#avatar-optimization
1 messages · Page 63 of 1
Am I able to do that in unity?
No, not really. That's why we keep saying to do it in blender
basically everything that's not like adding vrchat components is done in blender
yeah that's the thing. These are accesories. Separate entities like knifes, and hats
I can't toggle them in blender
They can just be weight painted to bones and scaled up with the same animations you use to toggle them
I get different skinned meshes for outfits, because that's a lot of bones and constraints, but for single objects, no excuse 
So what your saying it attach them to the mesh and make an animation that makes them shrink to the point of being imperceptible and then grow to being visible
Basically yeah
yep, scale their bone to 0
kind of the same idea as eye shapes scaled down in head with shapekeys
but with bones
You can technically use shapekeys for this too but that's more expensive for performance
Yeah
Would the animation of shrinking and unshrinking the accessory be done in unity or blender?
Unity
for non-deforming accessories [e.g. a shield or staff], that don't have a negligible triangle count(compared to the main body mesh), which is better performance wise:
- merging them into the skinned mesh, weight painted to a bone, with it's own material
- using a static mesh renderer, on or parented to a bone
Well if you atlas it doesn't need its own material
but in this case you're asking if one draw call is better than two
I'm talking about a case where atlasing the texture is not possible due to using different shaders
while it is true that if you need unique shader setups having a different material is good, you absolutely can use a texture atlas for that different material since its just the texture. a hat and a shoe can use different materials but can be drawn on the same texture, just shared. atlasing is still good in this case because it means less overhead is spent allocating space in RAM.
as for which is better, props should work fine with bones. things only get inevitably annoying with clothes. see "How does a shader based system compare to other methods" in https://gitlab.com/s-ilent/SCSS/-/wikis/Manual/Inventory-System
my question is more about whether or not, if you can't get around 2 draw calls, having the 2nd draw call be a static mesh would be better than having the 2nd draw call be of a different part of the skinned mesh.
if the difference is negligible unless you get into triangle counts of multiple millions it doesn't matter though
Scaling to 0.0001 is a better practice in case anything tries to divide by that scale
verts on a skinned mesh are more expensive to render than static meshes, but if youre talking about individual low-poly props then having them set to bone toggles is mroe ideal. see the message above
The best practice for this is to assign each object to a bone on the main armature, and then separate the objects to another single skinned mesh. Doing this, you'll have 2 skinned meshes total. Still at good perf rank. Then in Unity, set the bone weights on this second skinned mesh to 1 bone. That way Unity only skins that mesh once, which is all it needs. Then you can modify the bones for the objects any way you want
for any single assets that are approachign the amount of triangles found on, say, a quest avatar (5k or higher) id start worrying about spending the drawcall instead
What? What did I do?
slvr mentioned above that scaling bones to 0 is not great, which i didn't know
better to do 0.0001
ok
Hm, what i am trying to do is: Lets say i have a tight shirt on an avi, on the arms it clips very badly, so i cut the arm off on the inside and hide it. Problem solved, but i was wondering, what would be best in this case or cheapest in performance, make the removed part a separate mesh or a separate material that i can apply a invisible material to so it disappears until i need it shown
I personally use shape keys since I have heard from other sources that they aren't as bad computationally as some people make them out to be.
for a shirt, I'd make a shape key that shrinks parts of the arms and torso down so that they don't clip
Ohhhhhh, i didn't think of that, i would try that!
I would assume i just make the parts of the arms like i normally would and then move all of the vertices into a smaller spot as a shape key?
I will try that, thanks!
It depends on the poly count of the mesh as a whole. If you have over 50k polys (or around) then blendshapes begin to start causing heavier CPU performance hits when used. If you aren't 50k or lower I recommend using seperate meshes. Otherwise if you're low poly then blendshapes are cheaper
Another method is to use texture masking and mask out the arms to be transparent in the shader. That is if your texture allows the arms to be masked
Uh, I have a question. I'm making a Robo Fizz model and for some reason it says I've got 27b polygons and that the avatar doesn't contain an animator. When I switch the build to Android it just breaks the SDK
You are on a old sdk if there's no auto fix button for that, enable read/write on your import/fbx
whats the limits for quest avatars
https://docs.vrchat.com/docs/avatar-performance-ranking-system scroll way down |
The Avatar Performance Ranking System allows you to see how much a user's avatar is affecting performance via analysis of the components on that user's avatar. You can also use it on yourself to see how performant your avatar is. This system is provided to inform users what is likely the most perfor...
This page will describe various limits in place for the Oculus Quest version of VRChat. These limitations are in place in the interest of performance, user safety, and discouraging malicious behavior. Find more information about limited components on our Quest Content Optimization page. Although the...
aight thanks
should this be fine?
Drop 400 polys and you'll jump into Medium
if i drop any lower the hair will look blocky
i guess i can try and lower the clothes
I cant figure this out
it's so close though. Cmiiw but medium is the level where you stop being auto hidden
More of a #avatar-help thing but it's exactly as it says. Can't have both. If you're upload an avatar remove the Scene descriptor or vice versa.
the problem is, there is no scene descriptor
Or I just cant find it
Also sorry, I ddint know which channel
go to ur hierarchy and do "t:script" and just look through EVERY gameobj's components for it
alr
and if you see 2 pipeline components
remove both
1 should stay
if the scene descriptor is gonne
Uhh where do I type in t scripts? Still really new to unity
This is only the 2nd one I've uplaoded
screencap your entire unity for me
im to lazy to open my project
and it will take ages
Oh sorry
When I type that in, backups comes up and the avatar model
look through the componenets of each object for a scene descriptor or a stray pipeline componenet
Whats a stray pipline component?
pipeline component on a gameobject thats not accomapnied by an avatar or scene descriptor
so just something that isnt attached to the avatar? Got it!
yep theres no scene thing of any non attached items
Sorry for wasting so much time!
dont worry its not like im doing anything productive or should be
make a new project anf import the avatar agin
u probbaly just accidently added a scene one
I've already done that
unless the avatar creator was dumb enough to put a scene on it
it shouldmt stop u from afding or deleting componentys if its in a prefab
whenever i try it jsut says" children of prefab cant be deleted
but i cant delete the whole model
and theres ntohing detached from the model
*nothing
just unpack the prefab
rightclick the root of the prefab
and unpack
You only need a few and then use naraenda's tool to make constraints instead https://github.com/Naraenda/NarasUnityTools
do you still need to select legacy blendshapes in 2019 unity?
Yes
how do you stop clipping from gravity on hair without colliders?
Get shorter hair
It's already short hair
I just don't want it clipping through my face when I look up
Hi, I want to migrate some eyebrows to quest, but I use transparency on PC. Quest doesn't support transparency from my understanding so the texture bake fails. What options do I have for porting this to quest?
Sorry what do you mean specifically? Like go in and delete the black parts?
Or remove the eyebrows alltogether
use the knife tool to cut out the eyebrows without the black parts
then delete black parts
Basically since you can't have transparency, remove all transparent mesh
Cookie cutter style the eyebrows
As far as eyelashes go, that might look kind of jank so you could draw them on to the face
Unless you really wanted to cut them out
Can I mirror the knife tool? If so I can just cut once for both
Nah it doesn't look like it. But the knife tool would work for my purpose. Thanks! With this I think my shape keys would keep working too so I'll probs go with that
I would save a copy just in case
You might be able to mirror if you use a plug in or turn it on in sculpt mode. Probably not the mirror modifier though
^ take my advice with a grain of salt, I'm very much an amateur compared to a lot of experts
You could also try replacing the transparent/black bit of the texture with the skin color/texture so it'd blend in
that it would just look like you're flesh-colored chunks sticking out of your face
Thanks Mage, I was also trying that. It's a bit hard to get the exact skin texture, since I have base+normal+metal+ao. I might just go to one of my other models and rip the eyebrows from those.
yeah Rain, that's exactly what it ends up looking like. That may work for other models, just not this one in particular.
