#avatar-optimization
1 messages · Page 34 of 1
Anyone else using cloth automatically get poor or am I like a few polys to high. Just wondering. I’m at 400
I have 400 and everything else is green but the overall is poor, I didn’t know if it was the poly count or if it was just using cloth all together.
400 polys**
Yeah, the rating system still calls it poor
what system
never used it so idk
On performance.
cloth being poor is preemptive i guess
because there are people who do use cloth really badly
Well I wasn’t sure if anyone else has cloth under 400 that doesn’t sink them to poor.
it's never, ever, ever about "this doesn't lag me"
Duplicate the model 25 times, add IK to some of them, up and down some of the materials, verts, etc
Public avatars with 40+ materials >_<
Report the world or avatar
I had a good laugh earlier cause i only have one model using cloth, its at 440 tris and set me to poor. otherwise it would be medium due to a few dynamics in the hair.
ye the hair on that model is cloth
and i was gonna make the jacket cloth but went with dynamic bones instead
i'm only using cloth on the shirt
Yeah me too
but cloth hair...that sound weird
hair acts like cloth too
I decimated my skirt and it only uses 2 colliders
And I’ve used cloth on hair before, there was a performance decrease I removed it immediately
Cloth hair doesn't work very well at all, it looks horrible
probably bc of the polygon count, self collision, or etc
not on mine bc ive done it correctly
it acts a lot like actual hair it just isnt separable
not if uve done it correctly
I’ve only seen long hair tho
mine doesnt do that
With to many polys
only when u move down too quickly
like crouching really fast
otherwise it moves like actual hair
Yours is short too so it’s not going to get caught on or in colliders
and its brushable and u can pick it up
also has cloth in unity been updated because like
i found out that when i duplicate the hair it can actually collide with its copy
and i didnt tell it to
also the only colliders are in the head and on the arms
bc of the hat clamping onto most of the hair
it stays where its supposed to be
and i really like it
I’m sad it makes my avatars poor
Yeah but it counts tri amount.
I'm personally fine with being poor if its because 30-50 transfroms for dynamics. I've always aimed to be well under that if i can and make it still move the way i want it to.
Cloth i never really got into though did one test with it and that was it.
They suggested us 200 polys on optimization in their Pamphlet it seems using it at all is no.
my hair not have a lot of verts
i think too many polygons tho
but again somehow i gues it just doesnt lag people
ive asked them ive like switched to this avatar
As the person said duplicate that model 20 times there might be a noticeable difference.
yea but nobodys gonna have 20 of the model in one instance
But all my avatars are private. So there won’t be 20 of us.
When 20 of similar models if people don't optimize.
ive gone to the pug with this model
So I’m not optimizing my model for using clot?
already laggy from everyone elses avatar
but nobody noticed a performance drop from mine in specific
so im gonna assume its done well enough

I’m just asking not angry btw uvu

I just realize it could’ve seemed angry
Same with any of my models too, Wulf.
Never had anyone complan about any loss of performance even near mirrors or anything.
use cloth just dont use it too much or badly
I think 400 poly cloth, with low amount of colliders is not going to cost to much on performance.
yea they do
"hey does this avatar make u lag"
"*switches*"
"na dude ur fine"
it used to be a guessing game who to block
not really
Considering the rest of my model is green to go, it’s really sad that one thing can ding me and push me to red.
Just sure that if people are getting their frames taken and they see you with your avatar rank, it'll likely block your avatar. I been doing that even before this update, although I would go through a decent amount of people before I find them 
some people take a while to load in and then everyone notices the millions of shaders
idk how much more until trusted but am known
Even sadder when it appears just using that component flags you.
does it affect anything in vrchat
Hey out of interest, is the avatar performance calculated at runtime or is it just calculated at upload
I'm going through and trying get my avatars to at least "Good"
or is it just "heres how performance gonna be for this avatar"
@spark bison runtime.
Nothing I have is considered good because I use even the smallest bit of cloth.
to this day, when it comes to avatars with shaders though. The ones that takeover half the world with screenshakes or epilepsy inducing crap always get a block from me.
Aaaa I’m sad.
shame, then you can never really know how bad it is when using animation overrides
If an avatar renders over the walls I block it
or does runtime mean when you first spawn in?
disable an active item with cloth in it
and it can crash
the entire servers people
but again this is old unity
not a problem anymore
animations arent a problem
whats on/in the avatar is
im so happy about this update though, i hope people optimize now.
No way id ever get under this dynamic bone limit, but least everything else in green
does it disallow upload of poor avatars
whats the limt
most ive used is like 5-6
32 for Medium, but most of mine usually 50~60
Honestly i wouldn’t look at people with a “poor” avatar and say you’re unoptmized
so instant poor
Most of mine will be med/poor just from having 2 colliders and about 30-48 dynamic bones.
I'm also fine with that.
lol 50 - 60 dynamic bone scripts?
No
or bones
nah bone transforms
Bones
i always lower the amount of scripts in an avatar to 1-2 max
This is actually frustrating a bit
for dynamics
i also wish that cloth was easier to linit
I got 1 very poor avatar rest are good~poor
They could put a thing on cloth though
It’s counting cloth polys
so it doesnt move off
ive had avatars
with a LOT of cloth
one of them had a cape, a robe, hair, sleeves....
lmao
AYE!~

