#avatar-optimization
1 messages ¡ Page 48 of 1
I'd recommend using the new blender
Considering it's about to switch ui theres no reason to learn old blender ui
doesnt the new blender cuck up cat's or did i hear wrong?
OK
what do these mean ;-;
Just follow the tips above
is this stuff safe?
yes
the only assets youll find that are either not safe or are not good are the ones from the site that ends in -mods and anything from any unscrupulous people (read: crashers) you meet along the way
ok thanks
unity and blender are popular tools in all sorts of areas of game development and 3d renders from professional to indie, the sdk comes straight from the devs and cat's is a collaboration between community members and a tool we ultimately do not deserve lmfao
I don't even do mmd stuff but it's just got some neat stuff in it for regular ass creators like me
SOo i made a i copy of the tips you gave
oh no my game crashed đ
why do i not want to go past that version of unity?
The game uses 2017.4.28f1, and content built for newer versions cannot be loaded in older versions.
so i cant use this one?
That one is too old
Unity 5.6 comes before Unity 2017
And it's too old at this point, the game no longer uses it
how have i managed that?
It used to be the correct version maybe over a year ago perhaps you followed a very old tutorial
hurray i got it
Just so you guys know, cats does indeed work on blender 2.80
Watch a tutorial
And if you need help with normal modelling please use the #avatars-2-general channel
this channel is for optimization related questions and help
I have a material that I have never touched and it has always looked the same in blender and exported to Unity. It's very light pink, but the texture is darker. I can't figure out what makes it lighter, and when I bake it to a texture atlas, it become darker like the base texture.
what is the purpose of weight painting?
it makes you move
yes
So I'm trying to bake a texture atlas in 2.8 and it's a nightmare
I know nothing about cycles materials and I'm getting ridiculous amounts of shine on my textures
I don't have any light sources in the scene so I'm stumped on this
anyone know where this light could be coming from?
It might be something related to your baking options, i personally don't use cycles and neither do i have blender 2.8, so i sadly can't help, but i assume it's literally just the baking settings
Apparently the background gives off light in cycles
I'm trying a new bake after changing something with that
this is such a mess lol
I want to go back to 2.7 but I've already done the entire avatar in 2.8
(It wasn't the background)
I mean, if it isn't a backwards compatible blend format, you could just export an FBX and import it onto 2.7, if it's already rigged n' skinned there's a setting i'd recommend using on import as it correctly orients bones, if you made any shapekeys they'll be out of order, functional but out of order
it just closes on trying to open the blend file
and I'm really not in the mood to fuck up my model again
this is really ticking me off though because I'm only having issues baking the texture atlas
everything else is great
You could export the fbx then, the blend could be kept for 2.8 if it doesn't work out
Sometimes the weirdest solutions... just work
I'm also running into transparency issues where transparency shows as black
and the typical solution for that isn't... there anymore
you could set the baking type to diffuse only since combined pretty much bakes the rendered view
Actually right click eye bone and see what they are called
@dense crane I'm guessing it's Hi_L and Hi_R
thats shoulders
But yeah select one of the eye bones
Right click one of the eye octahedrons and it will say the bone name in lower left corner of 3D view
There are no eye bones đŚ
Hi is the highlights of her eyes
I got the eye to work on the poppi alpha model
?
There called Fc eye L and fc eye R but not show up on the options
oh and cats says it cant be fixed
Iâm pretty sure there are eye bones there can you take a pic from the side
I can try
It also says this
I have one character set up already it doesnt matter for now
does anyone give me some normal character bone please?
which of these would be lowlid?
Basis, you want the lowerlid to do nothing most of the time
what about vismes ?
Probably the ones that have 'mouth' in em. Those names are dogshit so just find ones that look like the character is saying AH, CH, and OH
There are ones that say mouth specifically
Yeah but the names don't tell me what they do so it's best to check them manually
how
Having the mesh selected, go to the right under the mesh menu which looks like an upside down white equilateral triangle and there should be a menu called blendshapes or shapekeys or something of the like
You can then drag specific shapekeys to 100 to see what they do to the mesh
It's in that black horizontal bar that starts with the camera on the left
No other shapekeys change the mouth?
smile, shout and frown.
If you can use some to mimic AH OH or CH then that's good, otherwise you'd need to make a new shapekey
I mean the lip sync is bad but its something.
Some movement is better than none
Yep
post pics of nice texture atlasing
(yes, the eye is about to recieve a number of important edits, dont yell at me >~<)
to be fair it was made by a professional
but like... its so pretty. and efficient. and compact.
makes me wonder what the world looks like
world atlasing is probably more interesting
because im trying to figure out what this world even looks like
its looks like this
I'm glad I did it, the world was supposed to be quest compatible and that helped me archive that.
big pill is watching
@pure bridge you should bake your atlas two times
So that the black parts are transparent instead
Are we still posting atlasses?
This was an atlas for an entire map,
And heres the Junkyard so far. empty space for future stuff
https://i.imgur.com/t7la3Ri.gifv (this map)
Oh i like that one. Colorfull
glorious
Lol my atlases are far messier than that
I always use 4k and use large margins, so I'm let with lots of empty space like
I don't have the patience to align it all
that indeed is huge empty spaces 
you should fill that alpha in, if its unnecessary. less alpha = better
@graceful vine fill it in with what, and more importantly, why?
Black, because alpha is expensive (runtime and file size)
in reality, its like 0 difference honestly but every little helps
You can blur out the contents to ensure proper color surrounds islands - helps for mipmaps
the texture already has some bleeding on it
@graceful vine file size is the same if I fill in the alpha
Unless I strip the alpha from the entire texture
But some bits are transparent, if they were opaque they would be black to begin with as the texture is baked, not stitched
I don't gain anything from filling in "empty" bits with black
Yeah, i figured that would be the case. Thats why i said if unnecessary
It's a common one used for mobile though, the bits that need alpha are kept separate to save big on filesize
Last time i ran tests filling in alpha vs alpha on png change the file size
What software are you using?
It might be photoshop changing the colors on 0-alpha pixels thus making compression less or more effective
But yeah the file size change is because of png compression
Compression settings between the two are the same
Do those pngs have different file sizes when crunched in unity?
If you don't I'll probably check when I get home as it's interesting
i will do a checkings
All of my atlased have alpha so I'm curious how much size I could cut if any
Ive just always put black where i dont need alpha because it seems right
Black = nothing, alpha = nothing
I've kept alpha everywhere as some parts need it and blender puts alpha everywhere on the second bake
All the Warframe's atlases come with the alpha spaced filled with dead texture as well instead of black or alpha
Unity is giving me a headache right now, if you test this yourself let me know what the results are
I'll be heading home in a bit so I'll just download your images and let you know
@graceful vine at the default RBG Compressed DXT1 they are both the same size of 170.7KB but if you crunch them with RGBA Crunched DXT5 the one with the unnecessary alpha is 42.3KB and the other one is 26.2KB.
If you change it from the Mitchell algorithm to bilinear it goes to 39.5KB vs 26.3KB.
Thats still something right?
Itll make a bigger difference on a larger texture, the test one was only 512x512
So if you had a 4k atlas with a bunch of unneeded alpha it could bloat it when you try to crunch
@surreal topaz Incase you are still interested in the black vs alpha thing, read above
We out here optimising like a mobile game
I think that a texture with a black background simply ignores/removes alpha (depends on exporting application as you can export a 24bit png which in essence is a png without alpha), although you can still tell a shader to cutout or remove any specific color from it, the reason, i assume, that DXT1 is seemingly the same size is due to either being opaque or being the 1bit alpha variant DXT1A which is identical to the former, but as the name entails, only uses a single bit of alpha (either fully opaque or fully transparent) making it rather effective and very lightweight but it also comes at the cost of quality compared to its other counterparts, DXT2 through 7 respectively, (although iirc unity uses different compression algorithms than your standard DDS file so that may not apply), rather unsure of how unity crunch compresses files so i can't say all that much about it, but generally a 24bit png is lighter than a 32bit png, due to not having an alpha channel, but that again comes with a drawback of either being fully opaque (gradient transparency is quite literally void) or having full alpha support (even if you fill most of it with black, if any transparency exists, it's still a 32bit png).