Alternatively you could paint the eyebrows onto the face texture. However, they would use shapekeys
Do you think my armature would move smoother if I converted my mesh to Tris?
iirc VRChat doesn't support using quads anyway so it should convert to Tris unless you tell it specifically not to
Edit: at least in the Unity I checked when you have an FBX imported one of the options is "Keep Quads" and it's not checked by default
I'm watching https://www.youtube.com/watch?v=SimRLR_lRcA (Your Skirt SUCKS: Zero to Hero)
But I'm using 2.93, how do I do the depth buffer bit mentioned at 13:30?
I doubt you're using the normal+metal+ao on Quest so wouldn't it be fine to just use the base/diffuse?
no it would still look silly because the model has shading
the quest does support the standard light Shader which does have support for all the same apps as the standard Shader
Ah because of some personal preferences I'd only use Standard on Quest for like metals and props
@marble ledge if you want Diffuse + AO on Quest you can simply premultiply them with addend
If you want I can put a simple couple lines of code you can throw in the Blender scripting tab to do that
can u get banned for very poor avatars?
If they are designed to crash people, yes, otherwise (and this will be most cases), no
okay so I want to reduce the number of polygons on an avatar in blender and I saw a video where someone separated by material to be able to work on a single part of the avatar at a time. but the avatar I got only has 1 material so I don't see a way to separate the model into different sections
that's what I tried to do but it took too long and loops extended beyond different parts of the model so it became too manual and after reducing it from 27k to 23 I eventually gave up and used Quick Decimation. but it'd be easier to do if I could separate the mesh into different sections
same for selection more/less
hey, so I want to bake the textures on the face of my avatar. On the left is the original, on the right is the bake.
Two questions.
- Where are the extra meshes coming from? I deleted those. I also made sure their materials were deleted.
- The eyes and some other materials didn't bake correctly. Is there a way to prepare materials to be baked properly?
Do you have an unapplied mirror modifier?
I don't fully know how to check that. I assume it's this tab? If it is, I only have an armature modifier.
could this have something to do with my issue? It's a setting under bake
Are you sure your bake target has no extra mesh geometry? It might be hidden with H (unhide with Alt+H)
yeah, I selected everything and unhid. I don't see anything except the one mesh I'm trying to bake.
I have changed the topology to 60% less than the original mesh size, and the avatar size is still at 10.77 What can i do to reduce it?
Did you crunch your textures ?
ah not yet
ok, so it hasn't changed anything

is there anything else i can do?
I just atlased the main body materials, but it's still at 13.
I also reduced all the bones that aren't needed for quest compatibility
You can use the vrworldtoolkit to get a build report of your uploaded avatar
13 materials ?!
one material for the body, other materials for extra items like a hat, a cane, a bowtie
Why don't you atlas the other parts ?
i have. The tophat has one material, the bowtie, the cane
the others also have one material
So how do you have 13 materials
Quest isn't really meant to support avatars with whole inventory systems
Hi, I'm using CATS bake and my eyebrows end up very big but everything else ends up small.
On the documentation I see this:
The islands end up super tiny or some object is way too prioritized after baking.
- Apply scale for your objects.
Source: https://github.com/GiveMeAllYourCats/cats-blender-plugin/wiki/Bake
What does "apply scale" mean?
ctrl + A
ty, ctrl+a + scale. Let me try now, but that seems like it most likely was the answer.
👍
Try mesh compression too
You shouldn't use mesh compression on skinned meshes
Why's that?
From what I can see yeah it's lossy compression so
it will lose some accuracy on your model and that could be more extreme on skinned meshes
but I haven't noticed anything bad personally using it
It affects vertex positions, normals/tangents and UVs precision, i really doubt you'd want that on something that is constantly moving and needs to be calculated at runtime for all kinds of things
I guess it depends how much you rely on your normals, for toon models, that may be less important, but still, i've never seen it recommended anywhere or by anyone in the past few years
Sounds like a last resort kind of thing, but I can't vouch for it
Not the most accurate but maybe if I "Build and test" once with it off, low, and high, and look at the VRCA file size
I can see how much it actually does for me
versus the quality lost
I'd recommend removing unused shapekeys before turning on any kind of compression tbh
often we see booth and models from wherever people can find them, have hundreds of shapekeys, that bloats file size a lot
Yeah true blendshape optimization isn't done as often as it should
Also not counted in the performance rank but I've heard does effect performance
It eats verts that should be there to help bend, makes skinning randomly bad
If I remember correctly it doesn't remove any verts/decimate etc it "just" stores where they are with lower accuracy
Anyhow could do this if I remembered where the test avatar files are stored
Used to have it memorized, just search for your unity project name one folder up from %appdata%
'%appdata%\..\local\temp' or something
Thanks
Seems "Mesh Compression" doesn't effect VRCA file size
and if it does it increases it
by 1kb
%appdata%\LocalLow\VRChat\VRChat\Avatars
Okie
Hey, I have an avatar that I want turned into quest and pc.
I have a base for it but I cant get some of the hair toggles, color toggles and body toggles to work. Can anyone help me?
Just getting into rotation constraints to turn hair from 30 dynamic bones to 6 but have an interesting problem, the antennae if constrained with the simulate chain rotates exactly opposite of how id like it to. Any quick easy math to have a rotation constraint be mirrored? Just making the weight negative doesnt work and idunno rotation math to manually fill out the constraint settings
i think that putting socks over a body doubles the amount of polygons at that place so, anyone has advice on optimizing that? could i just erase the legs and leave the socks there?
yes that's what you do if you don't plan on toggling socks
can always use the Legs mesh as the socks and just change the texture
same difference but that'll be the best way to do it imo
I would just use an extra dynamic bone for the antenna. Unless you really need to stay under a certain performance limit
Hey, I have an avatar that I want turned into quest and pc.
I have a base for it but I cant get some of the hair toggles, color toggles and body toggles to work. Can anyone help me?
Toggles aren't recommended on Quest cuz it really eats into performance. Also no fancy shader stuff available
You could do what NotA did but just for the antennae/banana peel
https://www.youtube.com/watch?v=rggL1eq8D2Q
Join my Discord for premade fake dynamics for models:
https://discord.gg/JRhjx8d2Wt
The premade packages also include helpful tips and links for making your own.
com vrchat oculus quest 20210829 054642
You make some animations for when you turn left/right, and walk forward/back and left/right
put em in a blend tree and set it up to move between the animations depending on AngularY and VelocityX/VelocityZ
Ohh, neat but i hardly ever use locomotion.. i just ended up using the ear bones as the constraint source! Something facing vaguely the same direction
It goes in your gesture layer not locomotion
Non humanoid bone, i think fx but dont quote me
Considering I have it setup for my hair and ears neither of which I'd describe as "humanoid"
¯_(ツ)_/¯
From what my friends done with 3.0 they say the documentation is just wrong sometimes
This is what VRChat says for the layers https://docs.vrchat.com/docs/playable-layers
In the Avatars 3.0 Avatar Descriptor, there are five buttons labeled: BaseAdditiveGestureActionFX These are Playable Layers. Each of them takes a Unity Animator, and they layer on top of each other. In other words, you've got five root animators to play with, and each of them can have a huge amount ...
^
@random tangle You're more knowledgeable here
I c looks a lil different from last time ive read it. Now says gesture is for bone transforms without mentioning humanoid bound
In addition, if you want to have an "idle" animation for non-humanoid bones like a tail, wings, ears, etc-- Gesture is where you should put it.
Oddly specific retcon
But yeah, just used the ears! Couldnt figure out negative rotation constraint
oh that was probably because of bone normals
it’s rotating the wrong direction because it’s bone roll is backwards
From the docs I read bone roll shouldn't matter as long as it's the same for everything
Obviously you can't use Poiyomi etc but I'd say you can do fancy stuff with a mix of Standard and Matcap Lit
does anyone use naras dynamic bone constraints?
I managed to get to 70000!!!
cool
Me 😜
I should have just pinged lol. Was wondering if your tool works in 2019 unity or something?
