#avatar-optimization
1 messages · Page 10 of 1
for worlds* ^^
now if they make a blit component for avatars like we asked in canny for years, that would be great ^^
🙏
Hi! So, I have multiple knifes on my avatar, and Id like to use the same equip sound for all of them for optimization reasons. But with my current setup, the knife with higher priority in the FX menu keeps the audio source disabled. Is there a way to fix this? Thank you in advance! 
There's already enough audio sources on the avatar, and I don't want to add any more if not needed 
You can't really trigger the same thing from multiple animation layers reliably. I'd probably put all those into one layer which branches based on a "which knife?" check rather than all separate.
if I'm understanding your current setup
Damn, I never thought about doing it that way, no clue why I decided to put them on separate layers haha. Give me a minute, I'll try to jam them into a single layer.
yeah if you'll only ever use one at a time, it makes the most sense
Yeah no, no need for multiple knifes at the same time.
So know I made it so
**Knife1 **is Whichknife = 1
and
**Knife2 ** Whichknife = 2
But how can I reset Whichknife to 0?
@cold yarrow
I'm assuming you're just toggling that to 1 or 2 via in-game menus? It should revert to the default in the VRC Avatar Parameters list
Currently it looks like this, but it isn't reverting to the default value
Yeah, I didn't realise that if you use the Toggle function, it reverts it the default, without having to specify.
Theeeeeere I go.
@cold yarrow I appreciate the help, more than you can imagine.
Dude, the amount of times where I stare at a thing that should be working and it isn't, then I realized I forgot something is practically infinite at this point. We've all been there.
Usually just saying the problem to someone, lets me figure out the issue haha.
Pretty sure I can imagine it 🙂
Rubber duck debugging ftw
Best way to troubleshoot anything, I've proven this to myself so many times haha.
Know I just need to figure out how to disable knife1 when enabling knife2
.
In that diagram you have ^ it'll be automatic if you're using Write Defaults = on
But I heard that using Write Defaults is like a major sin when it comes to VRChat avatars.
Yeah there's lots of things said about both sides. Pick one and stick with it. So if you have it off, you'll want a "disable knife" animation before you get back to the "wait for trigger" state at the beginning.
I just leave WD on and don't worry about what people say 🙂
I've built this already around WD off
.
But where in the diagram would I add the disable knife animation?
Since this already disables everything.
hmm I'd think that would work then, if that disables all of them
It's suppose to ye, but when I have Knife1 enabled and I enable Knife2, it kind of breaks.
Seems like the only path from 1 to 2 is via the "all off" state
So all is good when I enable the first knife.
Pressing the second, just changed the grip.
Is your condition for 1 -> all off "if WhichKnife != 1" ?
Think so.
no, the other one, the one that goes back
Oh, that's WhichKnife = 0
it should be "not equal 1"
You're right.
I forgot NotEqual was even a condition...
Aaaaaand it works now, you're a lifesaver.
Thank you so much again, enjoy your lunch! 👏
nothing outside of mobile shaders :^)
I think the question was more along the way of if there is a quest shader that can do metallics decently
@robust anchor answer is no, because mobile shaders is the only option @umbral echo
And that was my question: which of the shaders that can be used on quest handle metallics best
i dont know if i understand this correctly
there is only 1 shader for quest
the mobile one
you cant use any other shader
so that is technically the best as there is no other option
There are multiple shaders in the VRChat/Mobile folder, that VRChat permits for quest avatars last I checked
how i make an plushie have physbones and move around like some avis have? kind ahard to explain but some have little cats on heads that jiggle or in their hand
ah, apologies - i misunderstood. i do not know the answer 
Does anyone recommend any means of decimation/poly reduction as to give myself more breathing room to add models to my avatar? I'm sitting at an uncomfortable 18k tris (with a base model for PC turned to quest). Any tips?
the plushie would need to have bones in itself to attach a physbone to it. so if it was a bunny, you would need bones in the ears and tail and legs and stuff
how would i add bones? @tame summit
in blender
is there a turtorial?
probably not for the specific situation you’re in, but you’d have to learn weight painting and stuff too if you’re not familiar with blender
i would recommend just looking for a plushy asset that already has bones
i ahve no idea where to loom for them
you can check booth or gumroad maybe sketchfab has assets for that idk
Can I see the model so I know how much wiggle room you have in terms of how to reduce the poly count?
So far it's a base blended together by the CATS blender plugin
Even if I try to decimate it in blender, it begins to look unrecognizable and unpreferable
a base what, human?
Does anyone know the easiest way to make this get down to 10.00MB? It's at like 10.41MB and I'm pretty new to Unity so idk what I could delete or take off that won't break the avatar-
reduce texture resolution
texture = those png’s in ur assets
click them and reduce resolution on the right
aka inspector
Oh ok thank you! I'll try that out
is material count or polygon count more important for optimising avvies, specifically for Quest? The only part tagged as orange for my avvie is materials, which it uses 2 of, and I really want to make it usable as a fallback if its possible but I feel like I need at least 2 materials for an effect I really like on it. Its polygon count is just barely over 8,400
Hard to answer cuz quest is a weird one, with less load on the gpu (polys), its thermally allowed to be in a higher power stage on the cpu (mats+vrc)
And vice versa
not smart enough to really know what that means tbh... x.x
Its got 1 cookie jar instead of the 2 a pc has
ohhh ok :P
Just try your best to keep both as low as possible cuz they both come from the same cookie jar
(up to a hard limit)
so, if I could manage to reduce polycount enough, can still have "good" ranked performance even with 2 materials in use?
Right now Im at Medium performance on Quest btw
Itll be in spirit, the vrc avatar rating is very... Stiff
Oh, is it hardcoded to read is as Medium so long as have 2 materials? Even if its pretty optimized otrherwise in reality?
Unless maybe we can come up with a 1 material solution for your effect
I have a glowy effect on some parts of my avie (the eyes and some attachments), which use a 2nd material with an emission.
All the non-glowy parts use the same material
like this
Id have to check available mobile shaders, but does this one use a 2nd texture to define the glowy bits?
The glowy material does use a completely different texture, yes
I can open up my Unity, one sec
the 2 materials in use atm
having anything medium is going to make the Avatar always medium even if the rest are good @jagged lark
and honestly it's not the end of the world being medium considering that's better than most Quest avatars
tbh its probably pretty unoptimized, its just a pretty simple model too :P
This was my 1st major 3D project so don't know a lot of optimisation techniques and the like
yes so albedo by isself will make the effect of toon lit, and than a texture that is just the eyes and other glowy bits can go into emission
its possible to combine these materials
oh really?
Does the emission only apply to parts where the emission texture isn't, like, transparent or something?
like itll mostly be the same texture, just with all the parts you do not want emissive painted black
ohhh ok I guess I can try that out then!
just, uughhh hope can merge materials easily this time around... last time tried to dew it with CATs all the textures broke, so had to semi-manually fix the atlas x.x
its one solid color so if you wanna be lazy can just throw those two UVs into one tiny pixel and manually paint it those two colors
Atlasing caused this to happen, specifficaly
or- just put the UVs where the yellow and green already exist, it kinda looks like the eyes are already on the atlas
yeh, they were originally, moved them to the other material caus wanted them to glow too ~
mmhm, easy texture to make. in gimp or whatever select the glowy bits, select invert, paint bucket all black
boom, down to 1 material!
awesome! Will try it out as soon as eat 🍔 ~
have fun optimizing!
The way I see it, is if my avatar is medium, it's a hell of a lot more optimized than some if not most avatar people spend like $40 - $60 on
The main reason I try to optimize my avatar so much is because I can then cram in more props and other fun things 😂
@proper grail sorry for taking so long, but uh smol problem, is it possible to dew Emission with the Toon Lit shader, or otherwise get the Standard Lite shader to look a little more like the Toon Lit one? Ill post a couple images of what I mean, hold on
the Standard Lite shader has emission available, but it makes the polygons visible on the model which I dont really like. The Toon Lit shader stops the polgyons from being visible, but there's no emission option for it
Is there a way to, like, smooth out the surface with the standard lite? I don't think I mind the lighting change itself, but I do want the surface to be more "flat" like how the Toon Lit shader has it. Or, alternatively, add emission on Toon Lit somehow
hmm toon lit ignores normals, will be a blender thing but just see if you can go to the fbx import settings and change smoothing from import to calculate, and set the angle to 180 for smoooth
just to not open blender again
I guess thats a little better? Still looks off to me though...
If this is the closest can get to re-emulating Toon Lit, might just want to stick with the 2 material system tbh, at least now that I know being "medium" optimised apparently isn't all that bad actually
Still, I'll build & test to see if it looks a bit better in-gaym
yeah, itll look strange in unity as toon lit doesnt really use lighting or normals. Just a static color.
yeh, idk, still feel genuinely mixed on it
on one hand I dont mind it too much, on the other I hate it and really prefer Toon Lit for the main body
2 materials it is!
yeh, jus tested in VR too and donut lime it at all lime tha :P
Well, thanks a lot anyways though! At least I learned that selective emission with 1 material is a thing, Im sure that can be useful in the future ~
audiolink!
Material > Polygon count until a point. but its all depended. as more materials means higher cpu,gpu,vram useage.
ideally keep material count at max 4 if its not possible to keep it at 1.
anyway i can bring the MB size down without making it look like dog shit? (Uploading For Quest)
try lowering the resolution of the metallic smoothness map and the color map
and have you tried the crunch compression setting @grave hull
Yes
I’ll try that
mainly the reason I'm suggesting those as it mostly the normal map providing any details to your model whereas the color is Big Blocks of solid color
So tried turning the down and somehow it made it go up wtf?
weird
high compression stands for high quality compression, yeah i know
quality as in keep more details and be bigger in size
I have an idea just don't have a pc or laptop can anyone help
an idea? You could always commission someone.
Enable read/write
yeah I know, just always think its kinda funny lol
Ah yee it was shocking to me when i first saw it too lul
what's the reason behind that dough? Why does having read/write off make the polygon count skyrocket? Is that number just the max integer value Gluenity can count to?
ngl i expected it
but yes it deafults to all 1s except the last byte to avoid NaNs (not gonna write the bytechain again)
because it cannot read the mesh. so it just defaults to max int value. since there is no way to determine how many polygons without being able to read it
Idk what I needa do to opti it also now I can't even click the publish button (making it quest) so help pls pls
Optimization:
I'm about to put a 39k Polygons thing onto my avatar.
