the entire level, not per stone.
Decimation has been around for decades. Anyone could add the feature to any engine.. it's been in many engines. ..oh, sorry, I am probably ruining your exploring fun.. my bad. I'll shut up.
Hey there! No worries at all, I appreciate your input. You’re absolutely right, decimation is a well-established technique and it’s been integrated into various engines over the years. The AUTOLOD feature in Unreal Engine 5 just makes the process more accessible and streamlined within the engine itself. You can often get away with just decimating organic forms that won’t be animated: rocks, tree trunks, and so on. It’s all about finding the right balance for the project at hand.
Its not that is should exist , it is a where to find it that is the important part , there are so many hidden menus and stuff deep in the engine that most users dont know pretty standard things are just a couple clicks away
Hi guys! WOW! I wasn't expecting so many views so quickly. Thank you! ❤
I would love to hear from you, what topics would you like me to cover in the future? Any problems you have encountered or you want to do something but don't know how?
Leave a comment below and I might make a video about it! 👇
Thanks to the TH-cam algorithm for putting an amazing person like you on my suggestion list❤. Keep going bro any tricks you will post will be popular if you put an attractive thumbnail like this one ✨👌
@@oliverdive9759 Wow! Thank you very much for such kind words! ❤
Unreal Community needs more of these quick tip videos. keep it up, this is great content and super useful!
Thank you! Will do more of it!
Big, gigantic thanks. I was looking for ways to reduce some of the level bloat as it was starting to get out of hand, even if Nanite makes it easy to render on GPUs the sheer storage requirements were starting to hurt, this is great on multiple levels. Much appreciated.
Pretty useful for asset packs and Megascans. I usually export to Blender, decimate then import back but this is much faster.
Thank you!
your channel will become an irreplaceable source of knowledge; keep it up mate!
You have inspired me to give youtube a try, so hearing such enouraging words from you is crazy 😅 Thank you, mate!
Quality little tip, the algo did its work and brought me to you, nice one mate
Thank you! ❤
I mean, yes, I can see it very clearly, there is quite some difference in detail.
But then again, I pay a lot of attention to detail.
But the low poly version has enough detail for most people.
Exactly. There is definatelly a difference, but an average viewer will probably not notice it. My girlfriend made 2 guesses on the 5 rocks example and was wrong twice 😅
I wouldn't use AUTOLOD if you need a high quality optimized asset. But if you just need a quick result - its handy to have it right there in Unreal.
You gotta keep in mind that most people won't shove their face into models, and very small details are sometimes actually impossible for people to see because the camera simply doesn't allow them to get close enough.
I would say 50k polys is more than enough for big assets in a FPS game when you look at it up close.
amazing simple tip. Want more!
Hey man, thanks for this cool tip!
I was always optimizing within static mesh window but that only reduces the amount of triangles. This one bakes geo maps as well. Neat!
If you have packaged actors, is there a way to cull the geometry that can't be seen due to the actors' collision is hiding polygons inside each other's meshes?
Good tip but if you're using Nanite, don't you get the same result with changing the "Keep Triangle Percent" that you can see on the right in the mesh editor window at 1:13 ? You have it set to 100%, change it to 0.1% and compare the result. The information on the top left will adjust too, like the disk usage and such.
Hey Sweenus. I have set Triangle percent to 0.5% (to come to 5k triangles) and the size is 0.26 MB (0.1 nanite) vs 0.13 MB (0.06 nanite) from autolod. see screenshot below.
The mesh looks pretty good too.
The downside is that I have to go into every texture and lower its resolution too, while autolod bakes it for me automatically.
Also the original uasset is still there. (which in my case is 67MB and 750 MB from 8K textures). I would prefer the ability to just delete the original asset and save drive space.
In the end I work on projects for VR on Meta Quest and cant use nanite, so Im stuck to autolod for now 😆
But you are right! Its a viable option too! Good to know! Thank you!
drive.google.com/file/d/1T_PvP-yzERU8kn3cHXWpf42-wGxP67tg/view?usp=sharing
If it’s a nanite mesh, what’s the point? Nanite’s rendering complexity is proportional to pixel count, not poly count. It wouldn’t make a difference.
You can still see the difference in the video when I switched on nanite triangle view. I personally use autolod when I need the same assets for the close ups and for the background. There is no reason for the background assets to use 1 million polygons each, even if its nanite.
Besides there are cases where you cant use nanite, not all devices support it, DX11 or consoles (eg switch) dont.
It also reduces the disk size by a lot too
Disk size, and I guess memory usage would both be greatly improved I would assume.
Memory usage goes up. You are adding an extra 2K texture.
Nanite itself is expensive to run, not supported on mobile, VR, etc. On a lot of projects you won’t be able to use it. Also getting rid of the high poly assets saves disk space. In my experience using optimized assets with LODs is way cheaper than using Nanite. Nanite is good for next-gen systems and quickly designing incredibly detailed scenes, but not for games where maximum optimization is required
doesn't the nanite negate the need to optimize the amount of polygon on a mesh? or is there more to it than that?
yes there is more, the more polygons the more nanite has to work with, and 4K textures are 4 times bigger than 2K textures, if you have a game with thousands of objects that have million polygons and 4K textures it will run more poorly than with optimized meshes, even with nanite and take way more space on disk
You can definitely tell the difference between 1M Tris and 1,000 Tris at that close of distance, Jagged Edges and a difference in how shadows are received makes it obvious
Still everyone should optimize their Geometry and you can easily get better results for close objects by lowering less to maybe 3000 Tris and more for long distance objects
You are underrated as f-k, not gonna lie keep it up!