This just happens when I try to set up constraints
Uhm, it works fine for me. Are you on the latest version of the tool/SDK?
latest version of both. Ill try setting up in a new project with just the SDK and tool and see if it still works then
Can you check how the constraint looks?
Constraints only really work when you have multiple bones parallel to each other that you want to control with one chain of dynamic bones
In this case, multiple hair bones get mapped to the HairPhysics chain
For a single tail adding constraints wouldn't add much
(You can remove the constraints with the [remove constraints] button in manual setup)
I see. Thanks for the help.
Does anyone know how (of if) its possible to keep transparency by texture baking via CATS? I have a normal png file with alpha channel already included (didnt changed the blender materials)
When i try to bake it, i always get a black background instead of my alpha channel.
I also tried the third last option 'tranparency' but by hitting this button i receive a completly transparent texture ...
@normal nova ^
Yeah so when baking with transparency, set transparency to 1 and bake it normally so it doesnt black all the alpha. Than make a seperate bake target that copys over black/white alpha and manually compose it in gimp
so no direct way? I see its okay i will jump to gimp then 🙂 Thank you for responding!
Why is it that I cannot upload this avatar to quest anymore?? Before the new updates to VRCSDK. I could upload just fine. Did you guys decrease the amount of performance a quest can use or something???
The upload button isn't disabled
If it's not working that might be a bug, unless alpha isn't set up right in Blender
^ this is how Bake does it, essentially
its friday and im a few drinks in and i thought that was the bake panel, not cats. my bad
Bake panel in cats, yeah. I kinda wish I'd called it something else
This is my first time making an avatar, would anyone mind telling me how to combine materials/atlas textures?
Alright
Wait they’re gone
What
Where did their message go
WOTTT
It didn’t look like they did anything though
I think it was some misunderstanding apparently?
hopefully it was..
Yeah
Who knows how to make avatars on unity cuz this my 1st time and I wanna make one my self
so could someone help me
If not that’s fine
lookup some tutorials on YouTube
As a beginner as well I used vroid to make one you can put it in unity after you finished it
I was optimizing my avatar and when I put a material with a quest shader in a certain mesh, It didnt show the texture at all and it was just pitched black (shown below) (shader im using is standard lite)
I think quest shaders multiple the texture color with the vertex color. Make sure that the vertex color isthe same white (or gray) for the whole model
How can I find vertex colors in my avatar?
thank you!
Is it really impossible to upload avatars using mac. I just found this out and I already have a blender file i made on my mac
Hey, quick question.
I'm trying to make my avatar quest compatible right now, which has been an... adventure.
Going from 69800 polys to 5000 was rough, as were other things.
But after removing some parts of the avatar, trimming things down, et cetera, I'm at 4999.
I'm running into an issue, though.
Two, actually.
- I'm over bone allowance, and the bones are used for parts I've removed from the base model altogether... I can't seem to nuke them from the FBX file in Blender. Is there a way to get rid of the tail bones, for example?
- I need to get rid of materials by combining them. Right now, I'm using 4 material slots, and need to find out how to combine those down to 1. And I'm not really sure how to do it.
5k ? excellent is 7500 for quest 🐇 <10k for a fallback if you also want it for pc/quest , for bones just merge em to parents until there's nothing left (or active)
Odd, the SDK told me that fallbacks required 5k
Apparently very
is there a way to have particle trails without it costing like 2 material slots each despite the particle system having rendering set to none?
this somehow costs 2
No, Particle systems will always count as 1 (2 with the trail module activated). Having 0 Max Particles will count it as 0 but it's not a value you can animate so it's not helpful for effects.
damn that sucks, any reason it might be that way though?
doesn't make sense something that doesn't render and/or have any material counts as a slot
I'm not sure what it's counting in this case but I don't believe Particle systems were ever intended for logic. Plus, cmiiw, you could upload with the renderer module disabled to reduce material count and animate it back on at runtime allowing you to bypass the material count. I can assure you people abused this if it happened.
The polygon count for an avatar is based on the total number of polys if all component objects were active at the same time aren't they? Not just the ones currently active.
Yes. Game has to assume all meshes can be visible at once even if the logic won't allow it
Hmm.. frustrating, I've got an existing avatar that's in the 60k range currently and if I add stuff to it, even if I swap out, its going to fall over in to the Very Poor Category on polys
Just do 2 versions, one
and one
. The first can be your fallback so everyone can still see it. Just not all the fancy stuff
If you're too lazy to make a quest fallback just upload a x-bot model or something for that platform lol
I already have a custom fallback but that uses another model entirely. I'm trying to make sure people who have default safety settings don't miss the cool special avatar I'm building for an event next month.
How much over poly limit are you? Maybe you can still squeeze it under 70k
might have to consolidate on the things on your avatar wit meshs
or dissolve some more edge loops lol
I'm -starting- from an avatar at 67k.
I've started already be taking some of the high detail but non-animated bits and doing a little decimation first
that got me down to 48k
which is a much better starting point
Ouch, that's still rough
Indeed. but I managed to squeeze and stuff a protogen model in to just under 10k for that custom Quest fallback and still keep it looking good.
So this by comparison... XD
Yeah, it's still doable compared to someone sending me a 200k poly avatar to be optimized 😂
and that started at something like 27k to start with and was... spicy
Ah wait, I tell a lie, only 20k start.
This was how that turned out. Starting NKD type model on the left, 20,806 tris. PC appearance using Poi shader of my fallback centre, 9,880 tries. Quest appearance on the right with basic mobile shaders.
I occasionally have people go '...what happened to all your fluff?!' 
I suppose that's one way to do it to delete the fluff instead of dissolving dissolving Edge loops?
that works
is there any server i can VC With people that build avatars
i kainda need big help but typing will get me no where
ahey anyone up for a challenge? anyone want to teach smoothbrain how to fix his gfs avatar so he can upload it as a gift?
i got it from a dude who scammed her out of money so imtrying so desperatly to fix it.
The only thing preventing the upload is the shader, which it says how to fix for quest, you need to be using a Quest supported shader, which are the ones that VRChat provides, the rest is going to require blender work if you want more optimization
That should work, but if you changed that and there is still the error about shaders, you have more meshes in the scene that have non compatible shaders
where do i go to fix that. i retexture avatars and know how to change that.
but idk where in the hierarchy to look
if you click on the Select button it should select the mesh with wrong shader
there isnt a shader dropdown when i do that
otherwise you could use the scene hierarchy search for t:skinnedmeshrenderer
how many meshes are there for the body, is it all one mesh, or are there things that are separate, like tail, ears, hair being their own mesh
and again, try using the scene hierarchy search box for t:skinnedmeshrenderer
ok i found that the wings are using a standard shader, but its grayed out and wont let me change it
like its locked
you'll need to extract it, you can select the mesh in the project view, then switch to Material tab, and extract materials
it is the bottom panel, it is named Project
oh >~> my bad
if you want quicker ways to find your avatar assets, you can do right click and select this
and then right click on the item it selects and click this
you would just select the material, not the textures, and change the shader. If you can't then you need to extract, which you can do the first part from above to select the asset, and then switch to the Material tab and extract materials
also in that image that you have the texture selected, you will probably end up wanting to crunch compress if you go over the 10MB limit for quest avatar uploads
ill get that done after this, um how do i extract the materials now?
if you look at the image I sent for that, there is an Extract Materials button that is disabled, because I already clicked it previously to extract materials, you do that
right clicked and cant extract from prefab
wait my screen doesnt have tabs like those. im not sure if i did this right
This is your first step
Then you look at the inspector, and change to the Material tab like in this
yes, but you already have a material set up don't you? One that you can change the shader on? Just assign it to that mesh. If the mesh only has one material, you shouldn't get the error for wrong shader
time to try saving your scene and restarting unity
ok but the hierarchy for the wings was grayed out too thats why i couldnt alter the shader. are you shaying to swap the materials, change the shader from standard to vrchat/mobile, save and restart?
do as much as you can and save and restart
ok ill get right back to you
it let me build and publish it thank you so much! i owe you a beer good sir/madaam
Just be aware, because of the performance rating of the avatar, no one will be able to see it without manually choosing to show the avatar on quest
can i friend you so maybe you can help smooth brain me crunch compress it later?
no, because I am not always available, just post in the discord and if I am available at the time and see it, I will help, or someone else might
it just needs to function for her. her and i were desperate. she paid $280 this avi and the creator has been nothing but a pain in the neck for the both of us and a few of her friends. the dude cant even test them in vr cause he doesnt even own a headset.
thats fine i really do appreciate the help.
