And, for anyone wondering, yes, this same process can be applied to more complex objects as well. You'd be surprised how high-res you can make a low poly asset look with the right baking and textures :) The full intro to UE5 playlist can be found "Completely FREE" on TH-cam here: th-cam.com/play/PLZpDYt0cyiuu1ZKrZwxOWTRjTUDXMmGaX.html
8:35 This is probably the most important bit of information to fully optimize textures. For anyone interested, look for texel density, you can't even imagine how much performance are lost for 4k texture or 8k texture on miniscule props that will never be seen upclose. For example, in Gear 5 for walls and static meshes they used 400cm=2048px (so a 2k texture will cover a square 4mx4m)
Dude, thank you for videos like this. I feel like there's not enough content about how to approach game design from a standpoint of optimizing, procedures on file management and structure, etc. It's not a "sexy" topic, but it's so important and I appreciate you for it.
Thank you so much for this video it's insanely helpful! I know in blender you can use the multi-resolution modifier to bake high poly details to low poly objects as well but I'm not sure about texturing in that work flow.
A few months ago i upgraded from a few years in blender, to Maya, and man do i absolutely LOVE it a lot more for anything!! it all just feels a lot more legit and is just better, sadly there are a LOT less tutorials like this because it isn't free... but it is the standard in the industry. hopefully you can do tutorials with maya and unreal soon man! keep it up i love all that you do!
Generally, this is a good explanation. If I can only add notes from the technical art side, maybe it would be better to model only one base cube, texturize it in substance, and make a color mask. One master material is created, and then only the base color is changed in the instances. That's how we solved all the combinations with just one base color texture. Also, this way of modeling is not so useful when making a Tetris game, I'm talking about how the programmer will turn off the entire line when the blocks are lined up if they are all combined in the mesh. A better way is to make only one block, and then combine it in the blueprint.
Awesome work - fantastic explanation of the pipeline - this pretty much has it all (I personally would have also added collision into the mix but that's just me ofc) but this will be my go to vid when explaining to peeps what the stages are.
Great video, but there are a few notes. If your mesh is using Nanite, then having fewer polygons is actually LESS optimal than having more of them, especially if the mesh is enough to be used as an occluder for other meshes. Also, you can reduce the Normal maps as half of the Albedo and ORM a quarter. This is a rule of thumb, not necessarily true for all models, but it's also what Epic suggests to reduce texture footprint without losing visual quality.
@@Navhkrin less polys is better in general. but when a model gets convert to a nanite mesh (whats been done in UE5) a very low poly count might interfere with the workings of it. I dont know the full ins and outs of nanite myself, but calling someone's comment bs without giving a counter argument is bs on its own.
@@Navhkrin I think it means it is less optimal to get the details on a mesh using normal maps than using polygons. Not too sure because I don't know a lot about nanite
@@Navhkrin it is true if you compare nanite and lod, nanite is great with asset with a lot of polygon but if your level is filled with low poly asset, the cost of the nanite system wont cover the benefit from it, so you'll better stick to lod.
the only thing i would personally change here is instead of using linear color for the ORM texture, is id switch the compression settings to Masks instead of default(which will also uncheck sRGB for you)
this workflow is mainly for hero assets, could you make a more advanced video for multiple assets under texel density and combining as much asset in the same UV "maximize UV Space", and how studios goes about that matter? thank you!
Great video! But I got a bit lost in the UVs section. You mentioned that the UVs (when they are going to be worked in substance) should be sized relative to the model and keep stretching uniform because it might deform the texture. Yet in the video you were stretching the UVs. You mean that I can stretch the UVs to cover more area, but ONLY if I do this with all UVs that are properly laid out at the same time?