So it comes down to, keeping 2 mats, one fully opaque atlas, the other a transparency supporting atlas, or a single massive atlas that has full transparency support, or fully opaque at the loss of general alpha quality via cutout
Although i assume the more alpha needs to be accounted for, the larger the filesize, that may vary depending on image complexity however (as i think the location of the alpha does influence the filesize to an extent but i've never tested it myself)
You could of course use a 8bit png, if you're aiming for the most optimization, which instead of the 16million (16,777,216 to be precise) colors that both 24bit and 32bit support, you have 256 colors which is akin to a gif, being a palette variant, it doesn't support alpha either however, but being like such you can far more accurately cutout a color as you only have an index of 256 colors, which is one of the reasons that animated pixel art in gif format can be so easily edited or extracted (the act of removing a background or foreground but keep something, such as a character), so they all have their benefits really
Unity decides whether to import as RGB or RGBA but you can forcibly disable the alpha channel
Importing DDS files directly is very wonky and should not be done
Ok im a complete dummy when it comes to trying to lessen down the texture slots, becouse an MMD model ive used turns 5 png files into 30 material files. But the problem comes when unity's texture limit. I was wondering how one removes those unused materials.
the annoying bit is that all of the materials, are in a prefab.
And i then remembered that CATS exist and saves the day again.
You can cut down on materials by atlasing your models
Mmd models typically have more than one material for each texture for ease of editing and sharing, also because there are differences between the materials that arenât always apparent at first glance
There is also âcombine same materialsâ which, if two different materials have the same diffuse texture, it will delete all but one and use that
You have to be careful because it can kill cutout materials but itâs best to atlas down to a single material then duplicate whatâs been broken so all materials use the same single texture
You can always change shaders on the duplicates material but a single texture to download is still better
Am I the only one that an avatar with crunch compressed textures/particles/animations can take up to 10-20 minutes to upload? Have an avatar that's 10mb small, yet takes like 15-20min to upload.. I'm 100/100
wow..
it actually did turn on
will check the log too though
It should only be established once security is tight enough to prevent non authorised users from accessing it, aside being incredibly slow to build & publish with future proofing enabled, it's a free ticket for anybody with the right know-how to rip assets
Let's talk audio
Raw, bigger wavs
Or
Compressed oggs
Besides load times, what are advantages of ogg?
Don't hear a huge difference in sound quality, I've seen from 10mb down to 300~ kb
doesnt unity auto compress it anyways
WAV's get compressed to either OGG or MP3, i personally just use mp3 files directly
"Depending on the target, Unity will encode the audio to either OggVorbis(Mac/PC/Consoles) or MP3(Mobile platforms). For the best sound quality, supply the audio in an uncompressed format such as WAV or AIFF (containing PCM data) and letUnity do the encoding."
Further information on supported audio formats https://docs.unity3d.com/462/Documentation/Manual/AudioFiles.html
Develop once, publish everywhere! Unity is the ultimate tool for video game development, architectural visualizations, and interactive media installations - publish to the web, Windows, OS X, Wii, Xbox 360, and iPhone with many more platforms to come.
so when setting up dynamic bones what is the collider bit for?
@dense crane you set a collider script on your wrist (and give it a height setting to envelop your hand), and assign that collider in the section (for example on your hair's dynamic bones).
Be aware that colliders are a larger performance draw than just using dyn bones. Try to limit bone chains as best you can while using them.
If you ask what it is, just ignore it. That's the best for all sides 
Only in blender by selecting both objects and pressing Ctrl+J
What would perform better for particles - 8-15k opaque mesh triangles or cutout billboards
cutout still uses alpha test, so you'll incur a cost for that
oh, I could just use unity profiler for that, but yeah - was thinking how big of a difference there is with opaque triangles vs alpha boards
So, today I tried to upload my first avatar and I thik I got most stuff right, but his face keeps "smashed" when I'm not moving, does anyone have any tips? (Also I'm new here and idk if this is the right chat for these questions)
Try posting a gif of your issue in #avatars-2-general
that's just a weight paint issue
Find a tutorial that explains what it is, and how to fix it, and you'll be good to go
Cutout layers are always drawn even if thereâs something in front of it
Actually disregard that I was thinking of transparency
Is there a button in blender to draw the current materials onto a texture
I feel so close yet every one hour render has failed me
Figured it out! two uv maps, the target generated atlas map in every material, uncheck clear image, and bake diffuse color only
Rendering in blender isnât for texturing, by the way. Itâs for showcasing~
Well, most of the time it is
Well I mean, itâs still only a release candidate
@heady smelt it's not
Did it get released?
Yes, 2h ago
Osht
@surreal topaz it's not really a disaster, just a new thing to learn
Yeah, luckily blender is still open-source so overtime things will improve
And rather quickly compared to other programs like maya for example
And who knows, new method may end up being better after Hotox pulls the magic rabbit out of their hat
2hrs have passed so theyâll appear eventually~
It's pretty difficult to adopt cats to the new nodes since everyone can set them up completely different
yeah but in vrchats circumstance you could just assume toon/unlit just throw it through emission and it'd be no different from what you currently have right? (well assuming it isn't designed for rendering)
only issue would be potentially transparency
i suppose you could replicate popular* shaders within vrchat but at that point they would be better off just working on it within unity unless you're texturing from within blender
People are still gonna be using 2.79 4 years from now, no rush
pretty much
I know i am đ
really if people are using 2.8 they're more than likely capable enough to work inbetween 2.79a and 2.8 if necessary
@proper grail @surreal topaz @sudden zodiac You can paint from one UV set to another using the texture painting tools. You can see it being used here to do projection painting. https://www.youtube.com/watch?v=gz4qV5_zuMo
SUPPORT THESE TUTORIALS https://www.patreon.com/cg_matter MY WEBSITE https://www.cgmatter.com PART2: https://youtu.be/s_x62z1T9tk In this blender 2.8 tutoria...
The basic technique is to create two UV sets, then use the paint tool in single image mode to clone from the source image to the destination.
While you can use it for atlases, it can also be used for any kind of UV manipulation.
Like manually rearranging UVs or making the left and right halves of mirrored model unique again.
Interesting feature, but not really what I was looking for
This is the end result I wanted, combining a bunch of smaller UVs into one
select each one and move them manually
Oh no its super easy in 2.8. Can make a second UV map. Select all the materials you want on it. in edit mode use uv -> average and uv -> pack. Throw the same generated texture in each material. bake diffuse color
You don't want to do that if you have multiple parts of the mesh sharing the same UVs
It saves texture space to stack geometry that uses identical textures.
Using those commands will unstack them.
Also averaging them may cause you to close detail if the original UVs weren't already uniformly distributed.