To crunch compress:
- Find the relevant texture(s) in the [Project] window that contains all your files.
- Select the texture(s), the [Inspector] window should show them.
- Scroll down in the inspector, there will be 3 buttons: [Default], a PC icon, and an android icon. Make sure Default is selected.
- Set the following settings:
Max Size : 1024 or lower (2048 is OK if needed, don't use 4096)
Format : Automatic
Use Crunch : CHECKED
Compression: Normal/High Quality
Compressor Quality : ~75 (lower this if file is too big, higher if quality is ass)
- Apply, give it a minute to do the things, and enjoy smaller file sizes
4b. Alternatively you can do it more manually by selecting the PC/Android icon and use:
Format : RGB Crunched (ETC for Android, DXT5 for PC)
For the other options use the step 4 settings. Feel free to tag me if you need more help
Example:
What's the current tri limit for avatars?
PC or Quest
Pc
The Avatar Performance Ranking System allows you to see how much a user's avatar is affecting performance via analysis of the components on that user's avatar. You can also use it on yourself to see how performant your avatar is. This system is provided to inform users what is likely the most perfor...
70,000 if you don't want to get labeled very poor but it's not going to prevent you from uploading if you use more triangles
glad to be of help
20,000 is the number on Quest and very poor avatars on the quest have to be manually turned on
so most folks would just see a fallback Avatar if you have very poor on Quest
Also good to know, I was thinking of making the PS1 styled version of my avatar the fallback for quest
cool and if you're curious the max triangle count for a fallback avatar on Quest is 10,000
Great, I think mine is in the 500-1k range
huh, neat
Hey there. Trying to figure out how to upload 2D pictures as avatars. I’ve been following this video but I keep getting an error st the end that there is no animator attached to the empty object and front even tho I put the avatar descriptor on?? Help would be much appreciated
Finally made an update lmao. It's a shame it's a bit longer than the previous one. Well, I hope this was useful to you guys.
I forgot to mention again. Use Gimp's autocrop if you don't have photoshop.
Firefox extension for saving .webp images as .pngs (not needed if on Chrome):
https://addons.mozilla.org/en-US/firefox/addon/save-webp-as-png-or...
Did you have upload a avatar before?
Avatar Descriptor tells the SDK about the avatar's existence, and how it should function, an Animator is a separate component.
I have not
So then how do i assign an animator to a flat avatar that has no bones since i cant apparently upload without it
You should be able to upload without it, and an animator can apply to any object
Animators don't just dictate bone positions, also things like an audio source's volume, a sprite renderer's colour, whether or not an object in enabled.
Or you can just drag and drop the 2D avatar inside the object of the blue avatar an then set the avatar descriptor in the blue avatar
oh and disable the blue avatar mesh
https://github.com/Thryrallo/VRCAvatarTools - so this little tool exists, which gives you a rough calculation of how much VRAM your avatar will take once loaded in. Made by Thry
mhmmm atlasting 4 1024x1024 textures perfectly into a 2048x2048
this is what i live for
cool
what the different between unity/blender for attaching like ears/tail/hat?
Can someone help me upload an optimized avi 🥺👉🏻👈🏻 It's supposed to be optimized already but I think im doing it wrong :( I'd SUPER appreciate the help!
can you show us like the stats when you try to upload the Avatar so we can see how it's unoptimized
This ? I'm sorry im SO bad at this stuffs ;-;
basically you just have too much Dynamic bone the simulation going on
Do you know how to make it... not? Lol
This is the only red and orange stuff in the avatar stats in game, everything else is green c:
it gives instructions for that in the tooltip
What is a tool tip 🤔
I suppose so I haven't really tried doing Dynamic Bones on stuff that has lots of Bones
Do you know if this needs to be done on blender or can it be done in unity?
You could try using this to lower the bone transform count
https://github.com/Naraenda/NarasUnityTools
How do I get that ? :)
Dynamic bones can be lowered in Unity but material count kinda needs to be done in Blender
I think my materials are good, right? How do I lower in unity?
I think You download this and unzip it in the assets folder in your Unity Project
https://github.com/Naraenda/NarasUnityTools/archive/refs/heads/main.zip
they could be better but they're not terrible
Can it still be green with the materials I have now?
With current materials no
best it can be is Medium
Grrrr lol
presumably
Aighttt I quit lol
?
Thank you guys for your help ♡♡♡
medium is not that terribly bad
I need green for the events I go to :(
I'll probably just pay someone to do this for me in time lol
I'm far too uneducated on unity and whatnot for this stuff lol!
For getting the materials down you could make use of CATS atlasing
or if your avatar has a lot of solid colors that are duplicated you can just merge those
You only have 10 materials if you get it down to 8 you'll be good
I unfortunately have noooo clue what any of that means let alone how to do it 🤣😅😅😅
I believe in you
What is cats atlasing and how do I merge my solid colours 🤔
Here I'll send some videos and if they don't explain it well enough I'll try to guide you through it step by step
I'd be such a bother to get guided through all this stuff :(
I could do that
But I doubt you'll be that much of a bother
okay?
https://www.youtube.com/watch?v=YUZeBfYbfVo 5 Minute Material Fix old but still useful
https://www.youtube.com/watch?v=WTiBaiRLi_E File size optimization Probably not about lowering mats but still
https://youtu.be/D5UdA4uXHqM?t=706 Optimization Tips for your avatar in Blender I skipped to the atlasing bit
https://youtu.be/8MkgXcidxg4?t=315 PC to Quest (2021) Sippbox
Tutorial for Fallback but still shows atlasing you don't have to get it down to 1 material though for a "Good" PC avi
Feel free to dm or @ if the videos aren't enough
Also I'd recommend saving it as a different FBX so you still have the original unmerged materials if something messes up etc
This can be used to copy everything over from your setup one in Unity to the new FBX when you import it
Good luck @quartz grove
In Unity you're most likely attaching it by having one model as the avatar and another for the ears etc
Then just sizing and moving the ears to the head then parenting it to the head bone right?
Which means it's another skinned mesh (Model with bones) and generally people don't merge in Unity
Meaning you'll have multiple skinned meshes on your model
Whereas with Blender it's really easy to merge them so you only end up with one skinned mesh
Which generally doesn't lag people as much
lets say i did merge hat with the avatar could i toggle the hat off later
Yes in multiple ways
- You could keep it as a separate mesh on the same armature
- You could give it itsown bone that's parented to the head bone and merge meshes and then just move/shrink the hat bone
- You could toggle it by making it invisible with shaders
- You could shrink/move it with Blendshapes not recommended
etc
@regal lantern
for quest avatars is there a way to fix culling issues in blender?
Duplicate the faces on areas where culling is an issue and flip normals on them.
ill try that
Hey, I want to optimize my avatar more, but I'm still relatively new to blender does anyone know of good tutorials for material optimoptimirung or can give me A recommendation ?