Ignoring the material, roughly how much size will it add to my avatar (MB)?
size is 10.23, just more than 10mb. it must be 10mb or less to upload for quest.
Does this method work? or is decimation better?
how do i lower the physbone transform count?
have less bones affected by either removing bones or reference the root bone further down the hierarchy
Does anyone know if there are any decent Unity based tools that will automatically convert PC avatars for quest?
I have VRC Quest tools, but I still feel it's a little limited.
The model i want to convert is using Poiyomi shader, (which quest tools doesnt support), and has quite a lof of shader related animations (like skin/hair colour radial puppet).
The biggest frustration is with VROID avatars, and invisible textures... which become uninvisible when converted... Meaning i have to dig into blender to carve them out.
Quest is just that limited. you wont be able to do alot. let alone animations.
I have a quest avatar with animations. A little rusk that spawns a car and a few other bits. So defo possible.
Yes but no. Shader based animations (such as hue shifting) is not really possible. There are ways to kinda get around it (matswapping to predetermined colour combinations, or having flat coloured textures and using UV offsetting) but you won't be able to 1-1 recreate it in the same way. Same thing goes for anything to do with transparency, you simply won't be able to recreate that.
Additionally no constraints so any props you have will also be more limited in their functionality and then theres the hardcapped physbone limits etc...
The point being. You won't find automated tools to do this for you. You will need to learn and do it yourself because often times the solutions are too novel for them to be automated, or simply not realistic to be automated. (For example if you want to fix your transparent eyebrows issue, you will need to get the knife tool out in blender and start cutting)
iirc I've seen people do some hueshifting via emissions on quest
Guess I was wrong.
so realtime fluid dynamics are out of the QUESTion, huh?
yea
I have hue shift on a lot of my quest avis using standard lite shader and emissions. Change the albedo color and emissions color.
I have a question mostly because I'm new to creating avatars. How come I was making an avatar and it was a normal model in blender and unity, but as soon as I uploaded it, it turned into an image that filled up my entire screen?
What lol? can you take a screenshot
I'm ready to get crashed lol
For some reason I can't ss
So I'll send a pic through my phone
I don't know why it's doing this
good save
Anyway, I don't know what to do
nvm, I just realized everything was in the wrong place and it was the model creators fault
lol that happens sorry I'm working on a project
fun fact, i used to never optimize my models because i didnt know it was the easiest thing ever in blender, im talking 20+ materials and 70k polys and always verypoor with errors
so this is the most satisfying thing ive ever seen
Go to the Control Panel Settings and enable the option to Show All Stats. I’m curious to see what else you did.
Showing All Stats will expand the list exposing all other Stats that are also ranked Good or Excellent.
apparently i did pretty dang good
was just a vroid model that i optimized in blender
Damn! That’s a lot of Excellent stats.
Well done.
can someone help me plz to opti my skin idk who to do it its my first dl ^^ dm me
Do you need all 21 skinned mesh renderers? I'm betting it's a lot of props. I'd consider starting there.
So I got a avatar I made just recently, I optimised the heck out of it but I believe I messed up the material somehow for the eyebrows? When in blender, I joined the two meshes and I made the same material for the face that applies to the eyebrows. Can someone tell me what I have done wrong here?
looks like the eyebrows and lashes are transparent, very common issue. delete some verts and use edge slide to line them up with the texture in blender
that or make the rendering preset cutout instead of opaque but then it won't be quest compat
Thanks for letting me know!
Anybody know why the texture memory would report a different size in the SDK vs a Thry VRAM tool manager?
looks like shit but at least the questies will see my avatar lmfao
lol
It's an issue that happens sometimes, once you Build and Publish the avatar it'll follow Thry's number. Not sure why it does it, gave me quite a bit of confusion myself when I first noticed
When I build and publish it didn't change ;-;
SDK reports 76mb ot texture memory but i am only using 49mb in thry
make a big difference in textures if I can actually use 70 mb 😢
I'm thinking about adding twist bones to my avatar, does the use of constraints cause significant performance decrease for myself or others?
not significantly unless you're adding like 300 constraints or something
Oh okay then that probably won't happen I'm just looking to make it so my avatars base with move and rotate naturally I doubt I'd even reach triple digits for constraints
atleast not on my optimize version
Yeah that would be pretty weird.
I've come to a grinding impass with Unity and making things quest compatible. I have a bunch of props that I've used Parent Constraint to attach to the avatar, and it works in PC. However, apparently Quest doesn't use Parent Constraint. Is there an easy solution for this or am I going to have to try to jerryrig all of these props onto the armature one by one, and then re-do the toggles that I've made?
Depends on what you're trying to accomplish
Two things:
- I want to have props that are in my hand that I can toggle between (Staff, Baseball bat, glowsticks)
- I want to have a prop that I have attached to the brim of my hat
The hand props can just be added to your armature in the hand and same for the hat prop just add it to the head
which is a massive massive pain, but I guess it's literally the only way. Thanks.
I mean it takes about the same amount of effort as a constraint lol
And is less performance heavy
Toggles are the issue, unless I can drag from the hierarchy into the animation
...I haven't tried that yet. hm.
You can just click on the hierarchy in record mode an just hide it in the inspector
Just the record button on the animation tab
I guess I'll figure this out, thanks
I have to do further cuts on the avatar anyways, apprently one of my props was way too heavy for quest to deal with :sadge:
Lol yeah I have an Avatar like that
and yeah, I don't think I can mess with objects in the hierarchy and move them to the animator with ease, I have to navigate to each of them manually which is just... ugh
the one thing I hoped I didn't have to do.
well apparently having prefabs with parent constraints is... okay now?
Before it was pitching a fit whenever I tried
(nevermind, it was in the scene but not on the avatar 🤦♂️)
Any tips on how to get an avatar down to the 10MB requirement for quest? I typically start deleting clothing, change the quality of textures but idk if there is more I can do
If you know how to use blender there's always seperating the head and body into seperate meshes. Can also remove unused blendshapes in blender or use something like VrcFury which has a way to do it non-destructively
Most optimized Avatar
When combining UVs to atlas manually, some of my materials are screwed up and only show one color. When the UVs are separate, they are fine.
Is this a Blender bug? Is there anyway to fix this? I want to be able to combine this avi into only two objects and two materials.
I guess I could recombine them again, but I don't even know if I can make a new UV map that includes all parts of the mesh without manually combining them
Make sure your UV maps are named the same before combining
I did
i have 8 physbones, how do i fix this..
Sorry to ask this as I feel this would have been asked before but I can't find the answer in the search box. I have a model that I need for Quest, so I've neared the vertices to 10k in Blender but in Unity that 10k becomes nearly 20k, is there a setting I can turn off on export/import so it doesn't double in Unity?
The screenshot on the left from blender is showing vertices while the one on the right in the sdk has polys.
That explains why the keep quads helped with reducing the number in Unity! It did although spilt the mesh into 2 material slots for some reason
if you have tris or ngons, the model is split between quads and others
button really only works if your model is 100% quads
I am just unlucky today
Dissolve like, 3 random back vertices
I've been cutting corners whenever i could and this is where i'm at now, is this good for quest?
This will give you the Poor rating which will be hidden by default.
With that being said it's still better than like 90% of quest uploads
I see, i guess it's fine then?
Should be
Should be
Should be
I'm trying to switch the target to android but it's not changing, is it processing it on the background or is it just not able?
I'm pretty sure you need to play with the materials of the avatar
I tend to paitoon's toon shader
Poyomi*
I'm trying to port this to "android", but clicking on the button does nohting
what should i do here?
I figured I might ask this again since there might be more eyes to look at this... But it seems SDK is reading a higher number for texture memory than is being actually used... What should I do to fix it?
How do you know it is incorrect?
Generally it is showing a lower amount than other tools (like https://github.com/Thryrallo/VRC-Avatar-Performance-Tools) because it only counts textures, which excludes all the other stuff taking up VRAM.
Reason why is because my texture memory is REALLY low and I have run other avis with as much assets and didn't have to lower the textures this low
How many textures do you have and and what resolution?
I swear if you're using a 4k texture for anything other than text
What are the easiest ways to make an avatar 10MB? I got an avatar for VR models and it says it's at 11. Something. Could I have like a lil list of things that I can do that'll keep the avatar still looking alright? I don't wanna go into this without knowing anything and break it
A common and easy way is to reduce the texture resolution of your textures in Unity. This can be easily reversed if the quality is too bad so you can just experiment with lowering it safely.
How would I do that? I have the MAT opened up and whenever I click on one I'm not sure what I do from there
If that's even the right area-
question: how does the animator layer controller work?
If you have the material open there'll be a texture icon in the shape of a box or something similar you can just click on and it'll take you to the location of the texture.
Oh ok I got it! Thank you
How do I fix this? I don't know what this means 
Most optimized avatar to exist
Bone hierarchys are in alphabetical order, its just saying to have the first bone parented to upperleg to be more towards A than towards Z
Its for ik reasons
Hi, I’m new to VRChat and I’m now trying to create my own avatar but I have a few issues and I really don’t know how to fix it. It’s my all time first avatar and I’m trying to learn things. I only use VRoid and Unity until now. Would be nice if someone could give me some advice to fix issues like that.
if you are uploading for quest, theres limits to how many materials you can have. i think the limit is 4, so you'll have to combine materials/textures in blender with the cats plugin, or material combiner. you also have to reduse the amount of physbones on your avatar as well i think max for quest is like 8 or something like that? and mesh's i believe is upto 4? but usually you want to keep it at 1 mesh. if its for PC everything should be ok as long as youre using shaders that are VRC supported. you rmodel will just be Medium level of optimization. for more info i'd recommend looking at the preformance ranking system in the VRC Docs: https://docs.vrchat.com/docs/avatar-performance-ranking-system#section-avatar-performance-ranks-value-maximums-per-rank
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 perform...
also, you only want one avatar descripter on your avatar, i can see you have 2, one on the body and one on the face. from what your sdk is showing
Aaahh thank you, I think this will help a lot! I make it for PC 🙂
I don’t know why, but when I don’t add the avatar descriptor to the body too, the the builder tab won’t show anything, it’s just blank then. I will try again tomorrow 🙂
usually you should add the SDK to the top most part of your avatar, its usually the object everything is parented to/ falls under when collapsed so that it knows where everything in the avatar is
Is this crazy important? If I say yes will the avatar still be good?
It is not crazy important. Short version is: its safe to say yes.