It's not uncommon to make the face UVs larger than other parts of the mesh so they can have more details.
oh shit there actually is an avatar MB size limit? fuck dude
ya what is this why does it exist please remove it
it exists because you dont want avatars to spend a long time loading, have a significant footprint on your network, and to not take a long time for your PC hardware to work with, as well as just general space (such as on your RAM)
in other words, removing it is an awful idea
I'm guessing this is a Quest upload ?
most likely, 10mb seems too small for PC
reduce texture size and number, my first guess to a high number like that
wild. evem the unity build is set to PC? shit dude, i really havent seen that before
what assets are on / a part of your avatar?
both audio sources are very small
whats the actual size of them on your system in MB?
if its not the audio it could be texture size
is there the possibility that the latest sdk may have on accident made it so all uploads are set to quest
idk, thats a unity thing not an sdk thing so prob not
no thats not true at all i was able to upload models before
using an older sdk
i just updated my sdk just now and im now having this problem
Ruuubick is there a possibility that the sdk is not uploading to the PC version of the sdk/unity because its not defined by the project settings of the Unity project itself, but the SDK might be confused and think it's set to Android instead of PC (IgbarVonSquid)
so it appears that the issue is derived from unity thinking that the build platform is android by default the way to resolve this issue is to go to file and click on build settings and then click on pc and switch platform
So I tried to optimize my avatar. So far so good, everything is good, BUT the bounding box.
My mesh is wider than it appears, due to wings being able to unfold, and my tail is really long, when fully stretched out. Okay... so I managed to shrink the bounding box, and still seems like it'd still not cause too many issues for my avatar, but the SDK said it was under 5 x 6 x 5
I load into VRchat, and check the stats... it's still Very Poor, becuase Bounding box is for some reason... at least 28 x 2.0 x 10 and I don't know where it's getting these numbers from! numbers fluxuate at bit, but are around those numbers
What's more, I look at my older avatars, and they're fine bounding box wise, but the ones that I optimized the bones (by reducing bones in half in tail, removing bones in hair/feathers, and removing dynamic bones all over) and those are the ones that are rediculous! and I can't fiigure out why or how!
Ingame stats take into account the "real" bounding box scale, disregarding user input if i recall correctly
No, it uses user input
But if you have multiple objects on a character i.e big ass wings it can use those as a bounding box
I recall users with disabled pens having their bounding box at an absurd scale due to their way of resetting (sending the pen to a very far away location the second it's disabled to keep whatever was drawn intact but stop it from drawing) in essence completely disregarding user input, so does it just go for the higher offender?
It goes for the highest offender yes, pens need big bounding boxes so they dont stop rendering when the user walks away/ out of vision
So it uses the "real" bounding box radius, in which the highest values are what is shown instead of any user input (except perhaps if the user inputs values above what they currently have, but even then i think the system uses the highest values disregarding user input, but i may be incorrect) i recall having a model with broken visemes (of which i knew nothing about at the time) and simply rescaled the bounding box manually, not only was the bounding box stats showing the radius at an absurd radius, but the in-game model itself was being affected by lights around the map even without any viseme/shapekey currently active, so it enforces the highest box radius aswell
As long as you dont have fucking stuff on your avatar like pens, the user input remains true in my experience
Shape keys shouldnt change the bounding box, there had to be something else going on that you were unaware of
if theyre broken they can
How?
Broken shapekeys are known for having extremely offset vertices, so they add up to the radius
like a syllable that fucks up a vert
Hm, so if broken shape keys are present it disregards user input?
Straight to the trashcan as far as i can tell
Unity is a confusing beast
If they go really out of orbit, unity won't even render the model
Giving it like a -infinity box range
ive experienced it before by trying to make emtpy game objects in blender
specifically, my avatar was maybe 50m ahead, 50m above me
and i rotated ariund the thing while it floated in the air facing the ground in neutral pose (not tpose)
funny enough it wasnt visible, but objects i attached directly to it armature were
furthermore theyd bop up and down as if part of the foot
fucking weird
I recall trying to make a static cubic platform that i could stand on, it did not go well to say the least
i just thought of a good way of putting down a reason why optimization is important, but i cant think of any better place to put it than this chat, so enjoy if you come across this:
lets relate optimization for a second to the amount of people you can have in a world
usually this is set by the world author, or constricted by an event host, but lets move away from a hard definition and talk about it in a different way: how many people can you have in a world until you start to lose or drop frames?
oculus already has an answer, and its a policy for staying on the store, because games that cannot meet 80 fps will not be allowed on their store. their personal reccommendation is 1-2 million tris and 500-1,000 drawcalls (meshes + materials) per scene, the scene being each frame, and those components only counting when they are rendered / visible, so disabled game objects dont count!
were not gonna talk about how much these figures have changed nor how this relates to VRC itself, but we can use them as a good starting point
now think about an avatar that is just barely passable according to SDK standards - one that has 69k tris
how many times can you multiply 69k before you fill up 2 million? thats roughly the amount of that avatar you can have in a room before VR users drop frames
Divide 69k by 2million and youl have it.
Hang on
thus, you can also consider an avatar that has twice as many tris - 140k - not as one that is not just heavy and poorly optimized, but one that takes roughly the same amount of work to render as it would for two avatars, or two players. In other words, a 140k tri avatar effectively takes up two player slots rather than one on your PC, and in a way the room capacity has been lowered by one
Roughly 28-29 users
and thats not even considering the tris or drawcalls that go into worlds
or avatars with lower tris
dont forget that in that calculation, you need to factor in the world it self
wich usually should consume the most
plan around numbers like that, and youll usually end up with a an avatar that tends to work well in most settings
im not a world dev, but i usually plan my avatar weight around the idea that most worlds are gonna have 300-400k tris, still not sure about drawcalls
Now correct me if i'm wrong, but isn't parts of the world culled based on the camera orientation?
With a decently rendered world and the 69k tri limit, you'd probably get 11-15 or so in
they can be, and it's good to plan around that too
before major slowdowns
We need to take into account aswell that this is based upon currently in-view users (provided their bounding boxes are proper)
As anybody not in-view is culled and hence not counted towards such (hence disabled objects)
That's why your FPS can suddenly shift when you look between an empty area to a crowded area
exactly. My two cents, i havent developed worlds nor looked at unity's perf thing, only going off of numbers and "napkin" calculations, furthermore this isnt considering that vrc itself is kind of bloated and laggy and sometimes makes its own limits on worlds thru performance because of this
nonetheless this would still apply in the ideal future when the game is optimized and still holds some very real weight now
so for anyone who's optimizing right now id say that this is a good enough method to follow by
All things considered, i always decimated my models to bellow 20k before, untill recently that is as now VRC is (supposedly) using more than a single core
same dude. i hit 60 frames alone, 40 ish with people, and 20-30 in publics
jeez
i always strive to make my shit work as good as possible on as much hardware as possible cuz fuck this dude
I've got 4g ram and integrated graphics off a celeron
this is more on my smart ass anyway, buying an i5 6400 and all
Back in 5.6.3p1 the max polycount was 20k, they expanded on that fairly recently
20k would be a good number to hit for events and high numbers of people per scene, which i think is what they were originally going for
1 million tris worth of 20k avatars is like...50 people? i think...