Try these @real topaz
thanks
3.0 ? tracking control not animating certain things
detect prone/crouch then set your tracking control
can you DM me a step-by-step guide or a video pls
check the 2nd layer on that, it detects upright/crouch/prone and sets tracking/viewpoint (some tweaks might be needed its bit old now)
It's working, thx!
does anyone know the name of tool for unity to make custom shape poligon
It's really better to do it in Blender, but if you need a simple shape quickly you can use ProBuilder in Unity
So I got into a rant with someone on Reddit about atlassing materials and it led me to do some testing. 😄
I was tired of seeing "8192x8192 textures = lag, atlassing bad! :("
Here is my results from the same avatar before and after atlassing the textures:
Atlassed: (2 materials)
CPU frame time: 0.5ms
Render thread time: 0.2ms
Batches: 11
SetPass calls: 11
Texture size on disk: 10.5 MB (Has some white space > Bigger file size)
Texture dimensions:
- 8192x8192
- 4096x4096
No Atlas: (8 materials)
CPU frame time: 0.7ms
Render thread time: 0.4ms
Batches: 55
SetPass calls: 35
Texture size on disk: 8.63 MB
Texture dimensions:
- 110x84
- 2048x2048
- 288x265
- 1024x1024
- 1024x1024
- 1024x1024
- 2048x2048
- 4096x4096
The avatar is 69,100 polygons, has 1 skinned mesh renderer, and 1 dynamic bone component controlling 13 bones, no colliders. Note that these results are in an empty Unity scene with a directional light and a camera, some of the frame time will go to that. Only use these results to compare to each other. 🙂
TL;DR:
In this instance, there was about a 20% increase in disk file size (Same will likely be for VRAM), but, a decrease of 50% on the GPU frame time and around 30% on the CPU frame time. It's a trade-off, but it's not a bad deal IMO.
All said and done... She do be cute tho. 
can get rid of the increase with a 8192x4096 atlas. They dont have to be square, just to the power of 2. Nice!
vram is really the only cost but yeh worth it
Not worth using this much vram, no one said atlasing was bad, it's just bad when you go up to 8k
especially for an anime model 
also noting your UV stretching on your shirt, which tells me your UVs are really not optimal, which means wasted space in the UVMap to begin in
I think the main issue is the average GPU has about 6 - 8 GB of VRAM, and once you run out of VRAM then performance will suffer a lot.
Oh that's just scuffed weight painting. 😄
Just copied weights from body then did some minor fixes.
I make like 2 avatars a year, it's not a big enough concern to fix.
Still unsure how you need that many textures in the first place, care to show what the atlas looks like ?

burn it with fire/
I mean... Looking at it, I can relate. 😄
why does your model use 2 face textures that seem identical
But at the end of the day, it gave the performance benefits I expected and had performance decreases in the expected places.
You know, I have 0 idea.
Actually, you know what? I do know.
I saw this when I was importing the individual textures before.
The none-atlassed model had 2 different face textures, like the eyes were a different material or something...? I guess I forgot to change them to all one texture for the atlas.
¯_(ツ)_/¯
little did they know if they wanted to change the eye color
they can just crop out the eye part of the texture
I'll be real... I made this avatar like a year ago. 😄
I've learnt a lot more since then.
oooh
yeah, all that could be a single 4k atlas 
Do you have a recommended guide/tutorial on atlassing manually?
That could be super useful going forward. 🙂
I'm afraid my Blender knowledge is super limited still, but I am trying to learn it for some work on a world. 👍
I suppose then you could have GPU instancing even if you had 2 materials?
Or wait... Maybe I'm just confused.
If they're on the same mesh that sadly does nothing
Oh, interesting!
Also... It's obvious but to clarify: The graphics side of Unity is something I rarely touch. I just made the comparison as some actual results to put into the "atlassing bad" or "atlassing good" argument. 2 identical avatars, one in the typical MMD state before atlassing, one after CATs has screwed with it. 😄
We are trending to 8GB being the norm though! 👍
It's a shame it's not 11GB though.
The SteamVR compositor alone eats up a lot with modern headsets. 😓
Though there's likely on 6GB usable on an 8GB card when in VR as a result of that.
And now put 80 people in an instance 
I usually manually atlas my textures and then do ratio math of the dimensions of the original texture to the dimensions of the new atlas to figure out how much I need to scale the UVs
if that makes sense
🔥
I'm not sure that'll ever work really... 😄
It can, just not with poorly optimized avatars
hiding avatars by distance would also be a neat feature for big worlds with lots of people
if someone wants to choose to opt in on that
that wouldn't help with vram though
^ Still loaded, yeah.
avatars is only one thing, there's IK and voice, and all other kinds of data
oh so once its loaded its stuck in your vram until when?
you leave the instance or the other user does?
Was sure it was supposed to clear after a certain amount of time, or when whoever is using it leaves, but idk
Do we still have the 'change 2 worlds before cleanup' thing?
To avoid the unload hitch?
@calm spade (Sorry for @. :p)
Nope, it's much more dynamic now
Sweet!
and faster to unload
🔥
I have to say though... I'm really sad that newer GPUs don't all come with 11GB of VRAM.
Like 8GB is no longer acceptable.
me with 4GB
4k gaming with vrchat funny joke
Index is apparently 1440x1660 per eye. That's a combined 2880x1660. That's 4,780,800 pixels.
'4K' (UHD, 2160p) is 8,294,400 pixels.
'2K' (QHD, 1440p) is 3,686,400 pixels.
So I guess Index is just over QHD actually, by a 'mere' a million pixels. 🤷♂️
Realistically, you want a GPU capable of 120 FPS gaming at QHD for the Index I guess.
People forget how resource intensive VR truly is. 🙂
pfff resource intensive? "It works on my end"
"24FPS is fine d00d, just deal with it". 😄
I have no idea how people play when it's more motion smoothing artifacts than game, that'd make me very sad.
g2 > 9,331,200 pixels just controllers
meh
Index at 150% is 4320x2490. That's 10,756,800 pixels. 😄
That is now 20% over 4K.
There's a funny story there actually. I hit the digital resolution limit of my 3090 when I tried a PiMax 8KX as I have 2x 4K displays outside of VR.
Who was that thing even designed for? It can barely run anything on a 3090 at native resolution, hardware doesn't exist to run it. 🤷♂️
It is 3840 x 2160 per eye. 😄
Combined, it is 16,588,800 pixels.
for those people with like quad gpu setups ig
Has anyone checked how much blend shapes impact avatar file size as well as performance weight
And i mean extreme case
Not like 20 - 30 i mean 200 - 400 blend shapes
This also includes a question about max texture size for atlasses i want like 1024 x 1024 texture attlases that gives me 16 max textures if max size is 4098. Can it go beyond?
Like would unity struggle pulling 8k?
Or should I gave up and put a second mesh render and material
To be exact i need 20 textures so it is 4 above the damm limit
don't know about filesize, but they can be awful for performance
one of my older avatars, is medium ranked, and has a single skinned mesh, and I used blend shape for all my clothing toggles
that avatar performs worse than my worst very poor avatar
i've since fixed it up to remove blend shape toggles and just toggle meshes regularly, and the difference is about 2-3 ms gpu frame time when looking at the avatar 1m away on a 1080
its generally recommended now to keep blend shapes to a minimum, and to keep them on separate meshes when possible.
ie. head mesh with all the associated blend shapes, and another mesh for the rest of the body etc
Yeah i think about devision of that which would solve both the material and performance issues
they're also more expensive to constantly keep active, so if you're using blend shapes to change how your avatar looks permanently, its better to bake those transformations in blender.
that's probably the main reason why that avatar of mine performed so poorly, as it had a ton of blendshapes constantly active hiding clothes etc
I thought of using blend shapes to transform clothing as in change the collar shape or sleeves length
I think i will keep it to minimum and use alpha cut out with material swapping
honestly, a few here and there isn't so bad, just don't go too overboard, and keep in mind how many tri's that mesh has
my example is an extreme case with a single 69.9k mesh with a ton of blend shapes so
69998, i really wanted to be medium
I see. I will have to do some dark magic with the normal map and textures
I want to use those instead of poly for increase details
Thanks for help
i'd also grab this handy tool by Thryrallo that estimates how much vram your avatar uses https://github.com/Thryrallo/VRCAvatarTools/releases
Thx
Reminder to not use shapekeys for toggling things on and off when they can be weight painted to a single bone instead
so would it be more optimized to hide like blushing and other sorts of anime expression enhancers in Bones
For expressions shapekeys are fine, it's low poly geometry that's isn't constantly active
My case is to use the shape keys to change clothing shape not hide.
yeah, that should be bones then
there's an author on booth that makes models that uses bones for all their expressions, model has 1000+ bones 
the expressions look amazing as a result though
I know Blendshapes do have a pretty decent impact on filesize generally the more you have but I dont know what about a blendshape makes it take up more or less
Ive had someone delete blendshapes and go from a 29.9MB fbx to a 3.3MB one but they must of had some broken one or something
unique position in space per vertex
so generally the more vertices blendshapes control the higher the filesize?