Sure thing, I probably should have worded that differently. You can stretch the UVs, but if you do, you have to stretch them all at the same time. Don't stretch one island by itself. As long as you stretch everything together, you'll be ok :)
Everything you said was correct, but you missed out a lot of important stuff as well. Triangles in low poly meshes (equal tris vs long and thin tris), draw calls from UV islands, texture map resolution and TD in accordance with scale of the mesh, certain texture maps are a lot heavier than others, what file type to use for texture maps and their pros/cons (jpeg, png, psd, etc.), deleting invisible geometry in low poly meshes, atlas textures and trim sheets... I know your whole selling point are short and sweet videos but game optimization is wayyyy more broad than just the stuff you talked about here. Great video tho, keep em coming :D
hey man ! i'm really looking to go more in depth in term of optimazing my per, where can i get information from what you are talking about your comment ? already know about trim sheet, texture resolution, TD, and invisible geometry but everything else is confusing for me. i use targa as image type and im not to sure if it is a good solution
I defently gotta take a break from learning gamedev sometime, to get back into learning ue5 for environment creation again at some point. Especially now that I have upgraded to a newer laptop that has a proper modernized GPU now. Any plans on making enviormtental tutorials down the road royal?
I believe it still helps because it reduces the overall size of the assets size of the game. There’s a very big file size difference if you have 10 levels of 10 million polygon environments, and 10 levels of a couple thousand polygons -
It took me waaaaay too long to figure out what you were monitoring as you were talking. FWIW, the reason images are expensive is the GPU has to calculate for each pixel on the screen where in the texture it is. I.e., it has to do the UV Map work. And that's expensive. And if you're doing three numbers from the same image, they'll be next to each other in memory which means you can suck them over the wires from the memory to the processor element all three at the same time.
is that before using LOD(level of details) UE5 system? or just a separated way for optimization? I know that with LOD the game is bigger in size and it can be tricky because the asset has multiple version that get loaded and unloaded at run time with logic and that can create lag if not well done, but it's usually good for big open worlds game.
Out of curiosity, is there a way you prefer to “make” the low poly mesh? In this example it seemed like just making the asset afterwards, but, is there any merit to starting with a low poly mesh, duplicating, and then working with the high poly mesh? Thinking about sculpting workflows and characters, and I’ve been avoiding the topic (somehow) and could use some direction.
It depends. Some people who have practiced enough will make the low poly version of a character head with the correct topology and then duplicate that and up-res it and modify it for the high poly version. This saves time from having to retopo if done well. But if you are exploring shapes for your character you’ll probably want to just sculpt and retopo a low poly version after. Anytime I have a solid idea of what I’m making (if it’s a prop) I’ll make the low poly first, but it’s not always practical. Sometimes you need to adjust sizing or placement of more complex models before you know where the low poly would line up. For characters though, don’t worry too much about low poly accuracy first, it can slow you down, my favorite workflow for that is sculpt and using the move brush to get dimensions right, and then retopologizing after, it can be incredibly time consuming to try to box model complex characters.
For hard surface modeling in blender, people like the keep all the modifiers (e.g. higher segmented bevel, subsurface, any booleans/cuts) so work is nondestructive. When the model is ready, you duplicate it twice, once for lowpoly once for high poly. For low poly, apply all the modifiers you need/ remove what you don't need for the shape or adjust( for instance any beveling can have its segments reduced). For high poly, you apply all the modifiers. This should give you a base for both versions for any clean up you might need to do for normals/shading/uvs.
Bro where did you learn ts. Seriously, where do you even begin to learn this. Did you take a course, or just years of practice? Like the ORM setup, all of this is just blowing my mind.
Luckily, if you get the Steam version it's a 1 time deal and you get the license for perpetually for the rest of your life, so the steam version is the one I recommend!
And, for anyone wondering, yes, this same process can be applied to more complex objects as well. You'd be surprised how high-res you can make a low poly asset look with the right baking and textures :)
The full intro to UE5 playlist can be found "Completely FREE" on TH-cam here:
th-cam.com/play/PLZpDYt0cyiuu1ZKrZwxOWTRjTUDXMmGaX.html
8:35 This is probably the most important bit of information to fully optimize textures. For anyone interested, look for texel density, you can't even imagine how much performance are lost for 4k texture or 8k texture on miniscule props that will never be seen upclose. For example, in Gear 5 for walls and static meshes they used 400cm=2048px (so a 2k texture will cover a square 4mx4m)
Dude, thank you for videos like this. I feel like there's not enough content about how to approach game design from a standpoint of optimizing, procedures on file management and structure, etc. It's not a "sexy" topic, but it's so important and I appreciate you for it.
finally found the good area of youtube again
Thank you so much for this video it's insanely helpful! I know in blender you can use the multi-resolution modifier to bake high poly details to low poly objects as well but I'm not sure about texturing in that work flow.