Ehh, even then the IK was the major hoarder back then
yeah what am i saying of course its 50
I recall a world allowing roughly 60 users and everything was laggy as hell
shit dude community meetup (40+ people per lobby) is laggy as hell lol
Keep in mind the world itself wasn't that complex, at most it was probably even less than 30k at the time
The worlds wouldn't have been as optimised though back in the days of old
It was a motherbase from MGSV recreation that was in development
They sadly dropped the project after a while, unsure why, but generally worlds not being optimized back then isn't really accurate, alot of worlds were and still are, just a few outliers were laggier than others
By far the most optimized thing that must be optimized is worlds, if the world isn't optimized, it doesn't matter how eco-friendly the users in it are, they will all lag
cant wait for this game to get optimized in 1-2 years so the gospel of community optimizarion has more weight than it does now
mhm! optimized worlds = happy users, and more of them
'Aight, I gotta turn off discord and turn on focus for VRChat. I'll be back in a few hours.
kk
I kinda wish avatar LOD's were a thing, that'd bring a massive performance boost, albeit a bit heavier overall
World LOD's are amazing and should be used whenever possible
Especially for very large worlds that don't need detail to be constantly at full resolution
Levels Of Detail
i only remember that from the cursed unity asset
explain if you dont mind...specifically in how it relates to vrc
It loads a lower poly model of a section based upon the distance the user is from the location
dude holy shit thats genius
The transitions might not be seamless, but they sure as hell help performance-wise
yeah i can see that being heavier on systems but not necessarily on frames
id be up for that to some extent for avatars though, especially when screens are still too shitty to see things at distance with
i could see that being more useful for worlds tho
Yeah, more-so heavier on filesize is what i meant, as the lower poly models are still in the file itself
fuck dude you actually just answered a question of mine
about how to het away with really highly detailed ornamentation and other shit
LOD's also cull if you're not looking at them or are at a certain distance
cuz i was walking around in the library of congress and went " what would i have to do to model this," and the thing is the top of it was all highly detailed floating golden flowers
So they're handy in that regard aswell
should just be culling them or unloading them in place of flat tex at distance
holy shit now i want to make a library of vrc with this shit
"dude ur shitty fucking flowers are like 10k tris each and there's 50 of them! how am i not lagging???"
"wizardry and hours of work my guy ;)"
its time.
It's a unity built in feature, and it's also documented enough to grab onto it fairly easily
"Your set of Meshes have file names ending in _LOD and a number ranging from 0 to the total number of LOD levels minus one. For example, if the base name for your Mesh is Player, name your files Player_LOD0, Player_LOD1 and Player_LOD2 to generate a Player GameObject with three LOD levels."
I heard that it's bugged in regards to avatars so i wouldn't try to use them (i think avatars show all 3 LOD's at once instead)
You can have as many LOD's as you want, but keep in mind that the filesize will increase as you do
yeah, thats the thing >x<

I'd recommend 3 LOD'S in general
one thing that kind of irritates me is when a world is pretty but textures loik obviously low detail up close
LOD0 the most detailed
LOD1 mid-range medium detail
LOD2 very low detail
a lot of assets werent really made for you to truly get close to the eye with, and i feel like we'll be seeing more of people getting around that as VR becomes more prominent
is there any way to LOD textures?
Unity also uses mipmapping, essentially loading lower-resolution textures of a model the further you're from it, so you can still have some good looking textures
cuz time and time again i enter a world with massive ass bricks and a normal map to match and the floor looks like fucking soap
That is done automatically by default iirc
ah
LOD's are Mipmaps for Models and viceversa
so then whats stopping me or in this case any other world author from creating a high enough res repeat tex (say like 8k) and using that?
prob that it looks noisy at a distance id assume
Just load-time and generally if you had alot of them it could cause a drop as you got close
ah...
But even then, i doubt it'd be anything crazy, unless you went absolutely crazy
hm... ill probanly think about this more when i get around to makjg worlds
i already have some ideas from what ive learned today, and shit duud i love getting crazy đ
đ
Haha
Only problem is LODs dont work as expected in VR currently
The only downside i can see regarding LOD's is the filesize increase and how models can pop into existence
The distance is massively different compared to that of unity editor
oof...
Ah i recall that, for some reason VR is rendering the lowest detail LOD far sooner than it should right?
Ye making stuff look much worse and stuff dissappearing from very little distance away
Yeah, that really need to be fixed, LOD's are amazing but we don't need world-wiping as a standard xD
interesting

Is there a way to merge my materials together. I want to make my pc avatar quest compatible. It says i have 8 materials and it has to be 4 to 1 materials. Is there a way to do that in unity?
If it's not possible in unity is there a way to bring the unity avatar from unity to blender so i can edit it then bring it back to unity without messing a bunch of things up? So i can just replace the unity model with the new and improved blender model.
if that's not possible then can someone recommend a good tutorial for combining materials and atlas textures in blender please?
This Tutorial will show you how to Merge extract bones that aren't needed for VRchat as well as Textures into an Atlas to reduce draw calls. If everyone did ...
thank you.
Ideally you are supposed to start in blender. Right before exporting as fbx, save a blend file. Go back to this blend file if you ever need to update the model, and after your changes re-export as fbx directly into your unity project, overwriting the fbx with cats export.
If you never started with blender, or forgot to save a blend file, you can import your fbx into blender with cats but this will usually corrupt your prefab when you re-export and require you set up a new gameobject heigharcy in unity, so save the fbx as a blend before export so you never have to do it a third time.
So long as you always go back to the same blend file that originally exported your fbx (and it still has its prefab association to the fbx) you can freely update the model in blender and it wonât ever mess anything up in unity
You may need to set the fbx to ânoneâ and back to humanoid to force the skeleton definition to update if you moved bones around
I find regular blender fbx export (as opposed to cats fbx export) tends to cause problems updating pre-existing models in unity. No clue what cats is doing differently
Does anyone know how to atlas in 2.80? I got use to using the texture-atlas add on that 2.79 offered and now they've taken that away from me
shotariya's mat combiner should also be available for 2.8 as per the latest update
I really gave 2.8 an honest try but could not even figure out how to turn on texture render
Very frustrating feel like a newb again
I feel the same way
I'm forcing myself to learn it and I've actually managed to figure out how to bake a texture in 2.80, but it's still not quite right
however if you want to view your textures on your model through blender:
you can try importing an fbx with embedded textured if CATs doesn't work for automatically assigning textures
(i haven't tried CATs in 2.8 in a long time)
incase you wanna see a more toon like style aka flat
i've just always made it a thing to use blender with out Cats aside from the import function XD. Assigning textures hasn't been much of an issue.
new images for discord aren't loading for me reee
yeah you gotta open the original idk why đŚ
I use catâs pose mode because the rest/pose switcher is the most pointless thing ever
looks like you can see everyone elses images but your own right now
I do remember it having a use I just can't remember what it was for
ÂŻ_(ă)_/ÂŻ
so setting this up for hair, what should i do for only a slight hair bob?
So long as you go back to the blend file things will never mess up in unity
Actually things can mess up, for instance changing material slots may make you have to manually fix materials in unity, but all you need to do to fix that is drag the materials into their proper new slots
i have a blender file saved.
@honest solar Not really. But there is the Unity Cloth Component for cloth-based items (Skirts, scarfs, capes, flags and other basic things). Which may help your situation?
Hello, I don't know how Atlases work exactly, but this one seems to be using much more wasted space compared to ones i've had before, is there anything I have to do about this?
If you're worried about wasted space then the auto tool for CATs is not what you want to use. I would recommend watching Tupper's manual atlasing tutorial as that is the best way to atlas your models.
If you are insistent on using the auto tool, you can at the very least change some space settings in the newest version, but it is not perfect and you will still get some wasted space
Also holy cow I can only guess that black texture in that atlas is a 2k texture. You might want to lower the resolution of any low contrast textures. It will make the high contrast parts looks better as unity shrinks everything to a max of 2k on textures by default
@simple vortex combine through matcombiner panel and change size to 'automatic', panel inside cats forced to 'power of 2'
Alright, thank you all very much
You can save some space on that all-black texture by changing it to a black diffuse textureless material. Material Combiner detects and and bakes diffuse textureless materials down to a tiny spot texture so you save a ton of atlas space
Idk why but the bone merging only affect half of the skirt... Can't get it to merge the left portion of the skirt even though they all share the same parent bone
There seem to be a cap on 12 bones đ¤
I don't believe there's a cap, the naming might be different enough that Cats is getting confused
CATS be using Neitri's tools as the base for merging bones, doing such should merge selection to the appropriate bone parents, there must be an underlying issue regarding it, never heard about it happening before personally, but perhaps restarting blender could suffice, if not, you could also try having 2 blender windows open and copy&paste (ctrl+c and ctrl+v) the model onto the empty blender, and if that doesn't work, you might want to try contacting hotox to explain your situation
@slender saffron
The weird thing is it say it's going to merge Skirt_0_0 (12 Bones) but then 0_0 to 0_11 gets completely ignored, and 0_12 to 0_23 gets merged
Then, say I delete 0_12 to 0_23, the option to merge the skirt doesn't even show up on the bone merging tool
@marsh trail It not merging all bones is caused by the bone select algorithm, not the actual merging process
@slender saffron I already have ideas on how to improve that algorithm, but I don't have the time do implement it because I'm on vacation đ
is this good?