(its for setting up new visme)
I have one example...🥴
Thry's tool shows that as VRAM usage
That's way too many bones lmao, even AAA game rigs don't have this many 
Gotta simulate every muscle fiber of the face 😔
In a toon model 
Yet to see anyone use that model , its on booth, really hope they made it into blendshape then remove em after 🐇
so that explains why some models I have got like a million bones
This would perform even worse than if you just used Blendshapes
That was not what i ment, create gestures into blendshapes by using the bones, when got enough what they want remove the bones after (really have to spell it out for you)
Some xna models come with a ton of facebones like that , move bones save as shapekeys, do a few, remove em after
This explains why CATS has "Pose as Shape Key" option when in Pose mode, in order to making shape keys using this method quicker.
I suppose those people don't know about proportional editing?
Bones are better as it isnt destructive editing like proportional editing
also proptional editing isnt really gonna cut it well for the mouth expressions
Shape Keys aren't destructive. Basis is always where you started
I did full Vive Face Tracking for my avatar using proportional edited shape keys
that seems difficult
I do actual game programming so it's something I do often
That's not what I mean, if you move vertices with proportional editing you won't be able correctly put it back a easily as moving a bone back
Blend To Shape in Blender resets a shape key's selected verts to Basis
Whatever floats your boat dude. Just telling you why most game studios use bone facial structure. Better for most uses, easier to animate
I'd use the bone facial structure but just doing it quick job with proportional is easier and quicker
I mean, I feel like using Proportional Editing on a few verts is easier than selecting 50 bones to move the mouth =P
The mouth is at most a dozen points? Can't remember the correct anatomy off the top of the head
If you ever see someone animate a face with good rigging you'll see why its just better than proportional/blendshapes
I would also imagine this bone process would add on a LOT of work for the rigging portion of the pipeline, which would not be ideal for a professional game company with a harsh deadline
Facial structure is a mandated part of the rigging pipeline, if the character has facial animations they will have a facial rig
It's just how it works, I've only seen one or two games use a blendshape approach
Nothing like what's in the image you posted I've ever seen
I'm not talking about whats in the image, that's an overengineered nightmare made by a hobbyist. I'm talking about actual game assets
big studio movies have rigs like the above, but that's because they're worth millions and can't afford to look bad
Oh yeah movies do some crazy stuff. Videogames, however, can't afford the performance for that specific rig
Heyo, so um... I'm new (like really new. First 3D character model ever new) to this. But, I've been making a personal model for VRC I intend on just having as a personal, non-clonable model.
I've been using the CATS Blender plugin for some extra ease in making this specifically for VRC and I'm at a point where I'm pretty sure I'm nearly read to export into Unity, but, I heard somewhere as well that there's a stipulation for texture optimization in Unity where it only allows up to four model textures as a whole? Is this true? Is there any intuitive way I could merge all these into only four texture images? (I have the actual body mesh masked off underneath where the clothes are on the body so I'm guessing that makes that bit easier)
'Cause for the time being I have all of these I'd have to manage to either merge, or disable (I'm guessing, like for the arms and torso for instance, since those are covered by the hoodie.)
there is no such thing
TL;DR is this necessary? Do I need to merge these down into four textures? Is there any particularly intuitive way to do so?
the closest thing I could think you may of heard, is that you need 4 materials (I guess 4 textures then) for your avatar to be considered Poor perf ranking on Quest
Ohh... so that only applies to Quest baking? :T
it isnt a requirement for the quest although you should atlas anyways for both builds
4 materials is a good goal
anymore than 10 and you probably just didnt want to bother or dont know how to make it any better for avatars
I personally have an Index so I'm not too concerned with Quest optimization. But yeah, how would I do that? Cause... That's a lot of textures atm.
i wouldnt use the CATS automatic atlasing tool I hate it because it tends to leave alot of empty space on the atlas thats wasted and sometimes it tries to fit more than it can on a given atlas that causes some textures to lose resolution
I usually manualyl do it with photoshop making a 4096x4096 or 8192x8192 img and dragging each new texture as a new layer onto it until i arraged all the textures i want
and then when I take it back to blender I have to move around the UV mapping so that it fits this new atlas
Ohh, is that why there's so many? 'Cause realistically I only see about 8 applied textures on this model.
idk what u mean by why there are so many
That's from the CATS atlas screen.
But like to be fair, all the M_AMBs easily can be merged 'cause those are all separate pieces that all apply to the same hoodie mesh.
I was asking 'cause you said this:
the cats tool isnt why there are so many
it would still atlas and reduce materials and textures
does a shitty job at it in my experince though
Just trying different methods. CATS atlas managed to merge all those seemingly properly into six materials in this atlas.
Make sure the same texture isn't being used across multiple meshes
so im trying to bake the textures using cats blender but it says no meshes found?
Wdym "bake the textures", do you mean atlus them? Cus atlusing is under optimization
You can bake things like lighting or higher detail geometry into textures such as normal maps for lower poly models
like that's basically how all AAA game character models work is a stupidly high-poly model baked into normal maps of a more low-poly model
Normal maps are truly magical (until you learn and understand the maths behind them)
Normal maps are a godsend. Esoecially if your using a shader that can merge multiple. Same with detail textures. Both save a ton of polycount and texture space
how do i know if my model is full body tracking compatible without the device kit?
Trying to texture the eyes of this model and I turned on symmetry but it seems like other eye isn't taking in the texture correctly?? What can I do??
the uvs have to be on top of each other
for the parts you want with the same texture you're painting
Really hard to do when the texture I'm trying to put on the model gets separated into different parts of the UV
How the hell are you unwrapping the sphere? I would just have a single seam from the top of the sphere to the bottom going down the back of the eye
Also if two meshes are identical they will unwrap to the same position if you unwrap them individually
Ah I see, sorry Im a huge beginner so this is all new to me. The eyes are one mesh should I seperate them and then seam them??
No dont separate them, just select all of one eye, unwrap, then select all of the other eye and unwrap it
how do i know if my model is full body tracking compatible without the device kit?
there are cheap ways to do a jank FBT for a test, Driver4Vr is only like £15 and you can use a webcam as a camera for testing as an option
or if you have 5 phones with gyro slime vr is a thing
xbox kinect can only do 180 degrees and has lots of latency ,so i'd go against doing that
but i cant stop anyone,its just my opinion
remember testing != general useage
wouldnt recomend useing kinect/markerless camera as a primary FBT solution
its good for a quick test tho
true
like im making a model and want it to be FBT compatible, i'm good if i just follow the instruction on vrc, or i need to test it with the kit?
you can just risk it yeh, and aim for it to be correct
i need someone to test it for me lol
Hi, can anyone say if vroid avatars come optimized, or if I want my avatar to be visible to everyone I should do it entirely in blender?
there's plenty of optimization you can do in vroid itself
as it has a button to combine material and delete any completely transparent triangles so you just make all the body under the clothing transparent
they just realeased a new big update too
I was trying to upload an avatar and this happened I dont know how to fix it 
I think this is from Poiyomi Shaders, go to each material and manually lock them, then try uploading again
Ok ty
who else good gang
Excellent gang is superior
Maybe once avatar dynamics is here, can't go with absolutely no dynamics
With a male avatar I have the 'animated dynamic hair'
0 dynamic bones and the hair still moves
It's probably a .anim file.