Long version is, it's not a base Unity or VRChat thing, that's from the Thry shader inspector that you probably have because its bundled with Poiyomi's. In some lighting setups, a mesh will sample a specific position's light then use that for the whole mesh. If you have multiple separate meshes, like for add-on props, these points might be in different places and so these add-ons could be sampling different lighting conditions and look wildly different. Saying "yes" sets them all to the same place, so lighting is consistent for all the meshes.
Ah ok thank you
I know I'm asking a lot today I kinda feel bad aha hah but um I uploaded an avatar little while ago to PC/Quest, in game it says it's compatible for both but when I try play as that avatar it's just my fallback. It's like the avatars I have before the update saying the 10MB thing but it shouldnt of lemme upload it quest if it wasn't right? I'm so confuseddddd pls help 🙏🙏
Yuno moment
Surprised someones making a modern yuno too
Need to get my hands on a workable yuno model and see if she has bones in the hair for pysbones
Gonna give her a chainsword with kill particles
If I have a toggle on an avatar as a skinned mesh, can I convert it to just a regular mesh object and still have it parented to a bone?
Yeah should be doable in blender. Just make sure the new seperated mesh is a child of the bone it needs to be parented to.
It‘s me again
I just fixed some of my issues but now I have a few now issues especially with the physical bones. I reduced them and now it seems like I crashed my avatar 
I also can’t give it a test run because if I try this error (Second picture) shows up in Unity
Aaaah i got it fixed and tested her, she works well for PC :3 i just have a lot of work to do in blender for her physics and i want custom animations like sitting. But for now i'm happy that i can finally use her
is this the expiremental branch?
or are you using the old / discontinued cats plugin
The basics model from deviant art are rigged and gave bones in their hairs.
I recomend to edit the bones from the hairs because its not 1:1 on the mesh and weightpainted areas.
( this is only optional...)
Can you pass me the model
deviant art models usually cant be used for vrchat
Unless you know how to fix mmd bones
i do but i say that bc legalities
a new person isnt gonna know how to fix mmd bones
plus i also dont help with ripped/leaked/tda models
fair
not all models with mmd bones are tda models some people but random gmod models into mmd
there are sooo many mmd models that aren't TDA
https://docs.google.com/presentation/d/1yw8NZA37C71Z6YVvmMzZUhteXIj5vOqcFLKr1Ypoyvg Tupper's (great) intro to optimization. Haven't watched the pres yet (only read the slides so far), but you can see it at https://www.twitch.tv/videos/1836679848, about 7:20:50 into the video
(For those who want Ekkosangen's panel referenced in the pres, it's at https://www.twitch.tv/videos/1835895594, 1:47:30 in... haven't seen the material so I'm unsure if I can personally give it a stamp of approval per se, but I generally trust Tupper's judgment regarding things about technical optmization details enough to link it here)
like even just the slides from Tupper's panel was quite helpful
yeah... "material slots and texture memory are the perf-stats-visible things that matter most" is something more people should know by heart
and considering how many people use Poiyomi there's a lot of vram savings people are leaving on the table
The amount of avatars I've seen w/ poor/very poor texture memory and material slots when every other stat is medium or better breaks my heart
pure laziness
ive taken almost 1mil poly down to 60k with only 5 materials, i only see reasons to not optimize as excuses or laziness (every use case is different though inb4 "uhh acshully 🤓")*
I mean you can say that or people just want overly complex models I have 6 fallback compatible avatars but i also like epic particles and jiggly avis with smooth edges
basically don't design models with staring 2 inches away from it in mind if you want to be optimized
Is the only way to have toggles on a combined mesh from blender to have a shape key that scales it down to near 0?
that sounds more like a willing choice with awareness rather than a frankensteind avatar with no optimizing done
aka a lot of avatars
but if the boot im talking about doesnt fit you then dont worry
Most of vrchat is unoptimized monstrosities lol
shapekeys, bone scaling or shader based
you can use UV discard. it allows up to 16 different clothes or other things to be used without the need of more materials. etc and or blendshapes
can confirm, uv discard is magic
although because its shader based it only works if others have shaders enabled in safety settings, so organise outfits around that
yep yep
Would this have to be done with a special shader like poiyomi or would this work with standard. What about quest?
Poiyomi is the only one that can do it as far as i remember atm
Definitely not quest compatible
Before Poiyomi had it I remember everyone recommending Silent Cel Shading Shader (SCSS)
I never used it though so I don't know how many it can do
Silent's does, yes. Something else I saw did too but I forget which. Apparently it's quite easy to implement, so I'm not sure why more don't.
totally thought Mochie's did but I don't see it in the source
Silent's SCSS was the first one to do it, long before Poiyomi implemented a similar solution in his shader. Another shader that has pretty much same inventory system as SCSS is ACLS.
Oh cool, I didn't know that
I did actually go to look at the code, it's really simple
What is the common VR chat modelling workflow that seems to result in over 200,000 polygons? When I make models in blender, they're usually about 3,000 to 4,000. Same with extreme material counts. I usually just uv unwrap characters onto just one map, so I only need one material and one of each type of texture (albedo, normal, emmission, etc). It seems like you'd have to activley go out of your way to make very poor avatars?
a lot of folks using the subdivide modifier at too high a setting and like trying to model every single belt loophole in Geometry instead of letting normal Maps take care of it
Hmm makes sense. Another thing I hear is some models are made from bases on the internet. Are these bases rendering bases instead of game-ready bases?
alot of base models are ripped from other games. which is not exactly a good thing obviously. however for example. nearly all if not all base model creators in the furry community are only 20-30k polygons. leaving a decent amount for others to add their own clothes,accessory etc
If VRchat does not like the performance of the current cloth physics of Unity...
Why they don't use this optimized cloth physic asset, that uses Unity DOTS?
(I made a few screenshots how well it even performs, if you run it on a android phone:
Ig it's because it's not easy to implement things on vrc
If VRChat doesn't like that players use Unity's Cloth Physic for their avatars, why don't they make their own more optimized Cloth Physic system that runs on the latest technology standards?
Or they just add support for this asset, like they did back then with dynamic bones?
Only with the difference that this asset is multi-threaded and use Unity DOTS, so that it even runs fast on mobile...
Tea but the thing is, as i said, that implementing stuff on vrc is hard
And cloth physics are not that necessary from a general pov
I vote for Magica Cloth 🙂
that is magica cloth, but I even would recomend the new magica cloth 2 because its the latest and best version of it.
Definitely!
I mean they don't need to remake it from scratch like with phys bones, just adding support for it. you are not forced to buy it, only the onces that really want to have cloth physics because physbones would work and be enough for the most peoples.
and if you are really willing in using cloth physics, you normally should also have the 24$ to buy the asset.
yeah I kinda hope they don't make their own, it's a pretty large effort to build from scratch when it's already been done well, especially when canny is full of really stupid bugs and missing features
most likely because they have something they are making them self i would guess? and also they are kinda doing alot of in house things now rather then relying on 3rd party apps. that also would halt upgrading the engine when necessory. like we are getting a update to the next LTS i believe it was soon. that has massive performance increases to Blendshapes and such.
also it's most likely also a priority that they focus on things that are more commonly used. such again as blendshapes. cloth physics are rarely used in vrc because its not really required. you can get away with really good weight painting instead.
but if you want something better, you have to deal with the unity cloth physic, even if it is unoptimized as hell.
and there are many peoples that are using them already.
Its like with the dynamic bones before physbones existed.
barely anyone uses cloth physics. its to expensive.
of the performance, or what?
I mean 90% of all peoples have very poor avatars already, so if they are a bit over the limit of poor, they can directly give a F about it and screw their performance down the highway to hell.
but if we would have support for the magica cloth asset, it would not be that much work for VRChat and only the peoples that really want to have it, have to buy for the asset.
how hard was it for vrchat to add support for dynamic bones back then?
It would be mostly the same until they have something better in the future, but for now, would it not be nice to support this magica cloth asset for those where want to use optimized cloth physics?
dynamic bones is just one part of the problem they had to solve. they also had to make sure it would work in vrc.
dont really see the need for cloth physics at all
thats not really the point of it. and i can gurantee you if your avatar takes up 50% of a persons fps or outright crashes a few people you wont be popular. and in most cases just get kicked out of a instance
did you not saw how extreme optimized this asset is compared to the default unity cloth?
multi threaded and Unity DOTS.
I think it would be even fast enough to be supported for quest users.
@marble rain
no i dont think it would work well. due to the quest not having enough procession power.
a quest 2 can only handle about a 1 million polygons with light physics simulation
https://developer.oculus.com/documentation/unity/unity-perf/?intern_source=devblog&intern_content=boost-app-performance-525-mhz-gpu-frequency-meta-quest-2 here it was. and sorry it was Draw calls amount.
I'm assuming even just hooking in magicaCloth would likely be a non-trivial task. If VRC is mostly still running on monobehaviors it may be a bit difficult to figure out how exactly to slot ECS/DOTS into the workflow.
Probably yeah. And there's so much todo with the features they already have 🙂
does anyone know how to fix this?
Either remove some of your physbone components or add some of the physbone affected bones to the exclusion list in the physbone component. Excluding a bone will also exclude all the children of the bone.
mostly because they have other things on their to-do list they want to do first before making their own cloth system
I've seen a lot less cloth than dynamic bones, it's a lot lower down the priority list
I use cloth physics on skirts, dresses and long coats. I’m going low poly so it uses at most 500 ish verts. It’s still gonna rate it poor but personally it give me a more solid result compared to using physbone. Don’t get me wrong the physbones have made optimization massively better, but compared to the cloth physics especially for what I described it has been more straightforward to work with. I wish the limit was at least 1K verts. I hope in the future they do end up making there own optimized version it will help so much
An example of me using the sim; One with dresses and one with a long coat
The second model I’m currently working on so I still need to adjust the colliders
I will say learning uv discard has cut my mats from 6-5 to 2 at the lowest. I can’t go down to 1 since I use the second mat for IOR on glass or front of the eye to give a distortion effect
Im new to avatar uploading and one of my problems is that I have 10 physbones on my avatar but the maximum is 8 for quest. how do I delete them ?
find bones which have a VRC PhysBone (Script) on them and delete that component
you can search for Physbone in the box in the upper left of the hierarchy
got it, thanks !
actually I got one more problem, I have to change the shader of this but its not letting me
it's an embedded material in your FBX file, you'll need to export it. Just find it under the FBX file, select it, and ctrl-d to duplicate. Then edit that, and drag it into the avatar's mesh wherever it should be
ah it may not show up there
you can export them in the rig import though
oh no, now that I look at something that actually has an embedded material, that should work
Anyone got Honkai Star Rail's Kafka swimsuit model from aplaybox? I have it but there are insane amount of pointless bones and it's a pain to optimize so wondering if anyone out there worked on it
Hell
Hope it's alright, I just wanted to share this video because I think it is woefully underviewed.
https://www.youtube.com/watch?v=VnmYM3vkYBk
Fast tutorial for material optimization easily applicable to VRChat.