Great turorial! Thank you for sharing your experience!
A few months ago i upgraded from a few years in blender, to Maya, and man do i absolutely LOVE it a lot more for anything!! it all just feels a lot more legit and is just better, sadly there are a LOT less tutorials like this because it isn't free... but it is the standard in the industry. hopefully you can do tutorials with maya and unreal soon man! keep it up i love all that you do!
Generally, this is a good explanation. If I can only add notes from the technical art side, maybe it would be better to model only one base cube, texturize it in substance, and make a color mask. One master material is created, and then only the base color is changed in the instances. That's how we solved all the combinations with just one base color texture. Also, this way of modeling is not so useful when making a Tetris game, I'm talking about how the programmer will turn off the entire line when the blocks are lined up if they are all combined in the mesh. A better way is to make only one block, and then combine it in the blueprint.
Awesome work - fantastic explanation of the pipeline - this pretty much has it all (I personally would have also added collision into the mix but that's just me ofc) but this will be my go to vid when explaining to peeps what the stages are.
Great video, but there are a few notes.
If your mesh is using Nanite, then having fewer polygons is actually LESS optimal than having more of them, especially if the mesh is enough to be used as an occluder for other meshes.
Also, you can reduce the Normal maps as half of the Albedo and ORM a quarter. This is a rule of thumb, not necessarily true for all models, but it's also what Epic suggests to reduce texture footprint without losing visual quality.
also you can use RG for Normal map and B channel for a mask.
How the heck having less lolygons is less optimal? I call bs
@@Navhkrin less polys is better in general. but when a model gets convert to a nanite mesh (whats been done in UE5) a very low poly count might interfere with the workings of it.
I dont know the full ins and outs of nanite myself, but calling someone's comment bs without giving a counter argument is bs on its own.
@@Navhkrin I think it means it is less optimal to get the details on a mesh using normal maps than using polygons. Not too sure because I don't know a lot about nanite
@@Navhkrin it is true if you compare nanite and lod, nanite is great with asset with a lot of polygon but if your level is filled with low poly asset, the cost of the nanite system wont cover the benefit from it, so you'll better stick to lod.
Outstanding video!
the only thing i would personally change here is instead of using linear color for the ORM texture, is id switch the compression settings to Masks instead of default(which will also uncheck sRGB for you)
Thanks for the tip bro!
this workflow is mainly for hero assets, could you make a more advanced video for multiple assets under texel density and combining as much asset in the same UV "maximize UV Space", and how studios goes about that matter?
thank you!
Thanks.
Thank you!
Algo Comment, game changing
you da mannnnnn
great video breakdown :)
Great video! But I got a bit lost in the UVs section.
You mentioned that the UVs (when they are going to be worked in substance) should be sized relative to the model and keep stretching uniform because it might deform the texture.
Yet in the video you were stretching the UVs.
You mean that I can stretch the UVs to cover more area, but ONLY if I do this with all UVs that are properly laid out at the same time?
Sure thing, I probably should have worded that differently. You can stretch the UVs, but if you do, you have to stretch them all at the same time. Don't stretch one island by itself. As long as you stretch everything together, you'll be ok :)
Good shit👍
Sub 10 min to say what my last 100+ hour course said. Thank you as always
Everything you said was correct, but you missed out a lot of important stuff as well.
Triangles in low poly meshes (equal tris vs long and thin tris), draw calls from UV islands, texture map resolution and TD in accordance with scale of the mesh, certain texture maps are a lot heavier than others, what file type to use for texture maps and their pros/cons (jpeg, png, psd, etc.), deleting invisible geometry in low poly meshes, atlas textures and trim sheets...