I could probably just animate the hare and on my blend locomotion
and then give it an idle animation
maybe I'll do that
i could also animate my bell based on angularY velocity
Stop giving me ideas I'll never be done NOOOOO
That's what it is essentially
i cant i need sounds
i have plans for the other 2 sound slots
while staying good
Why do you need sounds
Maybe I'll make a Excellent one without purring
Any sane person has avatar audio at 0%
Purring, Bell Jingle, and in the future keyblade slash SFX + keyblade spawn SFX
I'd sacrifice those for excellent anyday
No star then
I'm optimal enough
Technically the stats mean nothing really
Audio sources count to performance because there is a maximum of active playing audio sources. It might seem negligible, but with 30+ people it will eat into performance, since AFAIK audio still runs single threaded.
does material swapping increase material slot count?
Nope
neat
what is better, some shader or a single spritesheet particle for displaying animated textures?
can you do hue shift on quest? I have a poiyomi puppet to change hue of a material
no hue shift, the closest thing allowed is to set a bunch of Standard Lite\Color parameters in an animation and cycle through them with a expressions menu float radial puppet attached to an animation state Motion Time
horray for answers that are more jargon than english buh
thats scary
i never ever heard someone say the word 'jargon', just learned about it today and then someone just uses it
im trying to follow a guide which should fix like colliders towards body / arms, only issue is i dont have a secondary thingie in my unity under my avatar
Does your hand go through your chest? Well do I have good news for you! With Argama's cureall unity fix, you will have more solid boobles in no time!
★ My other channel: http://youtube.com/argamawitch
★ Support: http://www.ko-fi.com/argamawitch
★ Twitch: https://www.twitch.tv/argamawitch
★ My store: http://etsy.com/shop/argamawitch
★Twitter: ht...
its pretty much this one
she has this
i only got 'body' not secondary so my question is
how do i get this to work, like can i just add the scripts to 'body' instead ?
by scripts i mean the spring bone
This person who did this tutorial doesn't really know how to do VRChat stuff right lol. Do you use Dynamic Bones?
i see yeah, i mean well to me it rather seems that the invidual forgot to generalize it and made it specificly for specific set models with set names / groups etc, i kinda gavve up on it tbh cus it messes my char completly just trying to and i think im missing alot of bones ? im not sure but
my body structure bones is like, ''chest'' and no children under it kinda, like no left boob or right boob or how to say :p
it has arms in details but ye
so what i did with my models is that i started them through vroid like 5months ago (before some big vroid update that just happened recent) and then i gone through blender to use CATS to fix bone structure and visame and so on, to then go into unity and then my bone structures are like this, tho it doesnt mess with my char like in VR headset my char works just fine ( few things that could be better ) but it works, like there is no dysfunctional bone movement etc like that
few things i want to fix or learn to is, my feet keeps gathering together so tightly it looks weird when i stand sometimes, when i lay down my hands clip through my body ( tho is not too much like i can live with it cus its mostly just if i have my hands down ), and another thing i found bothering is that
when i stand up it feels almost as if my shoulders need to be more forwardly for my arms to not clip through so much in standing position with arms down, but im not sure if its just me or not, from side it also kinda looks weird cus like half arm clips a bit
The tutorial there is only to avoid clipping with dynamic boobs with arms, not clothing. Clothing clipping should be fixed in Blender using weight painting
The legs and arms are a common issue with VRoid avatars, and can be fixed in Blender
so since im not using a female char anyways is kinda not relevant, i see, so i perhaps should rather look towards weight painting my char i guess
i see yeah especially with the feet being too close, its something if i move <- and -> with legs a bit and then angle them a bit it should fix the feet gathering too close ?
So I have a question, so let's say I have a avatar with clothing and weapons that I want togglable. Is there a way I can make them toggle and not be different skinned meshes?
yee u can make bones that have the sole purpose of showing or hiding a weapon. For example, a bone that has a sword max weight painted to it, and when its scaled super far down its invisible inside your hand/arm, but when its scaled up proper its on
blendshapes also work but they can get cpu intensive i think
AHHHH the bone way actually makes since. Yeah I figured blendshapes would work I just wanted to see if there was a more optimized way of doing it. Thank you!!
Why did it do this? How do I fix it?
e
any ideas ?
is like my texture is too small for the shoes tho it just seems very odd because i have another project with the same model in same height and same feet etc etc, where they are just fine, i even tried to drag texture from that project into this, but still this same result
im using opaque aswell, no transparent or any
- fixed myself doing adobe photoshop stretching the stuff under the shoes, tho i still wonder how this happened
i see avatars with trails on each of their fingers, is it possible to have this without having 10 trail renderers taking up 10 whole material slots?
er- or is there a way to do this without tipping my avi over the medium perf rating?
my avatar doesnt have crouch/prone animations. ive encountered this issue before but i never figured out how to solve it
Yes but it's very involved and complex to do. You're better off just using the 10 trails
@lunar stirrup so im having issues with the crouch/prone anims
i've encountered this issue before but never fully figured out how to fix it
Damn I knew a fix for that
Oh yeah
Fingers
Or just bones in general
If you’re missing an essential bone you won’t have full IK
i see
my avatar doesnt actually have fingerbones
or well fingers
i didnt think this would cause issues when trying to crouch/prone
You need finger bones yeah
You don't need finger bones for SDK3 avatars
this is not sdk3 :(
Yep exactly the issue lol
Well then there is your problem
ok so ill dummy some finger bones in blender
so maybe adding hand to the rig is out of my reach
i have no clue how i would go about attaching hands to the other armature
bit of a headscratcher
wait, i think i did it
yes
I need help combining all the textures on an avatar into one material
Currently there are 13 materials and to optimise the avatar really should condense them down to 1
Blender, bake or matcombiner (dont recommend cats one it gives odd resolution ones last time i used it anyway)
which version of VRoid are you exporting from and how many layers of hair do you have?
in CATS you can go all the way from 8 materials down to 1 if you generate a 2048x2048 texture atlas and setting Face to 100 on polygon reduction is basically a free 2K polygons saved
even at 100, the face mesh still has so much detail that each polygon is less than 2 pixels in size next to the camera
using CATS to generate a texture atlas and combining every material allows you to preserve all of the detail while cutting down on draw calls, which is a straight shot to
avatars
but does it run better than 10 trails?
Is there anyone way in-game to check how my fallback avatar looks? I've done a couple fallback and it would be nice to not rely on other people to tell me if it turned out ok.
I plan on going back into vroid at some point and atlasing the face if not on my own, then via vroid. They have tools to drop the number of materials
W h i c h
V e r s i o n
The latest version-
it’s easier to use CATS to do it because then you get a higher-quality texture atlas
You could in theory set it to be public for a minute let someone clone it make sure they set it as fallback and use safety settings to view it as the fallback?
or you could upload just the fallback avatar using the "Build & Test" option etc
Somewhat. Overall I believe so but like I said, it takes a ton of setup and isn't really worth it since it only improves by a small amount
Huh that's true, I'm not sure why I didn't think about the 2nd part whoops, thanks for the help!
Just wanted to start by saying sorry if this is the wrong channel. Hey guys, recently I got a VRC Avatar for the first time that was pre-made. Going into it I thought there would be some documentation to help me get it into the game, and once I realized there wasn’t I got a friend to help me out. I have a version of the avatar done to my liking but it has a problem, the head sinks into the character and like into the ground. It’s super screwed up for some reason, contacted the artist a couple days ago and still haven’t received a response to fix it. I also tried the same model in its original state with no modifications on it like glasses and it had the same issue. Again I don’t know anything about modeling, barely understood how to even put it together. So if someone can help me out it’d be appreciated, I can send screenshots etc but since there id a no redistribution clause in the artist’s ToS I can’t be sending it around. Thanks y’all.
Feel free to shoot me a message as well if that makes it easier.
you'd probably want to post your question in #avatar-rigging, this channel here (#avatar-optimization) is to improve the performance of avatars
first of all, i would check first to make sure your view position (in the VRC avatar descriptor) is set up properly. else, i would ask the people at avatar-rigging and maybe provide a couple of screenshots. second of all, there is plenty of official documentation to get the avatar in-game which i dont understand why the artist didnt include a readme linking to it or a basic description of what to do to get your avatar uploaded. you can read it here:
Will do
Requirements We're going to need a couple things before we can get started. Current Version of Unity. We strongly recommend using Unity Hub, available on that page.VRChat SDK - See the Choosing your SDK page to see how to choose the correct SDK for your use. Step 0 - Installing Unity Check the Curre...