Reduce the number of materials
That's really the only thing it has going against it, but you should consider turning down the poly count too
i cant without the model breaking
You can
Hey does antone know what's going on with material combiner? I've renamed, and moved the files around and yet this error still happens.
I think I've gotten this error once, some material had an mmd texture with an error symbol near it. Remove the texture and it should work
@tiny hawk
Problem is that it's not mmd, and removing the texture doesn't exactly solve the issue at hand does it, considering the whole point of atlasing is combining textures.
http://prntscr.com/opq8po
It doesn't have to be MMD to have the error, check all your materials for this section
The error happens when the Atlas generator is trying to display the image of the material in the list but the material doesn't have a texture image to begin with
removing that won't ruin the material or I don't think it will
I was able to do a fix. Basically I had to convert the files from png, to bmp, then use cats own png converted to get it to work.
Well that's one way to use that button đ
Yeah it's weird, but I think I'm just gonna stick with manually atlasing because the quality comes out a bit better.
@tiny hawk in what way is the quality better?
Basically the ability to decide what needs more upscaling in the atlas more than other materials. Aka the body's part of the image will be larger than the eyes.
@tiny hawk do you really upscale in that case or do you downscale all other textures?
Because you can downscale each individual texture in the material combiner
Maybe upscale is the wrong term for this, but essentially I can control what material has more room dedicated to the image than others, and usually I bake in 4k
Then it's not really higher quality, since you can do the same in the material combiner
How so? I might start using it more often to shorten my build times for avatars. Besides having to fix the whole error thing with the files.
Press the little gear icon next to the material, you can set the size there
@sudden zodiac Anxiety. https://i.imgur.com/czwkAkJ.png
@amber hemlock Oh đ
@sudden zodiac Is there a way to disable the MMD texture panels that Cats adds? I don't see what it's used for and it clutters things up for no reason.
@hoary juniper It is from the mmd_tools plugin that cats uses to import mmd models. You could go into the code and remove some lines to prevent it from loading
Hmmmm, I'll take a look, thanks
Where do I fix the bounding box?
Bounding box may get much bigger if you have some stray verticies far away from your model or be result of something added to your avatar like snail marker which I believe needs to hide the marker by moving it far away from your avatar. If you didn't add anything you may try fixing it in Blender by box selecting entire avatar, hiding it than pressing A to select all, deleting it and unhiding your avatar. Other possible solution would be to check your shapekeys and see whether there aren't any verticies going far of your avatar after using one of the shapes. You can also manually change boundix box size in Unity if everything fails but I read there may be some other problems if you do that.
Why is Join Meshes greyed out?
nvm, solve it
had to import the model with cats or else that would be greyed out
its a hide feature added in new 2.8
it doesnt "have" to be pulled in through cats, but blender takes old blender files, or fbx's with hidden items, and hides them across the board, instead of in the alt h/h menu
you can fix it by clicking one of those drop downs above it, and choosing to show all hide functions, then unclick each until you see which one is actually hiding it
Hey,
Does anyone know a way for Texture atalssing in 2.8 without using Node editor? Currently I have to set nodes in each of original materials so that they bake to textute
I think 2.8 took a big leap and went straight for rendering based on nodes, completely replacing blender internal, although i may be somewhat incorrect as i've personaly had no interest in researching about it much
@trail folio Try Material Combiner, it works in 2.80
anyone wanna help me lower the polys on my 150k poly avatar? it would be very nice to get some help
could try the cats tool for it
you see here. i have it but i rarely use blender and idk how to even like import it into blender nor import the model
ah ha! i need to hit import fbx huh
might be better if you start on a mega starter guide then
yeah haha
@candid sedge hold on, i import cats pulgin with the important tab correct? if so what kind of thing is it if you understand what im saying like a fbx or something else
6 million polys
there were a few of them where i found him,public german map is full of those people
brain: -
question
what would be better
A > one really high res (4-8k) noise texture
B > a really small noise texture (sub 1k px) repeated like 20 times
as in, what would be better for frames / stability?
hopefully this is a general question, such as if said texture could be anything from a normal map to an emission map, masks, etc.
tiled textures count as drawcalls? 0~0
holy shit
i need to like
go and fix a lot of shit
cuz i have a ton of things tiled to like 4-25
Tiled textures? I don't follow
Oh seamless?
when youre using an y normal shader
import like say a hair normal
and go ahead and open up the tile/offset menu and turn up the numbers a lot
I honestly don't know what you mean, likely because I've never used this system before
I'll cut you a better deal you can have 2 seconds
two seconds have passed
this
im asking if it would be better to tile that 256px tex or just to make one really large version of it
I thought you meant something else
yeahh
Smallest is bestest when it comes to basically everything
Hmmmm, yeah, that's not really good advice
Is it better to have a big high res texture full of random noise?
No. But that depends on the noise. Random white noise or blue noise can be small. Repeating patterns should be as small as possible. For noise that isn't meant to be repetitive, larger can be justified, or if it's visibly large on screen - like, cloud patterns, that would make sense.
Is it better to have a really small noise texture, under 1024x1024...
1024x IS large for random noise. In reference to above, 1024x makes sense for random clouds, but if it's just white noise, 256x is the highest I would suggest...
Which is better for performance?
Smaller.
@pure bridge
What exactly is blue noise and how does it differ to any other kind
I see math, immediately scared
White noise is purely random, so at a distance or when blurred it gets clumpy, but blue noise is generated to average out into, well, grey.
It's handy for shader stuff where you want to make something look blurry or soft without doing a blur or softening it đ
So nothing ill have to worry about anytime soon
Am Having Issues trying to atlas
I allready have a atlas that I had generated via CATS but am trying to make a new one Just to try to make my own textures
should i deal with any z-fighting inside a model or is it better to just ignore it if it can't be seen
Technically there shouldn't be any geometry inside the model if you can never see it. :P
quick question. do you find a T pose or half T pose better when making an avatar, which has better shoulder deformation?
A pose vs T pose?
A pose is better for shoulders because it is at the 50% extension of arms at side/arms high up. T pose scruntches the shoulders too much for me
You arenât limited to A or T pose you can actually have ANY reference pose
A pose is better because itâs more of neutral reference. Actually unityâs fetal / âmotorcycleâ pose is the most neutral pose but itâs hard to model that pose so no one does
I didnt know it had a name
I donât know if it has an official one actually!
All tutorials Iâve seen call it motorcycle pose though
Itâs just the result of all humanoid muscles at 50%
Do anyone have a link to a two sided shader?
Is it possible to merge skinned mesh renderers?
@midnight barn ye, just do it in blender, with cat's or manually with ctrl+j
hmm, i don't know what any of that means besides general knowledge of what blender is
this is an avatar i converted years and years ago but i want to optimize it, it's low poly and everything but the only thing that makes it poor performance is number of meshes and materials
cat's blender tools, popular addon for blender specifically for getting avatars ready for unity and vrchat
just take the fbx out of your project and import it normally into blender
select all of the meshes on your avatar and hit ctrl+j and they'll join normally with no conflicts like 99% of the time
as for materials, CAT's has a tool that will do that for you under Optimization>Texture Atlas / Materials (it's called material combiner)
it will combine all of the materials on your avatar into one and retain the quality
Sometimes joining meshes will break the UVâs, mainly if their individual UVâs donât have identical names.