Get the model I am VTubing in: https://camsaviis.gumroad.com/l/CamsRemi
Check out my Shop: https://camsaviis.gumroad.com
SimpleBake addon: https://blendermarket.com/products/simplebake---simple-pbr-and-other-baking-in-blender-2
UVPackMaster addon: https://uvpackmaster.com
is texture memory usage counter buged?
is because i have a texture toggle maybe
Shouldn't be bugged, lol let me see the rest of your performance parameters in the build window
ok
i believe in game is not counting the switch material since is used only with the animation
Weird
Yup, that's a known bug.
any optimisation tips
Keep material count as low as possible. ideally around 1-4, keep polygon count at 72.000 max, dont use more then i would say 4-8 phys bone transform components, and not more then 32 affected bones. no lights or cloth physics etc. gotta think in terms of how your avatar performance is when there are 40 people.
Wanted to show off another practical example using detail normals!
This one is more pertinent to furry characters where we want the flesh parts of our characters (nose, beans) to have a bumpy texture. Simply using a Detail Normal instead of a 4K normal, we save 20mb!
is there a quick fix for most of these (not the shader tho)
None of those are blocking issues, can still upload and use the avatar.
The LowerLeg one is just a warning about how the SDK thinks your armature should look, it could very well be fine. Can check this by going to the import settings of your avatar's FBX file, Rig tab, and clicking "Configure" on the rig. Take a look and see if there are any issues.
The angle one isn't a concern at all unless you're using full-body, besides that its like the LowerLeg one where it might not be a problem at all.
Material Slots, there's no quick fix beyond deleting meshes from your avatar. You probably don't want to do this.
Texture Memory Usage can be reduced by lowering the resolution of your textures, using fewer textures, using fewer materials (thus needing fewer textures).
Go into blender. use Cats plugin and create a Texture atlas or two. for the materials that is
Question how to I make the material size smaller in blender? A friend said you can compress each material to a certain mb but I can't figure it out. My avatar currently comes in at 400mb in materials, this includes particles and swaps, I made a newer version in blender with less materials required but I still have a high file size at 350mb. I would like to at least get the main model materials down to 200mb materials so it's not a 120mb avatar for people to load in.
No need for blender - click on the texture in the inspector in Unity, and near the bottom set Max Size lower and apply. Or if you want, literally scale the texture down with an image editing program.
i made this model on vroid so idfk what a phys bone is? do they have like a diff apearence on unity orrrr
i know basic optimisation but i still dk what a phys bone is
PhysBones is a set of components that lets you add secondary motion to avatars, permitting you to add motion to things like hair, tails, ears, clothing, and more! Using these well will make your avatar seem more dynamic and real.
I encourage everyone to actually read all of the docs there 🙂
o ok
the most ive learned is how to port stuff lol
I think I know the answer, but want to confirm something
Which is better for performance: one mesh with 2 material slots each containing a 2k texture, or one mesh with 1 material slot and a 4096x2048 texture?
I’d guess 1 mat slot is better than 2. Less draw calls
Right but it’s a technical 4k texture instead of 2 2k textures
so that would increase vram usage using the 4K texture honestly wondering if there's some techniques you could use as far as how your materials are constructed that could help not needing a 4K texture https://twitter.com/FaraVR/status/1668674960712146946?s=20
Wanted to show off another practical example using detail normals!
This one is more pertinent to furry characters where we want the flesh parts of our characters (nose, beans) to have a bumpy texture. Simply using a Detail Normal instead of a 4K normal, we save 20mb!
321
one material slot will render quicker than two, but use more VRAM - interesting compromise
Really depends on what features of shaders you're using. Like I'm 90% sure you can get away w/ a 1024x2048 texture if you're merging two materials which should be the same vram cost as 2 - 1024x1024 textures. Ofc if one of your mats had a normal map and the other one didn't then the new merged texture will need to have that deadspace included. Situations like that where you effectively have unused space in the merged texture is where your Vram increases will likely come from
Like I know people keep saying you need a square texture, but iirc you just need the texture to be on powers of 2. There's probably some caveats where square textures are needed though
1 material. the difference between using 2x 2k and 1 4k is just about 8 mb of vram. unless you have mip map the its about 11 mb difference.
pretty sure it would create more require more draw calls using details. and having a single 4k atlas is not going to hurt anyone. its only 16 mb or 21.8 mb if mip map is on
That's not what I would consider unoptimized as is
One mesh with 2 materials not atlased is like getting an A- on your optimization homework
I am curious at which point it becomes less optimial to atlas things together if for instance only some portions of a texture have roughness/metallic maps and some portions have normals, vs having separate smaller textures for the pieces that have more maps.
a Atlas will never be less optimal then having more materials. its a given fact. that more draw calls = more resources for your gpu/cpu used.
I think the only time where I could maybe see atlasing being an issue would be optimizing for quest where vram is kinda the main metric you want to worry about as maxing out the vram will cause the quest to crash. Though you really shouldn't be using super high res textures on quest in the first place.
You're missing that they're effectively trying to use a 1x2 texture rather than 2 1x1 textures. Usually people will atlas into a 2x2 texture which is where the vram increase comes from. But a 2x1 texture shouldn't be any more expensive vram wise as two 1x1 textures.
That’s what I figured
My model normally has 2 materials that are 2k and I just atlased them by putting them side by side and condensing the UV map essentially
So I just didn’t know if it was treated as heavily as a 4096x4096 texture
With ASTC compression and no mip maps it is well within the excellent standards for vram usage on Quest
that would still be one draw call using detail Maps initiating a new draw call would be swapping materials not an extra texture sample
The unity docs mention one format that needs to be square but I'm not seeing anything else mentioning other formats needing to be square
(see non-power of 2 - To Nearest)
https://docs.unity3d.com/Manual/class-TextureImporter.html
(this one mentions textures do not need to be square)
https://docs.unity3d.com/Manual/ImportingTextures.html#TextureSizes
I suppose it wouldn't hurt to see what the the texture memory stat looks like using 2k x 4k
if the Avatar is using PBR that number would be three times higher as a normal map & a metallic smoothness map would also be needed
You can usually get away w/ lowering the resolution on some of those maps as well. Like I usually have maps/masks at half or a quarter the resolution of the diffuse.
yeah that's pretty much what switch games do when doing PBR textures
no pretty sure it has to since it has to draw it again. but there should be a easy way to see it in unity
there's a difference between draw calls and texture samples
i know
tupper also recently gave a optimization panel and was recommending detail Maps which I doubt Tupper would be recommending if they were somehow requiring another draw call
and why would detail Maps require a draw call but normal Maps don't
from unity them self Normal mapped. This is a bit more expensive than Diffuse: it adds one more texture (normal map), and a couple of shader instructions.
which is a separate concern from draw calls how much texture samples
all through many people do say normap maps does not really affect performance. had to search it up abit..
cause i forgot most lol
no i read that.
i would suppose its about 50% less size. seeing its 4096x2048. which is half of a 4096x4096. in terms of pixels that is.
so instead of 16 it would be 8.
picture showing how many draw calls the diffuse Shader uses versus a fully detailed map standard Shader
and if it's a PBR texture set
increase the savings by about three times
so that's a lot of vram shaved off compared to a half blank 4K texture
The vram for one 2048x4096 texture will be the same as two 2048x2048 textures. So in the usecase of the person you were replying to there is no difference in vram usage so using only one material is probably the better option out of the two they posited
and probably good to see if any detail mapping would help the original poster with increase perceived fidelity
Showing the use of Detail Normals with RGBA Color Masking in @poiyomi . Still on my quest to reduce the use of 4K normal maps in VRC. This entire set is a single 2K normal but still has low frequency cloth detail with the use of DNs. DNs are all 256px
who even uses 4k normal maps lol.
a lot of people
i hope not. but i guess
I've seen screenshots of avatars with 1 gigabyte of texture memory
yea but thats prob also due to insane amount of materials.
also technically if you Atlas an entire up Avatar that could easily end up 4K
Most people are using PBR style shaders with normals, metallic s, roughness, etc. on top of the Diffuse. Not to mention emission maps, masks, AO maps, and more. That all adds up.
i dont actually see that many use PBR.
Yeah. It really depends on the person. I've heard a lot of people use poiyomi but then just use flat shading which doesn't really benefit much from the extra maps
Technically most people are using a pbr shader they just don't take advantage of the pbr part of it. Just using the diffuse
well realistic lighting is also more expensive to use. so.
It pains me every time I see a 4k texture on things that don't need them. Like eyes.
but then again as long as the total vram use for textures is below or at 40 your all good. lol
no need to really reduce below it
That's the issue, a single 4kx4k texture is already like >50mb uncompressed
you dont really do uncompressed.
It's uncompressed in vram
crunch compression does nothing to be VRAM usage
I tend to see lots of PBR myself but I don't tend to hang around with anime people
i dont hang around anime people either lol?
then again I also tend to hang around a lot more robots
oo
and also like PBR stuff is very common for maps on vrchat
so the whole a 4K PBR material is kind of expensive vram wise would definitely be applicable there
well for maps its a different story really
https://twitter.com/FaraVR/status/1668674960712146946?s=20 and here's a example of some PBR on furry stuff
Wanted to show off another practical example using detail normals!
This one is more pertinent to furry characters where we want the flesh parts of our characters (nose, beans) to have a bumpy texture. Simply using a Detail Normal instead of a 4K normal, we save 20mb!
346
Hmm a quick Google does confirm that textures do usually remain compressed in vram if they're in a compressed format. So I'll accept that I'm wrong on that front
its a common techique in games Rune. that compression saves alot of memory but also does hurt quality abit in terms of how the texture looks.
but at 4k textures the compression barely even affects quality
it certainly not using toon shading the nose
which would functionally just be the standard Shader pbr
i really do dislike running out of vram in majority of instances. 8 gigs but only about 6 available in vr.
and considering how popular pioymi is on avatars detail normal and color textures would definitely be handy for reducing vrm
This is why I have the download size limit in vrc set to like 50mb. Cause I know most avis that are over that limit tend to be very vram intensive. Like >250mb texture memory expensive
Though it's mostly just a holdover from when vrc didn't have download prioritization so it would just load all the avis at once ad I'd need to wait for the large avis to load first
Glad the download prioritization system is here now
It’s specifically a Quest avatar so no detail mapping available :P
The diffuse and normal maps are 2048x4096, emission is 1024x2048, and metallic roughness is 256x512
The avatar is about as unoptimized as you can get with a Quest avatar without dipping into Very Poor because of sheer features alone
Main body mesh is about 19k tris and 1 material, then it has a smartphone as an additional prop just under 500 tris. The screen and body of the phone are separate meshes so that I can swap between screens using UV offsets instead of constant material swapping draw calls. Then I have one more mesh that’s a pin on the body.
