Resistance 3 Generator
polycounter lvl 12
Happy Monday!!! (Tuesday)
New week New project!
Just wrapped up the model for my Bad Ju Ju
So i am kicking off another High Res prop
I found this Generator Concept from Resistance 3 that I thought was cool!
Here is where I am at currently
Let me know what yall think
New week New project!
Just wrapped up the model for my Bad Ju Ju
So i am kicking off another High Res prop
I found this Generator Concept from Resistance 3 that I thought was cool!
Here is where I am at currently
Let me know what yall think
Replies
really focusing on getting the cylinder proportion correct
Thanks man!
Today I have started fleshing out some of those high poly details
Really like this clay render lol
Let me know what yall think
Thanks man! I was thinking the same thing hahah!
Working on that now
Quick question though....I am trying to nail down my final in game polycount for the model....i have heard it needs to be 1-2k Max 5k and then ive heard it can go up to 10K
what should be the polycount goal?
I walk around games like destiny and doom and I feel the polycounts on environmental assets are in the 5-10 range
- Memory: Denser meshes are more costly in memory. Vertex count matters, UV splits and hard normals will increase this, so always be careful about it. Memory is always an issue even on next gen production (streaming of world chunks in open world for instance). It will also affect disc size, although this is usually less of an issue. An easy way to be efficient in your case is to separate in modules and re-use. Don't make the whole mesh one piece, but rather divide in pieces you can re-use.
- GPU time: Denser meshes are more expensive on the GPU. Tiny triangles on the same pixel will confuse the GPU so to speak, it will have to render multiple times the pixel (look up overshading GPU to see more). The big solution used in next gen production to improve that point is to use LOD. A lot of big production now have Simplygon integrated by default to automatically create the LOD. It's not always perfect, but it helps performance and with some love can be relatively discreet.
Of course, none of that gives you a number, I know. ;D
In general, I would say the rule is to use as little as needed. Keep in mind the context of the object. Is this a first person game where you'll go right next to the object? A third person game where you can zoom next to the object? More of a RTS game where you don't need as much details due to the top view? Is this a hero prop that will define a scene and be a major landmark or more of a generic gameplay element?
Basically you want to make the object look good for what it's used for, but make it as cheap as you can at the same time. Next gen productions can go way higher than 10k on some big elements in the environment, but again it all depends on the context. I would just advise putting the object in a game engine and going around it with a first person/third person view. If there's obvious issue (harsh curve for instance), you can probably add a bit more. On the opposite if you realize you barely see something which has tons of triangles, you might want to simplify that.
Porfolio work in general is always tricky. It's not at all representative of production work for obvious reasons. You don't want to go crazy and show no consideration for optimization, but you also don't want to have something looking blocky that no one will find amazing. Don't be afraid to add some information on the context you thought of for the piece. You can also make different LODs yourself, though it's not often done.
I realize it's a lot of words and no definitive answer in the end! I hope I didn't make things worse... Just don't go crazy on portfolio piece's vertex count and know that production work might have been different, that should be alright. ^^
wow thank you for the detailed information i really appreciate that man
I have decided to do 2 versions....I used to work at a really cool studio as QA and made a lot of cool friends and they have urged me to make it 2k so Im hitting that first....but I also will have a version that is 9k
But check out the results for the bake on the 2k model
Check it out
nice modeling.
@killnpc
Thanks man
@BucketOfNuggets
Thanks man yeah I need to go back and clean up the Normal Map, I initially used Xnormal for my bakes then it was being weird about my AO bakes so I did a test bake in substance and was able to pull a lot more detail out of my high poly, so currently I am using Substance
Final Poly Count 2,018
I'm unsure how that would help? Creating instances of different materials and then blending them together with a mask, wouldn't that increase draw calls since you are calling multiple shaders?
@BucketOfNuggets
Thanks man I was actually thinking that as well, contemplating doing some logo exploration with it to see if i can get anything to fit
ah yeah...I can see how that would be better...I could see about retooling my textures to work in this format....shouldn't be too hard
I am using substance painter....but exporting the materials should be to hard
So i know I said i was done....but i refined some of the textures properties and redid the lighting set up...i think it looks tons better...but let me know what yall think
Can u please share high poly to low poly workflow ?
Did u redid low poly from 0 ?
Did u remove highpoly details till u got 9k ?
Also how did u fix these issues mentioned before?
Thanks man I appreciate it!
The final mesh has a polycount of 2,000 and not 9k
I decided to drop it as low as i could go
-So here is my High Poly Work Flow-
1. Create mesh make sure proportions are good
super Important, on this asset i went back and forth so many times trying to nail the proportions
2. Start going through the asset and Turbosmooth
When the proportions match up then I usually start moving part by part through the asset making sure that I place edge loops where they are needed.
One trick that I do at this step that I have found produces really good results is I will focus on my smoothing groups and then I will turbosmooth with a power of 1 and with smoothing groups enabled and that helps me get my initial crease without an edge loop. Then i will turbosmooth again but this time with a power of 2
3. Duplicate Mesh and Disable smoothing groups (I usually put it on a separate layer)
So now that i have a super pretty model, I will duplicate the Highpoly and place it on another layer and Hide the Highpoly. The model that is to be my lowpoly I will delete smoother groups from my stack.
4. Hide Highpoly and start removing unneeded edge loops
These next 2 steps I hop around with....
But now i have a mesh with some holding edges and a lot of inner detail. I start by identifying the holding edges and deleting them
5. Delete inner detail while keeping silhouette intact
Also If there is mesh detail that does not affect your silhouette I delete it
6. Reach Desired polycount UV
This can be the tricky part because you need to look at your edge flow and identify how you can reduce it while making sure to not create shading errors doing so.
7. Export to Bake
I initially started baking in Xnormal....but I switched over when i was having issues with my AO bake
I saw a post on here a while back showing how Xnormal produced cleaning bakes than, Max/May and Substance.....
When the issues arose i tried various things to attempt to get a better bake....the results were inconsistent
I imported the asset into substance and the bake I got was really good, it grabbed detail that was lost in the Xnormal bake
There were some issues with the bake, but nothing that going through your bake and painting them out couldn't solve
Those weird issues you see in the render are 2 things.....
1. marmoset not having the correct settings for how to interpret Normals
2. I had some rendering artifacts from the initial bake/new Bake I solved these by painting out the errors.
I hope this helps, if you have any more questions please don't hesitate to ask.
sorry for the delayed reply no i didn't....i'm not sure how to do that in marmoset....ive used that technique in Unreal...creating shaders and then using a mask to put them in the right places....i can show that if you like?
@Tomiajayi
thanks man...yeah iterations is how you get to a better asset it was really fun working on this
Cool...so for this project (Still a WIP) I created a bunch of material instances like this metal i built in some parameters that will allow me to change certain aspects of the metal so that it can feel like the same metal but not look like I copy and pasted the same thing. so things like how many time its repeating and color, you can also build dirt and grim masks. On some of my other assets i have a sand material that has a mask that is controlled by wind direction that gets information from a wind direction node in the level and places sand accordingly
this one is for a base metal that i can drop into any material graph like below, as you can see this is made up of 4 different materials with the master Mask in the back sitting under concrete
this is what it looks like in engine, as you can see i created a material instance so that i can change all those parameters in the original materials (the metal) without having to go digging through material trees and do it on the fly.
These 3 Cars have the exact same shader and model. The sand shader powered by wind direction is why you see sand on areas directly facing the wind direction
you can also see it working on the hose that is at the bottom of the image