So I always make sure to rename them to the same thing (usually âUVMapâ) before joining
Perhaps Hotox could add an optional auto-rename UV's (much like combine same materials) to the CATS fix menu
I actually think cats does this already. Iâve never had cats mess it up so far
I just run into it when manually joining
having a transparency issue with unity after importing to blender
(color is just another skin, ignore that)
I did try adding a 2nd layer of it,
thinking it was just a problem with it being a single face thick
(didn't make any difference)
Did you try recalculating the normals after adding second mesh layer? It is possible both layers had normals pointing in the same direction. Enable backface culling in Blender to have better representation on what you gonna see in Unity (Press N ->Shading ->Backface Culling).
recalculate normals? alright, i'll try that
and yes backface culling shows me exactly that. much appreciated
Normals where the answer. thank you so much @cerulean holly
@marsh trail @amber hemlock Yes, cats already renames all UV maps to UVMap before joining meshes
Oh nice, i actually didn't know it did that already
Here is what it does: https://i.imgur.com/lxgfNas.png
Since half a year maybe even longer
Also, backface culling is on by default now in cats @jolly marsh , so if you see through it check to see if "separate loose parts" gives you two skirts. Make sure you're in lookdev mode in the future, it's handy
oh sorry this was answered, forgot to pull it down
if you need to create them, select a mesh object, tab for edit mode, a to select all, search for "mesh: duplicate" and then search for "flip normals" and you have something two sided without a shader
Kohi answered this one ^^
yay for quest compatible. Sweet! Slow day at work so I'm chilling here in the meantime if you have any issues with blender/unity
oh i do have some things i'm uncertain of that i want to get clarified before i continue which are over in #avatar-rigging
oh!
I want to call it shader masking but google doesnt come up with anything, Does unity have a shader that can select what shader to use based off a texcoord?
guys how do i remove the Dynamic bone collision check count
Remove dynamic bones
@proper grail Each polygon on a model can have only 1 material, 1material=1shader, its up to that shader to determine what is drawn on screen from given texture inputs and other data that same polygon has
Thats why shaders themselves support masking some of their features
Really wish I could wrap my head around shaders. Seems like everything cool revolves around them
well not everything but you can do some pretty cool things with them
@polar bolt Okay that sorta puts me on the right track towards 1 skinned mesh/1 material. So im looking for a shader that has the most options. Is there one on the vrc shaders off whitelist?
Ill be going through them i am just wondering if one pops up from memory
With Cats material combiner is it possible to combine both a diffuse and a normal at the same time?
@tiny hawk install 2.0.3.3 version of material combiner and use Multicombining through MatCombiner panel for that (don't work with 2.8 blender)
@novel grove thanks, I'll give that a try.
whats a good amount of materials for an optimized avatar?
why does the VRC blogpost tell me 10 is the max to keep optimization?
1 is best
so.... no model at all? 
shush, dont spread my secrets
I also use an exploit to remove my nametag, that's like 4 or 5 materials less, that's what i call optimization
2 l8 
If my model has 1 mesh, 1 material but 5 textures how many draw calls will it have? Is it the same number as for model with 5 different materials with one texture each? VRChat doesn't seem to account that in avatar performance rank.
it can't have 5 textures if it only has 1 material
^
1 material can only have one texture image
well do you consider normal maps and emission textures?
No.
no i mean kohi
They are not textures
well yes, but actually no
Only the Albedo is a texture, the rest is kind of "configuration images" that make the shader render things in a specific way
They are like settings conveyed through a image file
I see, but they do influence performance a bit I guess?
they're no bother
Cool, so normal map is like having extra details for free
Yep, occlusion too.
How dare I forget
Hey guys, can someone help me with gameobject is missing in the animation tab?
@still perch That's misinformation, don't say that
They're all textures to the GPU
However, the number of textures doesn't matter (*), only the material count.
*: Not speaking literally, but the truth is complicated. If you need a hard limit, 8 per material is probably it.
@quiet flame not sure if you still need help, but if you double click the name of the missing item itâll bring up a pathing type of thing
(Armature\Hips\Tail) or something like that. Simply just adjust it so itâs pathing to the correct item, if youâre not sure how exactly it should be structured check another item that isnât in the missing state and copy itâs pathing then change the directory to the item youâre looking for
do dynamic bones that currently have the mesh they are manipulating turned off still require processing power? i've been setting it up so they only activate when the object is active, but want to make sure i'm not just wasting time
the ranking systems says "2 Skinned Meshes " is "good" but ingame it says 2 Skinned Meshes are Medium .. why ?
before the update, the avatar was still classified as good
It was changed recently
@drifting trellis if you turn off the mesh for the dynamic bones the actual bones will still be dynamic and take processing power, even if you can't see them.
dybones are active when the bones (gameobjects) are active, even if the mesh isn't
Okay nice, glad I wasn't wasting my time. Thank you!
but u can also disable the dynamic bone component in an animation @drifting trellis
@mint kernel That's a bug, can you make a Canny post for it?
^
above, someone claiming that I can't have multiple textures on one material, even though I literally do that on at least 5 shaders (although not on any actual production ones, mostly on ones I was playing with to see whether there was a tidy way to reduce materials without actually atlasing. please don't hit me)
reminds me actually that I must put a hard limit on texture resolution for my new project, but first I have to come up with a sensible limit for that
at the moment I'm thinking 4k (after atlasing)
there is that, but it also makes some kinds of effect slightly easier to do if everything is theoretically doable as one material
I started enjoying having everything one material where possible on newer avatars
unfortunately the new hero avatar is unlikely to be like that because I'm likely to want custom shaders on at least eyes, glasses, and one other gadget, so that's already 4 :/
I know I could slap together multiple shaders into a giant megashader and use some data point to choose the shader
I just don't know how keen I am on scraping for that excellent rating on material count, when I have other areas which are usually way harder to even achieve a medium on (for me)
now that is a problem in its own way
but I think the Android version of my new girl will not have said gadgets to avoid the need to optimise them
hm
I'd say more but I don't really want to spoil everyone in existence on what could actually turn out to be a nice gadget that nobody has done before
but in any case it requires a custom shader, so it ain't gonna happen for the Quest build
the amount of work to make an alt version that uses only animations is far too high, and the number of polygons it would take is also high
before worrying about that though I'd like to try and get my dynamic bones down to sensible levels on my main three current setups :/
finding a way to reduce my Excellent avatar from 64 polygons down to 1 is the other thing on my todo list but there's some maths that I haven't quite figured out the right logic for yet
the 64 is a miscomputation in the first place but it's slightly motivating
technically the avatar has 0 polygons
but with the new way it's calculated I should be able to get 1
nah what I mean is, I know how the calculation is being done, and I think I can make a version which is still 0 polygons, but which is now miscalculated as 1
so that it doesn't look as "bad" as 64
huh
basically I'd go from having 192 vertices to 2 vertices
no polys at all, I don't join the vertices into lines
current stats are 192 verts, 0 lines, 0 tris
and it's 192 because I needed that many to draw a sufficiently nice sphere in the geometry shader
if I reduce to one line or something then I can just use a quad instead of a sphere and it should be fine, but finding the right dimensions for the quad has been slightly infuriating
ahh. makes sense
a sphere
it comes down to "find the smallest quad that completely contains the sphere in the camera"
so you'd centre it on the sphere I think, and then figure out how wide/tall it has to be to completely cover, but it sounded really easy until I tried to do it
Does anyone know where the Texture Atlas properties window is now in Blender 2.8 I can't seem to find it anywhere.