It ends up being like this:
19.9k/20k tris
4/4 material slots
2/2 skinned meshes
2/2 basic meshes
14/16 contacts
5/8 PhysBone components
VRAM usage is in the excellent category but I don’t remember the exact number
Most of my Quest avatars fall under Medium but this is my personal character that I wanted to be full of features
that's a helpful clarification
still better than most Quest avatars
I’m aware lol
I usually go for a balance of optimization and features
For this one it was heavily skewed towards features
I only have a single very poor quest avatar and the only stat in the red is tri count
I'm trying to make my Avatar compatible for Quest but It's showing red on most parts of the model
I dont Know where to ask this ( Given Im looking for somewhere to post this), but does anyone have a good bullet particle effect? Something Like what the musketeers have?
if I am animating for toggle PB off along with the object it belongs to,
should I?
is it really better for performance to leave the PB there even when I am not using the object?
I’m unsure what you mean by that, I’d it the textures turning red or your stats for the avatar?
I know texture will change if you use custom ones that have slight edits
The material combiner that used to go with CATS doesn't seem to be downloadable anymore. Anyone have it?
Oh is this the new one? Thanks!
not sure if its "new" but it works fine lols
Oh this is literally it! Hahaaa thank youu!
how did curiosity are you doing a color only Atlas or stuff with PBR textures @lapis hamlet
Just color, but out of curiosity is there something for PBR textures?
basically you would be doing texture baking if you were dealing with PBR
A simple tutorial on Texture Atlasing/Texture Baking using Blender 2.8+.
Links in Video:
Pumkins Tools: https://github.com/rurre/PumkinsAvatarTools/releases/tag/0.9.5b
=TIME STAMPS=
0:00 - Intro
0:08 - Disclaimer
0:30 - Combine Your Meshes
0:42 - Removing Leftover UVMaps
1:04 - Removing Unused Materials
1:17 - Creating Baking UVMap
1:40 - ...
I wouldn't recommend this tutorial as it contains information that may cause your avatar to perform and look even worse.
Haven't watched the video but I'll ask what info in there do you believe would cause the avi to perform worse? Each material slot on an avi is an additional draw call. atlasing/baking your textures down into less materials should usually be more performant
Now i wonder, is it even possible to have your avatar fall back to itself... I mean theoretically it should be impossible because your avatar (which is also a fallback and selected as your fallback) should never be able to be automatically performance blocked since it will never cross any limits, but if it did it would just fall back to... itself right? Too bad i'll never see people's faces when they realise my avatar falls back to itself
I didn't see anything either that would make it perform worse, just look worse. I think the key here was "may cause your avatar to perform and look even worse". The may here is key and the combination of perform and look may just be poor wording, OR would have been better. It seemed mainly targeted at the objectively worse look.
yeah, was just curious on what their reasoning was. Optimization and Quality are usually inversely related to each other and you usually have to make compromises in order to optimize. Though w/ some knowhow you can get pretty optimized with little to no noticeable impact to quality.
Well yea, it depends on what you attempt to optimize. Material slots, Texture Memory usage, Skinned Mesh Renderer and Mesh Renderer Count for instance can all be optimized in varying degrees without ever even touching visual fidelity.
Not the channel for that
Ah I see thank you. I actually have a paid texture baker for Blender that takes out a lot of the steps required for this tutorial. I thought maybe there was something CATS adjacent I didn't know about (I'm a bit too reliant on CATS for avatar stuff, ha). I appreciate it though!
yeah there are more Tools in your toolbox than just cats and mainly suggesting you learn about texture
baking because it's useful for PBR setups and even just color only you could potentially get more efficient texture packing then just arranging the texture squares together
how would it perform worse using less material slots which would cut down on draw calls
Because the impact of merged blendshapes on the overall weight of an avatar might be heavier than the impact of draw calls.
ah good point but you could still have a separated face mesh while having everything be one or two textures for the vram savings of a more efficient texture packing
whenever vrc updates its unity version, arent blendshapes gonna be less of a performance issue?
i vaguely recall a conversation about it because someone had asked about splitting up the head/body, but it counted against skinned mesh render
yee, skinning calls and blendshapes will be more performant
Fair point. Along w/ what Rainwolf said people really should also be baking in the blendshapes that they aren't animating in vrc anyways. Like in a perfect world you've got a body+clothes mesh w/ no blendshapes and a head mesh w/ the relevant visemes blendshapes and that's it.
But we all love our customization so it's usually never that simple.
auto fix should make that error go away lol
its unity just crying like a baby because it cant read the file
depends on the baker. it isnt worth paying for a plugin.
I bought this one many moons ago. Was inexpensive and gets updated very regularly for each version. https://blendermarket.com/products/simplebake---simple-pbr-and-other-baking-in-blender-2
ooh, passing that to a friend who seems to keep doing that the hard way
That's actual life changing plugin for blender.
totally is
not an optimizing plugin for blender but another set of tools thats nice is "gret"
https://github.com/greisane/gret
yeah! There's a few things in there which are really cool
I wonder if AI is getting so big how it is today. when will VRChat have a option to auto optimize your PC avatar to quest?
but with a high quality algorithm how avatars have to look like and fuction...
not that it looks like this after the optimisation.
(just an idea)
They are working on impostors, which is not 100% what you mean but it does its job and it's very good imo
And an algorithm for these things would be extremely complicated from what i know.
A neural network won't be very reliable either imo.
wonder if I would have looked better if you baked the clothing onto the body mesh for your quest build and the chains so you're not eating your polygon budget trying to do metal chains in geometry
that is not my avatar. Its a random that told me he gived a F about the hard work and just reduced the poligons until it fits the limitation of quest.
The question here was, to make an AI that will create a (good looking) quest version of your avatar... maybe with something like ChatGPT...
An AI that sees and understands your avatar and priority the parts that needs many polygons and the parts the can be reduzed without having a mess like on the screenshot.
I hear that Windows 12 will get something like ChatGPT build-in, so why not thinking about the VRCSDK?
"AI" is often suggested for things that it doesn't make a lot of sense for, I'm not sure why it'd be needed here.
windows 12? 💀
ye i agree, before ai was used as an excuse to make things sound advanced, neural networks were used mostly to do repetitive tasks that needed recognition or any other action that relied on human tought, i worked with some in the past.
Even if ai has advanced a lot (ironically advanced where it shouldn't have mostly), doing algorithms traditionally is still more effective and clean than training an ai for something such as optimizing an avatar.
An ai couldn't effectively optimize an avatar unless you trained it with a lot of variations of optimization methods.
TL:DR not everything needs to have or be an ai, normal algorithms can be better in case of processing 3D models
Exactly this, yep.
If ChatGPT is capable to reprogram minecraft from scratch or make you some programms just by telling what you need... why would it not be possible to have something what will create or optimize you a VRChat Avatar?
make a avatar from scratch or upload your avatar to the AI and let it improve it.
I saw already ChatGPT features that is connected to the Unity engine and you can spawn objects in different types of behaviors just by saying what you want to the AI.
@buoyant holly @cold yarrow@ivory sluice
you can try it if you want, but a chatbot won't help if there isn't code or algorithms publicly that do this specifically.
I wasn't saying it isn't possible, just that it makes little sense to use AI for this.
Avatar optimizing is a lot more of an abstract problem that you can't steal code from to do
AI is getting bigger and bigger today, very soon, we will get ChatGPT 5 and Microsoft want to build-in AI into their next Windows...
I just wonder how long it will take until it also affects VRChat and it's SDK?
- Intel is making now AI cores in their next CPUs
You need human engineering to create something that a chatbot barely knows.
and also gpt is a clear example of why ai is not reliable for 3d model manipulation, it might be more dynamic and easy to make but the results won't be exactly what you want (i lost interest on gpt helping with unity issues when i saw that most solutions were made up)
I'm starting to think you have only a very superficial idea of what AI is actually all about. It's not magic that you can say "I'll add AI!" and suddenly your problem is solved.
(also the ChatGPT company never responded to me applying for a job there, hmmph.)
"1 meter neko anime cat with red eyes, black dress, white sneakers"
and the will generate it for you. just like with the AI arts and the newer AI 3d models and enviroments.
it's an ambitious goal, do let us know when you've got that working.
Not me... but maybe the VRC Team?... far in the future?........
with a bit hope and fantasy?......................
yeah, you need to train it, and with a lot, and i mean a lot of samples
These trendy AIs can be easilty trained by just sending them images / descriptions or browser results.
Training an ai with something as sequence of inputs and decisions would require very advanced algorithms and an absurd amount of samples
Blender modifiers could use ai but they use algorithms because it does its job correctly and it's efficient
dude, they can't actually get around to working on most of the stuff in canny, you think they'll put time into this?
"far in the future" ?
"far" yeah
I just saw that AI is trending as hell and I can not even hear Radio in my car without a hearing a moderator talking about ChatGPT....
even the boomers did noticed it.
Vrchat is not going to spend time making it's content creators obsolete
my local radio just plays reaggeton and house all day
I have no idea what mine play
optimized Avatar is not something you can quantify into a math problem
I do it myself but 99% of the game doesen't know how to optimize or just don't care about it.
and most times an avatar is optimized by not going bollocks by adding random models and complements, base avatars are rarely unoptimized
Doesn't have anything to do with my statement of the feasibility of it
This is a proof of concept from a Unity developer about integrating ChatGPT into the Unity game engine. Will simple text commands work in Unity, or will it fail hard?
The answer... yes... and yes. ;)
Honestly, I think this video is a glimpse into the very future of artificial intelligence powered/assisted game development tools.