I know your whole selling point are short and sweet videos but game optimization is wayyyy more broad than just the stuff you talked about here.
Great video tho, keep em coming :D
hey man ! i'm really looking to go more in depth in term of optimazing my per, where can i get information from what you are talking about your comment ? already know about trim sheet, texture resolution, TD, and invisible geometry but everything else is confusing for me. i use targa as image type and im not to sure if it is a good solution
I defently gotta take a break from learning gamedev sometime, to get back into learning ue5 for environment creation again at some point. Especially now that I have upgraded to a newer laptop that has a proper modernized GPU now. Any plans on making enviormtental tutorials down the road royal?
Do you still need to care about polys in UE5 if you use nanite for assets?
I believe it still helps because it reduces the overall size of the assets size of the game. There’s a very big file size difference if you have 10 levels of 10 million polygon environments, and 10 levels of a couple thousand polygons -
@@TheRoyalSkies agreed!
It took me waaaaay too long to figure out what you were monitoring as you were talking.
FWIW, the reason images are expensive is the GPU has to calculate for each pixel on the screen where in the texture it is. I.e., it has to do the UV Map work. And that's expensive. And if you're doing three numbers from the same image, they'll be next to each other in memory which means you can suck them over the wires from the memory to the processor element all three at the same time.
Look up face-weighted normals to fix those shading issues he's got going on.
is that before using LOD(level of details) UE5 system? or just a separated way for optimization? I know that with LOD the game is bigger in size and it can be tricky because the asset has multiple version that get loaded and unloaded at run time with logic and that can create lag if not well done, but it's usually good for big open worlds game.
Out of curiosity, is there a way you prefer to “make” the low poly mesh? In this example it seemed like just making the asset afterwards, but, is there any merit to starting with a low poly mesh, duplicating, and then working with the high poly mesh? Thinking about sculpting workflows and characters, and I’ve been avoiding the topic (somehow) and could use some direction.
It depends. Some people who have practiced enough will make the low poly version of a character head with the correct topology and then duplicate that and up-res it and modify it for the high poly version. This saves time from having to retopo if done well. But if you are exploring shapes for your character you’ll probably want to just sculpt and retopo a low poly version after. Anytime I have a solid idea of what I’m making (if it’s a prop) I’ll make the low poly first, but it’s not always practical. Sometimes you need to adjust sizing or placement of more complex models before you know where the low poly would line up. For characters though, don’t worry too much about low poly accuracy first, it can slow you down, my favorite workflow for that is sculpt and using the move brush to get dimensions right, and then retopologizing after, it can be incredibly time consuming to try to box model complex characters.
You can do it either way, as long as the shape dimensions are the same it should be good -
For hard surface modeling in blender, people like the keep all the modifiers (e.g. higher segmented bevel, subsurface, any booleans/cuts) so work is nondestructive. When the model is ready, you duplicate it twice, once for lowpoly once for high poly. For low poly, apply all the modifiers you need/ remove what you don't need for the shape or adjust( for instance any beveling can have its segments reduced). For high poly, you apply all the modifiers.
This should give you a base for both versions for any clean up you might need to do for normals/shading/uvs.
Bro where did you learn ts. Seriously, where do you even begin to learn this. Did you take a course, or just years of practice?
Like the ORM setup, all of this is just blowing my mind.
The promise to like your videeo even i did not need the infos in the video
Thanks bro - That really means a lot to me, hope you have a Fantastic Day-!!
@@TheRoyalSkies the thanks for you bro 🔥
This did not feel like a 9 min video
can we not just let FBX die? :C
Just put a bevel modifier + weighted normal on a cube and duplicate this cube its 10x better, easier, better looking
Another guide for Adobe Software 😫 50$/month 💩
Luckily, if you get the Steam version it's a 1 time deal and you get the license for perpetually for the rest of your life, so the steam version is the one I recommend!