They completely changed the whole interface in Blender 2.8.
you dont need to atlas in most cases. it just improves the avatar rating @severe isle
No it also improves performance
you should 100% atlas whenever possible
even if if you're still gonna use 4 unique materials, it helps to just have one texture for all of them
Does unity re-use the same texture if used on multiple different materials with different shaders?
I'm having a bit of a blank here, if it indeed re-uses the exact same texture without any duplicates of any kind, then using a single atlased texture sounds like the best idea, whereas if it does not, you'd have massive texture duplicates with "empty" spaces using up more memory than need be
The benefit of atlasing is the notion of reducing drawcalls, decreasing the number of materials essentially, multiple textures or a single one per material is negligible at best however
The idea of atlassing also relies on the fact that you should do as little materials as possible
You can use the same texture on multiple materials but that doesnt decrease draw calls so its the same as using different textures, aka: the atlas is pointless
Oof so maybe decimate what you can before atlasing?
Im learning making models btw, very interested in optimize stuff
Good!
Kareeda has some good optimization guides for porting avatars
Pinned in the channel ^
okay guys, so I've noticed that some of the content in vrchat was very low poly due to the fact that the models of the dmc 5 characters have a lot of polygons
I've tried to import it into unity, but when I saw it, there were a LOT of polygons over the limit. right now, I'm doing what I do best and imported it into blender 2.79.
why does my model have this like holographic effect? it shows up in some worlds and it gets annoying can someone help me remove it
might be an outdated shader?
@weak path Are you using an old version of Cubed's Flat Lit Toon?
There was a bug in a fairly old version of Cubed's Flat Lit Toon that caused data to get mixed up in the ForwardAdd pass used to render point and spot lights.
It lead to weird ghost images like that.
@empty pier you have a link to the new updated shader?
Cubed's shader hasn't been updated in a long time (Sept. 2018) you might want to try switching to a newer one.
Like this one https://github.com/Xiexe/Xiexes-Unity-Shaders
Thank you!
*Gonna be reporting this as a bug eventually (seems to be an issue relative to blendshapes)
the solution i came up with is Separating the Eyes to a new Mesh and naming the Mesh for the Eyes "Body" in order to function correctly
and proceeded to strip any unnecessary vertex groups (you should be left with four vertex groups) and that's it
*Incase anybody is reading this i'm going to state that i did indeed encounter a change and what i previously stated seems to be somewhat incorrect
*it's super weird but i'm still trying to determine what exactly is making it tick so try taking what i've previously said with ALOT of skepticism
Hmm... i've never heard about eyetracking performance issues myself personally, but i think it's likely a bug (obviously uninteded behaviour but y'know).
Do you perhaps have any further information regarding this?
And if this was done on a hierarchy without any eye vertex groups (but eye bones present) does the same occur?
I assume it must be quite impactful since 2 meshes are more performant.
@woeful ember
not sure about the vertex group part but i did see one avatar that had the loss but had blendshapes intact for eyetracking with vertex groups for only one pair of eyes
but the eye wasn't actually tracking
so depending on how some people set it up if it has the shapekeys intact it should still attempt to use eye tracking thus a loss should be incurred
i initially recognized this performance loss earlier this year but i was quite confused at the time and only recently have i actually done countless tests to finally deduce this result
I think this should be taken up as a bug-report, this sounds like completely unintended behaviour
this is definitely something that could be improved on their end or within CATs
anyways the loss from the eye tracking seems to far out weigh the additional skinned mesh
assuming you test this "properly"
incase you're wondering i was using Intel GPA Frame Analyzer to recognize this issue (but i don't believe it was absolutely accurate in some circumcstances)
i've correlated this with actual ingame performance while stressing the GPU in VR
I'd imagine it should be taken up by the devs/programmers as an increase in drawcalls to fix an issue with a "built-in" feature sounds counterproductive
OS - Windows 10 Enterprise Edition
Mobo - Asrock z97 Pro4
CPU - i7 4790k 4.4Ghz
GPU - EVGA FTW GTX 980 ti 1285 MHZ (Latest Driver and i tested with older drivers aswell)
RAM - 16GB
VR - Oculus Rift
really the loss for an additional skinned mesh is extremely negligible compared to the loss from the eye tracking
anyways yeah i'll consider making a report
If it's that bad, then i agree, but i'd reckon it'd be best for both worlds if it gets patched, increasing performance over a give/take scenario
but yeah if you wanna test this yourself and give me your results that would give me more samples
again this is only valid if you do it "properly" though
then again perhaps somehow in my process i might be hitting something unexpected giving me these results
but i've kept it as narrow as possible for comparisons
56,137
i separated the vertices weighted to the vertex groups for the eyes which had a bunch of junk cause of blendshapes since it was an MMD model
so that group was fairly large at 1,824
Very well, i'll do a bit of a test myself, which map did you use for said test?
i've attempted multiple scenarios but the one i particularly used to stress my GPU in VR was within baked lighting "Thulens avatar world"
I however do not have VR so the stress will likely be far lessened
i had to rely on looking into the mirror and using a camera to get below 45
really this probably isn't the most optimal method for this but it worked
i guess
I can however open multiple instances to stress my whole setup
there's a chance it was bottlenecking CPU since i didn't confirm this but the data mostly correlated with GPA frame analyzer
Unfortunately i can't have both instances on the same world instance as before, so i have to use different world instances, but that should be fine regardless
@woeful ember does this seem proper?
yeah but make sure you still have shapekeys
The 4 main ones right?
you can retain misc shapekeys aswell
yeah but i left all of them intact for my final test
I just wiped ehm all out honestly, but i'll re-create them if need be, as the only thing present on that mesh is the iris' which are the only moving eye part
The mesh with visemes is in BodyMain
Body is the iris'
yeah just make sure you retain the ones necessary for eye tracking
make sure the import settings for unity are identical
avoid overwriting/reimporting fbx since i suspect that might be causing some issues
Oh yeah, i'm making 2 different dupes for this, without any DB's or anything, just plain model
I'm editing a pre-existing avatar i had, makes it easier
i did the testing as barebones as possible
I'll even use the standard shader for authenticity haha
i used Robot Kyle as a baseline for performance since he didn't have blendshapes far as i'm aware nor eye tracking
Alright, going to upload both variants now for testing
Seems like the SDK decided to be a pain for me today
Triple checked everything and even applied from none to humanoid
đ¤ˇ
Behold, missing bodyparts-man
it just works! ree
I'll try re-exporting it, but the other one is clean apparently
Guess i'll restart unity to check for stupidity
Y'know, the weird part is that the multiple meshed one has no issues and it's the same model haha
đ¤
who was the villain lurking
Never tick this boi
đ
It casts some black magic on your unity deleting your body or someshiz
1 out of 2
Both uploaded
Hmm, i think ima need something heavier as i'm getting the same outcome
are you hitting 90 frames?