Links
h...
is this a troll
Programming code is a different thing from 3D modeling
It's a bit scarzy how advanced AI is today... and this is still the beginning.
maybe a little
and I'm not even pulling the "I'm a pro software engineer" card here haha
genuinely having a difficult time telling lawl
ngl i'm kinda getting tired of seeing youtubers using gpt or stable diffusion just for the sake of views. Because i'm sure most of them don't have a clue on what they are doing so they let the private message beep boop to do the work
ai is never going to be superior in as niche things as Optimizing Avatars xd
gives me the same feeling as cryptobros thinking that the entire planet was going to go to 1 currency and use NFTs as the currency
it can be, chances are low but never zero
but i'm sure most people are aware that using an ai for that would be a terrible waste of time when they could just code their tought process on an algorithm
which would be very hard to do but relatively easy compared to training an ai
yee i call these people AIbros
because as cryptobros they just jumped on the trend wagon for their quick money while their previous one burned down to crisp
U can quantify everything I to a math problem. Avatar optimization for vrchat is just a series of steps. An ai would easily be able to optimize an avatar much like cats "fix avatar" feature only with more steps
Take decimation for example. that's already done by a simple specialized ai. Just set a series of targets and base steps and run an ai through a generational learning series till you get an ai that optimizes an avi to an acceptable level
Algorithm != AI. Decimating doesn't use AI.
None of blender's native modifiers (nor CATS) use ai.
An ai is basically an algorithm that "programs itself" by learning, just how a human mind would do, that's why it's also called neural network.
An algorithm doesn't vary its programming on the runtime, it follows a series of steps that simulates how a human would process it, but the instructions always stay the same.
I think the only thing that blender has that relies on AI is the denoisers, because it's something that can be trained , it doesn't need exact results and it's faster than an algorithm
Hey I'm looking for avatar creator with reasonable price range to make an avatar quest for me as don't know how to it myself it's old avatar model so should be simple enough
Search for commission in VRCTraders server. #community-servers-old message
Also try not to include the word "simple" or something similar. You really don't know if the task is simple or not.
Alright thanks
its far more advanced then that lol. just reducing polygons is one step out of many. teaching a AI to do it properly would take month's and thats just the polygons. then you need to train it to optimize Textures to get the best result on the lowest possible resolution and highest. and then there is the almost endless amount of combinations of how much something impacts X Persons pc. since there are so many combinations of hardware it would take far to long to train an AI to auto optimize it to each selective hardware combination.
doing it all yourself is much eaiser and simpler. since majority of the performance loss comes from to many materials typically in Vrc atleast. or to high vram costs.
i never said it wouldnt take a while. ofc it would take a bit. but alot of the tools that are used to optimize avatars already use a basic form of speicalized ai. but as with any machine learning once its made its made. the only part of the process that would require human input is the inital set up and then seeing if its good enough. theres already a few tools in unity that you can use to optimize textures down to 1% of what there current file size is. most of the steps to optimize an avatar are a few clicks away for someone optimizing an avatar. would be as simple as the ai exicuting a chain of clicks/commands. tbh with the stuff thats already available you could probably even set up a basic optimization on a macro to get the bulk of the work done. no ai needed.
ive optimized a couple avatars for vrchat and didnt think it was that complicated. took me like 2 hours and i didnt even know what i was doing. all my unity and blender knowladge is self taught via guess and check for 95% of it. then theres that 5% of stuff i got stuck at then googled it. im working on my first from scratch avatar atm
you can calculate the entirety of a human body, its proportions and what makes it up with an equasion after all
they really dont. use AI at all lol. they just use logic based on what people have found
there is several forms of ai most of the time when people talk about ai they are talking about generalized ai
but there is speicalized ai thats used in every day life
there is none through
google search is a very advanced form of speicalized ai
it is
nope
Artificial intelligence (AI) is a branch of computer science concerned with building machines capable of performing tasks that typically require human intelligence.
Google Search is litterly just a Advanced Algorithme with alot of conditional logic.
you use speicalized ai multiple times daily
your recomendations on youtube are the work of speicalized ai
nor is a cpu.
Internet browsers use algorithms, AI is for tasks that would require tought process and would not need exact results
theres different forms of ai
not really.
the ai everyone thinks of when they think ai is generalized ai which doesnt not exist
what no
machine learning is a vastly different thing lol
has nothing todo with any of this either.
from generalized ai yes
im saying you could make a specialized ai to optimize an avatar.
no you couldnt
An ai is an algorithm made to learn and create its own "tought process", an algorithm just follows a set of instructions already created by a human, which is what's used on vrchat tools, blender addons, and internet browsers
specalized ai is an ai that is very good at one specific thing but cant realy do anything else and doesnt actualy understand the problem
cause it would not be an AI
i dont know how you got to think that everything is AI.
but its not.
hostile your thinking of generalized ai
nope
you should do your research stereo knows what im talking about
i am sorry to burst your bubble but Google does not use AI for Google search
nor does youtube.
they both do https://searchengineland.com/how-google-uses-artificial-intelligence-in-google-search-379746
its all purely algorithme based and information on what you search and such
yea no i am not gonna click that.
this is getting chaotic
the algorithems your talking referencing are speicalized ai thats what your not understanding
no they are not lol.
has nothing to do with AI.
a AI would be something like Google assistent etc.
ive provided you with 2 sources proving that you are missinformed on what speicalzied ai is. you have refused to remain reasonable
google assistant is a psudo form of generalzied ai
true generalized ai currently doesnt exist as it would have to be able to change its own code
nah i just dont believe everything on the internet. when it comes from untrustful sources.
want me to get you a scientific paper?
want me to define AI for you?
anyway its the wrong place to talk about that in here.
its for avatar optimizations
That's all well and good everyone. But Imma have to ask you to stop since this has nothing to do with avatar optimization at this point.
we were tlaking about how an speicalized ai could be used to optimize an avatar
Such a thing doesn't exist at the moment so it is functionally irrelevant for Avatar optimization in here
thats the thing i was trying to explain tho rainwolf. alot of the tools used for avatar optimization already use a form of speicalized ai. you would just need to set one up to exacute the steps nessacary and check to see if its good enough.
https://docs.google.com/presentation/d/1yw8NZA37C71Z6YVvmMzZUhteXIj5vOqcFLKr1Ypoyvg/edit tupper recently gave a panel about Avatar optimization recently and hears the slides from it
Does anyone know how to fix the top one?
not the ai optimizing conversation again lmfao. cats and blender are not ai plugins and do not use any ai besides denoising
stop relying on and thinking ai will be able to do every little niche thing ever
it just means your bones arent parented correctly in blender
so open it back up, click on armature and go into edit mode
not able to get screenshots atm but i Think after you go into edit mode, the little green stick figure that looks like its running has the bone relation settings
it was supposed to just be "ai != algorithm" and now i'm left with more questions than answers, wtf is specialized and generalized ai, and what does it have to do with traditional algorithms
what
thought this was common knowledge and have seem to confused some people. speicalized ai = a chess bot/the ai you fight in a bot call of duty lobby. generalized ai would be iron mans voice in his suit.
true generalized ai currently does not exist. the ai in your phone is an advanced form of speicalized ai that can do many things but doesn't actually understand what your asking. it just knows how to reconize what you say and input that into a search engine to give you the result you want.
theres a third form of ai but thats called super ai. think the terminator. we are no where near that one
people make specialized ai all the time. im by no means an expert but there are people out there that could make a specialized ai for optimizing avatars. you could probably commission one much like people commission avatars.
Yeah man that's not artificial intelligence those are just algorithms lol.
uh, anyways, how bad is it to leave the head+body conjoined
Depends on polycount of your body.
yeahh thats what i was thinking?
something like blendshapes checking every poly every frame? finding it difficult to find anything Solid
Bear in mind, this becomes a nonissue as soon as the unity upgrade drops
wonder why there's so many folks in here recently that don't know the difference between neural networks and algorithms
They saw chatGPT and started using buzzwords.
because the common mis conception about AI is that anything that seems smart is AI while its not. we really dont have any AI today at all. the closest we have is automated drive
iirc it's just the cpu skinning cost that'll improve. The other reason I've seen people seperate the head for is file size. I've def decreases the filesize of an avi just by splitting the head and body. I think it has something to do with the bit length of the vertex index having to increase after a certain number of affected verts or something.
who do i gotta pay to optimize my avatar for both pc and quest?
i dont have the brain cell anymore
lol
A lot of time or a good amount of money
i have some money, but thats why im asking here : who is available so i can ask lol
Don't pay people here, all you'll get is scam bots use vrc traders: #community-servers-old message if you plan to spend money.
Other than that advice is free here lol
thanks for the help <333
Ariva Victoria check your DMS.
This is likely to be a scam, of course 🙂
depends on how much work needs to be done. and who you ask. and how bad the current avatar you have is.
lol i dont know how you get to that conclusion are you saying that we are all bots? or are you the bot scam?. plenty of people here would be able to do it and not scam.
people are redirected because bots and scammers will zone in on conversations that could involve money
not everyone is a scammer, obviously, but theres been an influx of them trying to poach people using the #avatar-search-old channel
Money avatar buy
Sorry Bossman but cats' material combiner reeks.
Not everyone is a scammer, obviously, it's just that this server here has no repercussions if you do get scammed. They can just keep doing it.
In contrast, at vrc traders, you must be verified to sell. Either person or can be removed from the server if they are a scammer.
So basically, the community here just sends everyone to traders to deal with money. And we just tell people here don't pay anyone here. It's for their own safety.
is there a comfortable way to remove all bones that are not linked to any mesh
Hm. I don't think there's an easy way. You'd probably only want to remove bones that have no children anyways since bones w/ zero weight but have children might still be useful for organization, constraints, components, etc. when brought into unity.
You can probably do a quick search for end bones (ones w/ the _end suffix) those are appended to the last bone in a chain if the setting is checked on export (it is on by default) and have zero weight. But again they may be useful for physbones if you don't want to add an endpoint distance on the component.
Though I've definitely seen some models that have _end_end_end bones cause they were exported and reimported into blender multiple times as an fbx.
I don't know you will saying that also is here everyone of here bee a scammer
honestly rather than a specialised ai an avatar creator with an asset store that automatically sets up triggers, radial menu, creators can upload outfits and body types for free or sell them in the app and apply outfits to specific body types and sort them by search
all it would take is one very open cutomizable avatar creation system with an asset store to transition the avatar community from whole avatars to selling assets for creating avatars
pls no more ai stuff pls pls psl
How can I check what the custom shader keywords are?
on the material's inspector, click the three dots and toggle debug, you'll see a text field containing all enabled keywords
@ivory sluice Which of these are custom?