This is very encouraging, thank you!
great tip, thanks. How do you make the modelling icons *(UV, deform, model..) in one column?
LOVE VIDEOS LIKE THIS! more video about optimization please. maybe fog or skeletal mesh optimization if there's any
Thank you! I'll look into fog or skeletal mesh stuff. If you have any more topics you would want me to cover, let me know too!
But why are Nanite Assets not "Retopologized" in any way like other 3d models used? Seems pretty inefficient to me
Hi Unreal ArchWizard!!! I love this tip. How would you go about using this with something like the city sample assets in the marketplace? They are all in blueprints made up of many different static meshes. Pull each section out and replace it with a AUTOLOD'd version?
Nanite is reason alone for devs to adopt UE. All the retopology and normal mapping that I get to skip -- awesome.
Now, Idon´t know sh*t about unreal engine other than it´s a graphics engine and looks great. But is it possible to change the triangle count realtime during live run, so that the triangle count can change according to distance to camera?
Yes, thats exactly what games do. Usually, the meshes change between like 5 different Levels of Details (LODs). But in Unreal there is now Nanite which does that automatically and saves a bunch of time for developers.
@@unreal_archviz So cool. I look forward to play games with these kind of graphics. I'm old enough that my first computer was a Sinclair ZX Spectrum 48K with really shitty graphics. So I really enjoy the development. Even then my friends and I was talking about how realistic the graphics looked when the Amiga came.... 😀
Nice, but they are nanite, could you show the difference in performance ?
Great work!! Thanks for sharing bro!!
onpoint .. nice
Wait it downscales textures too? Damn nice thanks!
nice video hope we see more optimization videos maybe foliage ?
Thank you! I added it to my list of topics to cover!
So any way, I started subscribin’.
Great tip! Thanks brother!
thanks bro, amazing tip!
Great fun video bro 💪
Ty exactly what i needed. keep it up
thank you very muchly ❤🔥
Thank you!
really good! ❤
We need this kind of video, quick, simple, and important information for development! Thank you for creating this video, I hope I can see more video like this from you again.
Thank you!
Is that "chào" 😲 at the end of video😮😮
Yes 😀
Of course you can't see the difference if you keep the camera 20 m from them.
This works for simple geometry, but not for complex ones such as bushes and trees.
This is simply moving size on disk to size in memory, by adding an extra 2k texture to the streaming pool.
If you are using Nanite, this is simply a bad idea, as Nanite compression is fantastic.
If you keep both of the models, yes. But it you delete the original and keep the new one - you are saving memory. Also you cant always use nanite. I do many VR projects and I cant use it there. Its still useful to know that Autolod is right there.
You sir are irreprace... irrepray... irreplacible!
Haha. 😂 Thank you !❤
its clearly noticeble
Does not work in Unreal engine 5.4
cool tip
just
does that work with outside assets or even weapons/enemies?
It works with all static meshes. I wouldn't use it for the weapons and enemies because these are some of the most important objects in the game and the player's eye often focuses on them. It's better to optimize them properly. AutoLod is more for situations when you just want a quick result for some mid to background meshes.
Are you Serbian?
You are not the first person to think I'm Serbian. But close, I'm Ukranian. I used to hang out with Serbians at the Uni. You could tell I wasn't from there because I was the shortest in the group 😅
@@unreal_archviz A hahaha, yeah guessed cause of English accent, Serbs sound the same speaking it, and I feel you, I'm only around 180 which is short for a Serb
I mean I could tell
Honestly, I was barely able to spot the difference (when you only showed 2, when you showed 5 it was not possible due to the video), its the blacks and an educated experienced unreal user will tell the difference, but thats you going from 1 000 000 down to 5 000, like, if you go down to 20K which is still relatively low, I do not think I'll be still able to spot the difference anymore...
LODs.
Hi good evening I want to be your editor I send my proposal in PeoplePerHour
Clickbait! It doesnt have 1.000.000 triangles but only 999.996! :P
You caught me... 😅
Then they wonder why games have such poor optimization, because idiots make models of stones with 1 million polygons. There should be 1 million polygons for the entire level, not per stone.
Decimation has been around for decades. Anyone could add the feature to any engine.. it's been in many engines.
..oh, sorry, I am probably ruining your exploring fun.. my bad. I'll shut up.
Hey there! No worries at all, I appreciate your input. You’re absolutely right, decimation is a well-established technique and it’s been integrated into various engines over the years. The AUTOLOD feature in Unreal Engine 5 just makes the process more accessible and streamlined within the engine itself. You can often get away with just decimating organic forms that won’t be animated: rocks, tree trunks, and so on. It’s all about finding the right balance for the project at hand.
I think this knowledge is still useful to learn, especially nowadays where "optimization" is not a word in most 3D designer's vocabulary
Its not that is should exist , it is a where to find it that is the important part , there are so many hidden menus and stuff deep in the engine that most users dont know pretty standard things are just a couple clicks away