Before getting started ensure you have a Unity project with the SDK set up. Step 1 - Find a model Arguably the most important part, you must find a 3D model to be used as your avatar. As this is your first avatar we recommend getting one from the Unity Asset Store as they usually come fully rigged m...
Enabling Expression is a fundamental goal of VRChat. Allowing users to create, upload, and share their avatars with others is a key part of that. There have been some minor updates to the Avatar system since we launched on Steam back in February of 2017. Some of these updates have enabled a ton of c...
but you really only need the 2nd one right now if you already h ave a project set up
also check out this if you need video-guides instead of text guides: https://www.youtube.com/watch?v=dTlZXDJueks&list=PLt8qiEFqNHFkPqpKLnVRx-G78mGvjPbFt
Welcome to my updated tutorial on the basics of uploading avatar with SDK3! This video goes over the simple aspects of getting your avatar up and running in SDK3, including eye look, lip sync, and view point. My previous tutorials were lacking a bit, and i've learned a lot more since then, so i'm going to keep making updated versions of the fir...
Sure, sounds good. Before moving to the other channel I’ll quickly provide a screenshot. This is really helpful so far though thank you 😭😭
It really probably is just the view thing but I remember my friend mentioning that and we did something to the model and it didn’t end up working. I remember he said something abt a Prefab tho.
Model was done in Blender from what I’ve gathered so there’s that too but yeah. I’ll take a look at this stuff and send a screenshot asap.
Yeah so here's what happens to the model, makes zero difference if I'm in VR or not btw.
Oh its ghostlys shark. You'd might want to check the help channels on ghostlys discord. I dunno what it is about that av but it's super allergic to being modded for some reason, that isnt even the worst collapse I've seen. Normally it works fine uploaded straight from the scene file tho.
Will do, yeah that’s really strange 💀💀💀
Just gonna join that on my other acc or some and figure it out, appreciate the tips guys!
Hey, so how do you guys store your avatars? Do you make any changes after optimizing them, or do you keep the detailed version, edit that, and optimize every time before updating?
Or keep separate detailed and optimized versions of meshes, so you don't have to optimize everything for every change and delete and merge them before exporting?
What do you mean by "optimized"
Keep everything in the same .blend file and use collections to organize stuff
I was thinking polygon count, that takes the most time to reduce since you have to do it manually if you want the avatar to still look good.
Oh, whether people start of sculpting in blender or zbrush, they'll keep the high poly mesh for detail and then bake that down onto the lower poly retopologized version of it
and folks would probably be doing any modifications like adding clothing/accessories on the final low-poly version
Fixed it btw
Polygon count limits aren't as severe as some years ago, most models should be fine without decimation now
unless you're trying to make models for the quest which basically has the old vrchat limit for PC as its current limit
Hello, so I got an avatar from someone and when I try to upload it it says there is 72 materials but I have no idea how to combine them or if i should delete any of them, can anyone help me with that?
Can you import it in Blender ?
I tried to but i don’t know if I did it right, and once I did it looked really off
Idk if it’s because of me or how the person made the avatar though
I did get it in blender at one point to combine the materials but it said “no materials combined”
So idk
Blender, in the material tab
But when I bring it into blender the whole model gets messed up D:
messed up how ?
Soooo I bought the avatar custom made and when I bring it to blender to try to make it have less materials it gets like a dress or something and the face gets messed up
Idk if it’s my fault or if the person who made it didn’t do it right
if it was a unity package then yeah it will more difficult to optimize
i wouldn't worry about too much tbh, it's unlikely that you'd be able to optimize below very poor given how low performance it is by default
So is it going to run poorly for other people
Also I think I found the problem? The hair has 52 different material elements I guess
So like they are all seperate materials
Even though they are hair lol
or do you think it’s not a big deal and I can play fine
are you able to contact the person who made this
ここって日本語でも大丈夫ですかね?翻訳したほうが良いですか?
Should I translate it? (Not at good my english)```
yes
that's definitely terrible
How do I fix it though 💀
can you show what it looks like ?
this is the avatar
which is correct
and it looks like that in unity
buuuuut
this is blender
the character has a dress and stuff for some reason
and weird textures over the eyes
idk if its my fault or what
and when i bring it back to unity it has the black dress
@calm spade
also i did finally get the avatar into vrchat
but the knees bend into eachother kinda
and again i have no experience or anything and i messaged the person who made it for me
You should probably ask the person who you got it from to help you out, cause that's an absolute mess lmao
way too much work for a beginning to get into right off the bat
how the heck do you mess up the normals that bad
sry
I don't mean like you personally I meant like the person who made the model
especially since that appears to be just a VRoid Studio model
yeah they put actual effort into making it terrible
It doesn't look bad at all, it's just awfully optimized
I asked them for help with it
Definitely the first and last time I get a custom avatar I guess
you could try asking them for the VRoid file
as that would be better anyway to have the source file
as that makes it easier to swap clothing
All I have is a VRM file
I’ve been looking for the crisscross sitting animation
This should help you out with all your issues! Sorry the person who created the avatar didn't give ya all the info https://www.youtube.com/watch?v=Qbwl9XIONuQ&t=1357s&ab_channel=PHIA
My beginner VRoid to VRChat videos using the VRM converter in Unity 2018 were exceptionally popular, so I have decided to make an in-depth tutorial for those who are looking to make a PERFECT VRoid model :D Using minimal blender skills and Unity 2019, this is an up-to-date tutorial on how to fix your model to perfect. This video includes fixes f...
👍
is there any easy way to convert a pc avatar to a quest compatible one
Feilen made the bake addon for Blender
can anyone help with materials being weird after CATS fixes the model?
i import my model and then match each material to its image texture and things look the way they should:
but after clicking "Fix Model" some parts are grey :/
is there a way i can optimize this tail to be merged to the avatar
how do i go about optimizing these bones together to be merged , is there like a video tutorial for this
You should just need to keep the tail bones ?
What is the large bone above the tail supposed to be ?
Probably some sort of leader for the physics bones
Probably like make it just 2 bones and change it so the 2 bones will effect it so it is a smooth transition for it to bend smoothly
Geez now I wish I could make vrchat avatars. I tried taking a model of my favorite character and slapping it into unity but I couldn’t figure it out
Same thing with maps
I don't think the tail bones need to be changed
Oh ok
I want that as an actual avatar that way
Looks funny
use Fix Model before applying textures
wassup
Is it better to keep an avatar's head separate from it's body?
It helps to keep the amount of verts being moved by blendshapes low, so depending on the model yes
worded a lil funny, even if the vert doesnt move it is still taking up resources in regards to blendshapes
I need some serious optimization
Uh how does the avatar look like?
🗿
It probably doesn't have that many, the VRC SDK is extremely glitchy and sometimes shows an insane number
A model of that size would be in the gigabytes easily
Read/Write Enable needs to be enabled on your import settings of the model
Does anyone know how to optimize one of the ghostly avatars?
Is the error for "We said no, there is an error". In the console tab click clear and see what stays. If its empty press the play arrow and the top of the screen and itll light up blue and just click it again to exit play mode to get the real error
It's not clearing
and its still the only error?
There were a few
46 phew, double check you dont have an assets folder inside your assets folder
i do
thats breakin everything cuz you have duplicate scrips
like a assets/vrcsdk folder AND a assets/assets/vrcsdk will make assembly with name errors
yep, that is exactly what that error is, you have a copy of a project inside of your asset folder
when i press play after it cleared it said no cameras rendering
eh, just have the camera disabled. no big deal
it still wont show the upload menu
hmm as long as there is no more errors it should work. Mightve broken something with a duplicate so maybe delete the vrcsdk folder and reimport the latest