_anisotropy_on _transparenthilight_on _vertexoffsetrule_on
_reflectionprobetransparent_on _rimtransparent_on _usereflectionprobe_on _usevertexposition_on __speculartransparent_on
_anisotropy_on _nonereflectionprobeusematcap_on _secondspecular_on _transparenthilight_on _transparent_on _usebacklightrim_on _usereflectionprobe_matcap _usereflectionprobe_on _vertexoffsetrule_on
_rimtransparent_on _usereflectionprobe_on _vertexoffsetrule_on __speculartransparent_on
_anisotropy_on _transparenthilight_on _transparent_on _vertexoffsetrule_on
_anisotropy_on _secondspecular_on _transparenthilight_on _transparent_on _vertexoffsetrule_on
_anisotropy_on _secondspecular_on _transparenthilight_on _vertexoffsetrule_on
and what would pressing Auto Fix do?
Is there a risk of it breaking anything?
If these keywords are local, which i assume all of these are, then you don't need to autofix
And auto fix will open the keyword utility panel, removing the keywords won't error the shader but it will disable all functions that these keywords enabled
im having trouble with my pc freezing when i turn my avi to a android model so i was wondering if i could find someone out there that could upload for free
Kinda
Removing a keyword from this list is equal to manually turning it off on the material.
But i'm sure all keywords there are local, so you don't need to do anything about it.
Anyone know a good asset store product that'll combine meshes, materials, atlas texs etc? Tons on there but not sure if anyone's tried them.
https://github.com/d4rkc0d3r/d4rkAvatarOptimizer
Can do the first 2 things.
But not aware of anything that can do good atlasing in Unity and not aware of one that does skinned meshes.
However there is polytool that does do all those things, but I am not sure how flexible it is, assuming you need something that doesn't just automatically optimize https://markcreator.gumroad.com/l/Polytool
Thank you for the quick reply and info, I will check all these out.
Is there any way to set the default compressor format for Android to be ASTC 12x12? It's set to 4x4 for every single texture I have, and I can't change multiple at once. Going through every single texture and changing it to 12x12 then waiting for it to finish uploading is tedious as hell
If you have all your textures in one folder you can select them all and change the format, it'll do all of them at once
The format is the only setting it doesn't let me change when selecting multiple textures
Maybe it's the file formats
heyo, lmk if you still need an avatar optimized. Been working as a 3D artist for 6 or so years. Not sure how to convince peeps I'm not a scam? I'm inbetween studio gigs. I'm new to freelancing, so idk how to market myself lol. Here's my portfolio if you wanna see my work, I specialize in optimization, low polycount/texture amount. If you want, DM me the model details and I can give a better quote
Man if that's not the most sus thing lol.
50/50 a real person.
yeah always feels sus but i think theyre real
had a more human conversation with them in the rig channel at least
about using mixamo and how they changed the naming conventions on the rig you export from it 
They also have an avatar world #avatar-showcase message
It's art station. That's like one of the go to place for portfolios
yea
Is there Any Documentation in what way Hidden Objects/Assets can Draw Performance? I know that Having an Avatar with a lot of Stuff can draw Performance, but i have run tests and found out, that even without the Extra Animatorlayers, the same avatar with and without the Assets runs a good 2-7 FPS worse.
I got a Ryzen 9 5900x, RTX 3090, 32 GB 3200Mhz Ram, Rendering at 150% on Quest2 Wireless
In Desktop
~128 FPS Avatar without the Extra Assets in my Homeworld no Mirrors
~107 FPS All Assets and Animators (22 Layers in Total no Anystates used unless the Standard Additive has Anystates)
~118 FPS with the Animators Removed and just Standard Animators
I know that a Full VRam requires the system to use the RAM and slows down the Rendering, but i doubt that this is the Case here.
Collisions of Contacts and Rendering of Cameras on the Assets should be disabled as far as i know.
Are Audiosources in any ways processed even when disabled?
its probably a CPU thing, but which components would draw CPU even when disabled and why?
The Avatar has Quite Many Audiosources, Contact Recievers, Contact Senders, Static Meshes, Skinned Meshes, Particlesystems, Lights, 2 Cameras, Materialslots/Materials but all that stuff is not active on standard.
idk if this'll awnser your question, but here's some benchmarks:
https://notes.sleightly.dev/benchmarks/
might or might not contain what issues you're experiencing
def worth a read either way though
Yeah, check out what @pale pollen linked.
Also, you should consider adding Thryrallo's VRC Avatar Performance Tools to your projects. It has an Avatar Evaluator that can tell you detailed information about how performance is affected based on what's on the model. https://github.com/Thryrallo/VRC-Avatar-Performance-Tools
Disabled objects do not occur any performance loss. they only take up memory and no objects that are disabled can take any cpu resource. its not possible.
i already checked that out, but my testav doesnt have custom animators
and as far as i know, vrram usage itself when not rendered shouldnt affect fps
it still needs load in the memory. even if its disabled.
no. it doesnt matter how much vram you use. however as soon sa you go beyond your capacity then it will struggle since it moves overflow to regular memory and thats slow AF
yeah but this is not happening with my AV
you have alot of lights, bones
yeah but most stuff hidden
does not matter. all the materials still need to load into memory.
i tested the Same AV with and without the Objects and even when disabled the one without performs better
how is that draining fps
shouldnt this only matter when the capacity is full?
eh it wouldn't affect it. but that many materials will cause anyone to complain when they load you in.
but now we have a problem here, by your explaination i should gain same FPS on both my versions but i dont
iirc theres also a slight performance cost to moving object or bones around. It's not much but can def start to be felt when there's multiple 500+ bone avis out.
if its parented/child bones for sure lol
Googling some rather sus things at 9 am like 'can disabled children affect performance unity'
How many parent constraints? iirc that's not counted in the perf ranking yet. They are tracking them in the avi details in-game though
51/52 no objects
51/52 no animator
49/50 animator+ objects
udonbenchmarks
79/80 no objects
78/79 no animator
76/78 animator+ objects
tinyworld
49/62 no objects
60/62 no animator
59/60 animator+ objects
greatpug Populated
hmm, maybe my previous numbers where changed by something
Looking through this entire thing
Everything hidden or not has data to even exist in the first place
All that data needs to be loaded whether or not its shown
Hidden objects have materiels and meshes that needs to be put in memory
Hidden bones need to still send position and rotation data even though you cant see them
The more components like lighta or constraints give the object even more data
Polygons on objects can also cause lag but hidden objects wont directly/visibly cause fps drops since its in memory vs needing to be rendered right now
Dynamic lights take up more resources than baked lights
hmm. i wonder if there is a way to actually get decals to look not so blurry at 2k resolution or 1k even.
turn off the filter
When setting up toggles, should I have two seperate animations to turn something off and on? or is there a way to do it in one animation? does it affect performance?
performance not really
and yes
cause otherwise it would toggle on and not toggle off
since nothing is telling it to
Technically you can do it in one animation using motion time but honestly it shouldn't matter too much.
The only thing to consider is the amount of Layers in your Animator Controllers. The more you have, the higher of a performance impact. Use Thryrallo’s VRC Avatar Performance Tools to evaluate how that looks like in technical detail: https://github.com/Thryrallo/VRC-Avatar-Performance-Tools
Add to VCC: https://vpm.thry.dev
Honestly, if you have enough toggles as layers to have that much of a performance impact you're probably pretty far into the very poor rating anyways
Hmm, unless it's all blendshapes I guess.
Hey everyone. I've built an avatar that only uses 2 materials, but has a lot of toggles for weapons and stuff. Each weapon is a separate mesh. The builder screen shows I'm using 18 material slots. Of course its counting the amount of times a material is used as opposed to how many materials in total are used.
Is there any way to optimise my workflow so I can keep my weapons, but not have to use a separate material slot for each one?
I suspect the only option may be to use blendshapes to hide the weapons inside the body, but that sounds like a pretty unoptimised way of doing things.
could someone dm me and assist me with a question regarding blender and unity. ive tried everything I can think of
If they're all using the same material you could merge them all into one skinned mesh with a bone for each weapon. To disable them with this you can scale them to 0 and bring them back to 1 scale when needed. It's not entirely optimal since you're technically still rendering them but it'll save on mats.
Could also look into uv tile discard for hiding merged weapons if the weapons are already pc only.
Honestly though. 18 materials isn't too bad especially if you are using all the weapons at once cause you're only incurring draw calls on meshes that are active.
So with the current setup I have, I in total have 3 separate skinned mesh renderers. When I use Thry's Avatar Evaluator it says my blend shapes are medium ranking. my "Body" renderer incudes my avatars base which has blendshapes, and his face which also obviously has blendshapes, would it be better to separate the base and the face? If I understand correctly having more skinned mesh renderers would probably be less performant than the blendshapes. But I just want to check with people in here to see if my thought process is correct
well. depends you may want to see if u can merge it all. and have one skinned mesh only. however if you really want to seperate face from the body then understand your going to have a fairly ugly cut off line from where ever you seperate the face from the body. the performance benefit from it is not really worth it and with unity 2022.xxx whatever version we are getting its less of a problem
and potential other issues with shaders and colors not being the same as well
Merging it all into one would still have performance effect when it comes to blendshape but less so than seperate skin meshed renderers? I plan to keep the clothes and body separate because I do certain things that I think might break or I am not super confident about.
skinned mesh renderes are fairly heavy. and honestly the face seperation is not going to massively improve your frames. infact i am pretty sure its not that big even.
i would just keep it merged. but you can obviously test it if you want
have two versions one with one skinned mesh and one with two
the face and body seperate
and check the frametime of each being the only object in the unity editor
and again once we get unity 2022 LTS blendshapes no longer really demands that much
Oh that's smart! I'll try doing that! It wouldn't probably hurt anyways!
I'll make another version of him with just one skinned mesh renderer, I could probably just use separate materials like I am already currently doing but reduce the the draw calls because there one less mesh renderer
worth a shot!
theres a whole ton of reasons to avoid skinned meshes as much as possible. like each skinned mesh also has to calculate light being bounced off etc and such
Yeah that makes sense, I can still use separate materials which I need because different parts of the body require certain things from a shader that I can't do all in one material, but If I understand, it's better to have 7 materials on one mesh renderer than 7 materials spread across two mesh renderers
yes its alot better to have it all on one
however, merging these two together are going above the shape key polygon count which the wiki says not to do
yeeea i know but its not that crazy
atleast not when we change unity version
so its not realllly worth doing ngl
I'm going to make a separate one, and test the frametime of both in unity and thry's editor to see if it makes much more difference in my specific case
how many polygons are we taking about in total?