go to "Thulens avatar world" and turn on the mirror
I am currently there
depending on your system that should get you below 90 as long as you keep the avatars in the background within focus
ah turn on the heavy mirror
well you can see a difference the one with two skinned meshes should be the winner
in VR at 40 fps i was getting a 10fps difference
with a camera enabled
make sure you have all of the avatars in the background on screen though
they're laggy as fuck
ah a better discord user 
specifically the one in the corner with the wizard hat
Actually, it depends on which instance is currently on
They seem to vary between 70-80 fps regardless
yeah the difference definitely seems higher than it should be
the fix should solve the lost in performance almost entirely
since i was getting pretty close to robot kyle in terms of performance with 5 materials
The odd thing is that both seem to be giving out extremely close results
I tried to gyazo it
i should probably try a world with a more stable framerate
i might just mash up something to lag my GPU but i'm kinda lazy at this point
Perhaps also test it with other users' inputs, switching between both and testing with other users should provide a fairly reliable test
it's eratic for some reason even on VR it was jumping inbetween like 36-40* fps
anyways the fact that you're running two instances probably isn't helping
It was the most reliable way to stress my system at the moment, i just get a 90-93 constant fps range otherwise
i get down to 70 with the mirror turned on with only once instance
there might be some weird stuff going on with that world i'm not positive
Also for specs
Win 10 64bit (build 17763.678)
Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz
NVIDIA GeForce GTX 1060
16 GB RAM
Asus ROG GL702VM
Anyway, as it stands i sadly can't vouch for it working or not as the test seems rather inconclusive, but you may reach a reasonable conclusion at some point, going to wipe the dupes now
it's already a rather reasonable conclusion from what i've done
but perhaps a better way of stressing it to get a more accurate measurement of the cost might be in order for measuring the actual performance loss
alright so i did a simple setup in which i basically placed a whole bunch of mirrors within a world with some stuff in the background with post processing and i ended up with pretty much the same result
more mirrors seem to cause it to be calculated multiple times as expected
*Incase anybody is reading this i'm going to state that i did indeed encounter a change and what i previously stated seems to be somewhat incorrect
*it's super weird but i'm still trying to determine what exactly is making it tick so try taking what i've previously said with ALOT of skepticism
I have an avatar that has opaque and transparent parts. Is it better to have 1 material that is rendered with a transparent shader on the entire model, or 2 materials: one for the opaque stuff and one for the transparent stuff?
Cpu wise, yes. Gpu no. Bunch more reasons for no too. Transparency costs a bit more even at 0% and there are gonna be visual glitches like that weird you overlap people effect. And than there is still worrying about what shader itll fall back to when blocked.
I see alot of models with face emotes that are on desktop where they use. Shift and the f1 etc buttons is there a video that shows you how to do that?if so can somone share a link
how to create them? uhh id make a video rn if i was home
there's one on kareeda's channel
Ik how to make em dont you have to make an animation?
Like how you would make the blinking animations
His video on it is more for props. you just record an animation on a duplicate, copy and paste the keyframes so they're the same, and then drag it into your animation override
Where on the override is it for the f keys sorry for all the questions
Shift + F1Hand gesture - Idle
Shift + F2Hand gesture - Fist
Shift + F3Hand gesture - Open Hand
Shift + F4Hand gesture - Point
Shift + F5Hand gesture - Victory (Peace)
Shift + F6Hand gesture - Rock 'n Roll
Shift + F7Hand gesture - Finger-gun
Shift + F8Hand gesture - Thumbs up
@proper grail ayyy thank you so much!
Also I believe left or right shift controls which hand does it (tho you trigger the same gesture regardless) this lets you do two at a time like vr users can
Does the author of the Phoenix Wright PC avatars have plans to make them compatible with quest
How would I go about fixing this issue?
Check your rig in unity and make sure everything is in its right place @carmine blaze
would it be possible to get some vc assistance because I'm not sure I'm doing all of this the right way
actually I'll hold off on that for a minute
ok nevermind I got it
@vague hearth đ
Examples are pointless, do some basic research and you'll discover how to optimise game assets @agile radish (moved to this channel because relevant)
Just because an avatar looks good does not mean it is good, or well made. Which is 99% of VRChats avatars
@graceful vine I can optimise, I think you miss point of post
You dont need 50+ mats or even 300k tris to be a good avatar.
I not say that
Articulate better then.
Perhaps not make assumption would help
Regardless you said "all good avatar break rule" this is untrue
Sorry i put two and two together and assumed you were uninformed.
As an example, heres an avatar made by a user in here "UniformVixen#0094"
5300 poly's, one material
Breaking 0 rules and looking stylish as fuck.
Yes, I agree that avatar very nice. If only more have same skill
what rules? there's no more standards lol
they dont care about unoptimized avatars any more 
đ
if they did we'd at least see attempts to limit public-set avatars
its easier to just keep giving us tools to disable stuff
Its not easier, its almost the only viable solution right now - People kept and will keep bypassing stuff
Only so many checks you can do serverside, so why not just offload them to runtime (safety system)
Well when we had limits very few bypassed since it meant world removal
and avatars forced private
significantly slower
But its just not gonna work iguess
i noticed a change once worlds got "rotated" out and favs did as well
3 fav slots helped a lot with rotating out the trash, but forcing avatars private wrks even better
"Basic research" isn't a source of knowledge. In a game where a large portion of the game is user-made content "basic research" is a pretty poor piece of advice to give a new user @graceful vine
I don't frequent this channel very often, but i haven't seen another user even mention going to
https://docs.vrchat.com/docs/avatar-optimizing-tips.
Ofcourse its not i was just pointing out how flagrantly wrong the statement was.
fair, i just get spicy when people don't bother to give them any direction and just point them to google
I always try to help when people are genuinely interested.
I don't use cats blender, but does it have any issues with using it in blender 2.8? I've always done everything manually so I'm curious to hear what other people have to say about it
Doesn't seem to have any issues so far that I'm aware of in 2.8
I might have to give it a shot. I've heard there are some pretty cool things in 2.8, but it just looks a little odd at the moment.
Cats can do a lot for you
But are they purr-fect?
XD
Not really sure if people have made full videos about it tho
I mean there a re lot of cat videos out there, being there since the early 2000s.
... Wait.

I'm thinking of the wrong kind of cats, aren't I?
X3
But anyway if they have not even put a drop of effort in it can feel pointless for other people unlike some one saying i tried this and i dont get it or what video would you recommend ya know
there are optimization tips in the pinned messages
yeah it takes a bit to get used to 2.8's layout but i like it now that ive learned more of how to work everything @lament garden
i straight up bought a course just so i could learn the new UI the right way @azure berry
there's a lot of things i haven't put to use in blender so i figure why not
lol yeah thats useful, sinces theres tons of features in blender that alot of ppl dont even know about or use properly
right?
I'm already making some objects from scratch, but I want to make full models at some point
then start brushing up on your blender tutorials so you can learn how to make stuff
There are some popping up
There's literally tutorials for everything at this point lol
challenge accepted
There are pretty much no avatar tutorials for Blender 2.80, that's a big niche right now
why cant i get my atlas workin?
what's currently wrong with it
every time i try to turn it into an atlas it says this
pretty much our error message meant something gone wrong I don't know enough about pythons do you actually tell you why it's gone wrong you might try turning on and off your plugins a couple of times or reinstalling them
fine...
sorry I couldn't be more helpful there
Could not identify file type .dds. My first guess would be to convert the textures to .png files
@dense crane if you could, what's the name of that base avatar?
I need to get my avatar to not be so short. How the hell does one do that. These online tutorials suck. Dont mean to be a BUG
Well, scale factor in unity.
For a propper reference add a cube in unity. The default cube is 1x1x1 meter
Of use the example Avatar from the SDK
@still perch i figured it out.
So im adding my materials map which is in PNG format but why does it look like this? It takes so much just to upload avatars I swear. This is a custom one. Can anyone point in some type of direction???
Optimizing for a game engine is a little bit more, but for someone who knows how to create models from scratch it should be a finger snap.
@dense crane I happen to know the fix for that one
you have a material that has no mesh association
basically, there's a material that doesn't have a vertex group associated with it
or, you cannot access the image file the material is referencing
it's always one of the two with combiner, and yours seems to be the second case
the main issue is here:
find pc00201_body_col.dds, and, in blender, make sure it can find the file
easiest way is to select it in the dropdown on the uv mapping tab (the 2d grid) and on the right, expand the menu, click the file icon to open the reference image, and re-point it
alternately, delete the material from the list, save your .blend file, and close and reopen blender