Should I combine all my skinned mesh renderer it'll be total of 63,314 polygons
Currently have my body, clothes, and hair as mesh renderers
not to many
hair is separate because I didn't want to go over the 32k polygon limit the wiki said, but I can always still merge it
mhm
If you transfer normals from around the cut area on the uncut version onto the cut version there shouldn't be any noticable seam. I've done it on one or two avatars.
kinda still is. which is a huge disadvantage
Weird. I really haven't noticed it at all on the avis I've done it on.
even in triple A games it still happens 😄
can confirm, i have data transferred normals from a solid mesh to a separated mesh with just the seams selected via vertex group and i do not have the split edges problem
unless a shape key affects that edge, than i just disable shape key normals ez
Hi Is it better to have a single atlas for the entire avatar or divide the textures between each materials, how does an atlas improve optimization, does it make the avatar size lower?
less draw calls with an atlas, only combined resolution matters for performance (4x 2048 = 1x 4096)
Atlas is far superior. as it results in being able to for instance if you have a 4k atlas it allows rendering of 4 2k textures as one texture only. which greatly saves on material cost. obviously you can do 16x 1k textures inside a single 4k atlas and so on. its basicly doubling every time you step down
however having a single 4k or two 4k textures does not really hurt anyone.
also for people who don't know, having an alpha channel literally doubles the file size of any given texture, also if you pack 3 greyscale textures into a single RGB texture, they collectively take up 1/3 the size of what you had, using this, you can get away with several 4k textures instead of 1 or 2.
Hi! Do I have to delete some textures to convert avatars to quest?
maybe for optimization but you can have multiple materials on quest avis
It fails to upload because of size limit, it is 24mb and I should reduce it to 10mb
Or just reduce the resolution of the textures, you can set it separately for Windows and Android.
I'm currently re-compressing a lot of textures for mobile, as I'm finding out the texture formats available for crunching textures expand to fill incredible amounts of VRAM on a quest. think like, 1 meg becomes up to 40 meg. fun! so now I'm recompressing to things like ASTC and PVRTC to be kind to the very limited VRAM of an oculus quest 2
Now that I'm done on the mobile side, I find that the PC side of VRAM texture compression (not crunching. crunching is only on disk) is actually pretty hamstrung... I guess the only option I have is DXT1 as even BC7 is... bigger than expected for a texture that has large areas of completely flat color.
textures that are just big islands of flat color probably could have the resolution decrease without much detail loss
its the transitions that make that not an option...
Question about UV tile dissolves and geometric dissolves. From what I understand, meshes aren’t rendered when using UV tile discards, but what about those 2? Are the meshes still being rendered when they are dissolved?
If I’m not mistaken, they will still be rendered since the Mesh Renderer is still active. Even if you use Tile Discard instead of Dissolve, the original shape of the Mesh still exists even though it’s being hidden by the Shader.
Thank you!
One huge benefit of using UV discard is that the material doesn't need to be cutout or transparent and can be opaque, meaning you won't dissappear behind transparent surfaces.
Can confirm. This is true.
iirc from a friend who uses uv discard he's had some issues where some shaders in worlds will still be affected by the discarded mesh. Like some water shaders will show an outline of the discarded mesh still. That's pretty much the only issue I've noticed aside from the normal issues you'd get if people don't show your shader.
That is caused by a projector, which applies a shader onto any mesh, no way to stop that except turn off the mesh.
Lots of Polygons make the MB size of an avatar go up right
Just polygons on their own not that much, but if there is tens or hundreds of shapekeys affecting a lot of vertices, then their memory and download size can shoot up.
can somone help me with quest optimizations
Yes. The exact amount varies but its at least around 44 bytes per vertex. However that per vertex number goes up with the number of shapekeys since each shape key I believe adds 12 more bytes (3 4 byte floats) even if that particular vertex doesn't move. Also every UV map adds to that number as well with UV maps taking around 8 bytes per vertex.
Triangles/poly themselves don't add that much size, but you end up with more vertices as the number of triangles goes up.
Thanks!
Yo, Im trying to make an avatar again after not playing vrc for 2 years, i frgt if I should do this or not, but should I delete all the faces of the diff objects that arent visible, like the back side of the eyes?
you wont see it so
Hi, guys
Is there any Unity addon to batch copy components from one avatar to another?
Let's say I have two similar avatars with a similar bone structure (the main bones are the same and have the same names) and I'd like to copy every component from every available bone from one to another, but without copypasting it manually
Thank you ❤️
Is it expected for a particle system to take a material slot if the render mode is set to "none"?
I guess according to https://creators.vrchat.com/avatars/avatar-performance-ranking-system/#avatar-performance-ranking-stats particle systems with trails use two slots for VRC's ranking. Do the particles still actually have 2 slots if render is set to none or is that just how vrc is calculating it?
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.
Ah, I see there's a topic for this on Canny already https://feedback.vrchat.com/bug-reports/p/particle-systems-with-no-materialdisabled-renderer-module-still-count-towards-ma
I can't remember can you enable/disable components of the particle system via animation? I could see that being where an edge case might appear in that case but then you could just disable the trail renderer by default and enable it at runtime to 'save' a slot since iirc you can also swap materials into a slot via animation
Though yeah, it'd be nice if particle systems w/ no renderer could be made to not count against your material slot count. It'd save a couple of slots on some of my avis that use the persistent particle workaround.
How do you optimize an avatar for quest 2?
Change shaders to android optimized
Can you help me walk through it?
Yeah
Did this video help you? Consider sending me a tip on Ko-fi! https://ko-fi.com/sippbox
Making optimized Quest models is hard. But it doesn't have to be! With recently updated polygon limits and fallback avatars, it's a better time than ever to create something that represents you to Quest users! Hopefully this tutorial will help you out!
0:00 ...
edge loop dissolving, change shaders to mobile
theres a tool called easyquestswitch you can use too
I have easyquestswitch, but idk how to use it
I sent ya a friend request
I'm trying to cut down on polygon count for my avatar, as it's the only thing still in the red. Would deleting polygons under clothing be a good way of doing it or would it screw up other parts of the avatar?
if you dont toggle those clothes off then yeah its a good way to reduce polycount
dissolve edge loops too
Great, most I'm planning with the clothing is pallete swaps so I should be good
yea, dissolving edge loops is a great way to go about it
and stuff that you straight up wont see ofc
Yeah, the before is just under 100k so hoping I can cut enough
Sadly have to put it on pause cause work.
One last hail mary I'll do if I'm still not down past 70k on an avi is start decimating some of the bits that are mostly weighted to one bone or don't have many blend shapes affecting them. Like nails, pawpad on furry avis, etc. Mostly cause that usually only removes a couple thousand at best.
I'll keep that in mind
im trying to add a couple of physbones to my model. But when i add the bones to the armature and weight paint them to the mesh in blender, it for some reason makes the file size jump from 18mb to 150mb? I'm just exporting the mesh and armature too, so i dont know what other export settings could be causing this
i was in blender and used the quick desamtion on the cats plugin and it made my models mouth open and i was wondering how to change it back
When you say "the file" do you mean the FBX you exported?
What were your export settings in blender? does the blender project have animations in it?
it has a couple of limited animations, im exporting just the mesh and armature, but i dont know what other settings would effect it. im pretty new to blender and unity
Trying exporting with "Bake Animations" unchecked. I think even the most recent version of blender that actually means don't export animations (blender can't export unbaked animations to FBX format)
okay yeah, that cut the .fbx all the way down to 2.5mb, thank you
i think i optimized too hard and now my clothes clip into the person
how do i add edgeloops
or somthing similar
are the clothes going to be toggled off at any point?
no
then just remove the parts of the body mesh that are underneth the clothes. it will reduce the size and poly count and there won't be a cliping problem anymore
like delete those edges? or
faces, vertices, just any part of the body mesh that will never be seen. always keep a backup before starting. when deleting mesh along a seam it can mess up UV maps sometimes
oh alr
everytime i try to save the atlas it says error
this is whats poppin up
what do i do
Hey
Can one of you guys hop in a call?
I’m having problems with performance along with too many polygons and materials slots
i just use this one yt tut
the thing is im stuck
Pure guess, but 2.83 is old, newest, non-developer version of CATS support 2.93 (Developer version supports up to 3.5) https://github.com/absolute-quantum/cats-blender-plugin/releases/tag/0.19.0, possible that material combiner would only work on 2.93+.
Quite easy to have multiple blender installation at once so I suggest you give that a try.
so my blender ver is just really old
Hm, apparently it is a LTS version (only 2 years old), but I don't see why anyone would stick to such an old Blender version nowdays.
maybe they have a really old computer and cant afford a new one
hmmmm
(thats me)
Wasn't aware that newer Blender was that much more demanding.
Hopefully at least 2.93 works.
Either that or you go back through the releases of material combiner and finds one for 2.83.
ill do the latter
when was 2.83
nvm
i found it
was givin this when i tried to enable
i remember it working one time
i think thats what happend
Question:
Even if the total polygons of the avatar in Avatar Details page is high,
If most of the polygons are in the items(like outfit, props etc) that are toggled OFF at the time,
will it still affect performance?
they would still be contributing to the avatars total vram cost and Avatar performance ranking but yeah they wouldn't be but yeah it wouldn't be incurring the cost from actually drawing the triangles if they're disabled?
i see, so even if i hve more polygons, as long as i keep textures small and toggle off stuff it wouldnt be so bad
What does "PhysBone Transform" mean?
how many bones are being moved by the physics bone system
and I might suggest uploading a second variant of the Avatar that doesn't have all that extra props for when you're in club worlds that enforce no wearing very poor avatars
ah no problem, I have a fallback that looks the same as my very poor variant
Cool
does disabling a physbone component make the avatar more "optimized" when the physboned object is disabled, its a question that ive had for a bit, but i was just wondering if not disabling the component changes the performance of an avatar
Disabling it doesn't change the perf rating. But it certainly stops it from working which make it no longer taking fps.
Yeah the sdk will just sum up all comments in use for perf rank.
But, yeah disabling an object effectively nullifies the compute cost. Data will still be in vram though.
Ok that's perfect exactly what I want, thank you
Is there a easy way to optimize an avatar after making it? Or if it's a bought one but isn't optimized
you need access to the Avatar files to do the modifications to it to make it more optimized in blender or Unity
so if you bought like an avatar off of Booth you could do modifications to it to make it more optimized but not a a publicly uploaded Avatar that you didn't upload
What would I need in order to optimize it?
probably blender to do modifications to the model

thank you