Well there’s not much I can do about cloth it appears. I suppose I’ll have to just have a poor avatar. I know that they are still fine tuning the ranking just a bit sad that you get grouped with people who don’t optimize for 1 component.
The avatar score needs to be a weighted average based on performance impact from the line item
I doubt being over the poor rating of some of these items has the same performance hit as being over the poor rating of many others
but if the goal is to not exceed any maximum It does accomplish that
I just fear people will give up optimizing once they realize they have low capabilities of reaching good for the use of dyn bones or something.
it's probably going to be forced in the future
I can see a time when very poor avatars are blocked by default
yea when its actually accurate
couldnt it also be judged by the users performance before and after the avatar is introduced
Having a "low performance requirement" toggle would be the best way to deal with this
as opposed to some guess
i think the update like this would be good as for stats nothing like oh A poor performing avatar let me just auto block all of em
if its just for stats ok but otherwise it will be bad
but actually that should (at the moment) never be added
if a person blocks you because of either your user rank or your avatar rank, is anything really lost 😛
its just helpful for finding out who is the laggy person in a game that wasnt lagging
bc the systems not that perfdct
almost never happens tho
I think it should’ve been added without being public to everyone, until it’s been fine tuned and tested.
Like, everyone can see their own stats.
It did need more testing, but when is that not the case
both their company and mine released right before Christmas which is a no no too 
has anyone found a fix for the snail marker bounds issue yet?
Check if you have compressed on in Blender atlasing, it make it horrible ,no idea if cats one does that, but manual using the Materialcombiner have it as default.
You can also click on the texture in unity and set compression there, default normal
So for those with cloth, I just learned if you use cloth but your vertices count is 200 and under, you should stay in the yellow and not hit poor. :>
They have recommended staying 200 polys for cloth since it got big. So I don't think it's completely biased.
It's something they've stood by.
ik but i think mines 300 or so
this update is absurd
dont remember
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 purely informational at this...
all the specific numbers are here
why is it called "Overall Performance" if it just displays your lowest ranked category lol?
how is having 1 poor the same as 10 poor

because single things can tank performance
1 bone over limit, boom you are poor
light sources yeah
a bunch of dynamic bones but everything else good will perform poorly
sure, but there's no weight
the system doesn't consider dynamic bones worse
it thinks they're the same as everything else
¯_(ツ)_/¯
i dunno, making it a hard line makes it less of a guessing game
weighted categories would be nice but would also take more research and time to implement
each category needs to have a weight, then the overall score be an average of the categories
for future avatars i can plan around this
agreed
yeah it would, and they should do it like that instead of making it live as it is
i can plan around the amount of dynamic bones and stuff needed
the skirt optimization video I'm looking at makes dynamic bones mostly not needed, it's like a 4 bone skirt lol
yea but it can be done by testing fairly easily
like they could make an avatar thats "poor" on each category
and see which avatar is less performant
and u can compare the frames and figure out some percentages
The truly sad thing is even after I fix everything I'm still gonna be very poor unless I remove snail marker 
lol
Anyone here that use marker on your av had problems with the bound size of your av?
yes
did u find a way to fix it?
performing well is so important to making things pleasant for everyone
not yet
same :/
no way to fix it, after using your marker in a world your bounding box will be astronomical
it affects bound size even if disabled on load
actually for me it didnt
is it possible to instead make a transparent trail and the trail is bounded inside your model instead and also have the game object disabled
it should work, but everytime we stop our drawing gesture it will erase everything
yes
exactly
and then you change the trail color when you need to use it
so your bounds stay small and the marker cant be seen if not in use
I've never seen this done but ill try it because it seems like it would work if you can change the trail color through an anim
bound size shouldnt really be counted
you can't have different colored segments in the trail renderer
so animating the color won't fix it, the mechanism that shoots the emitter into space when you lift your pen is the cause, and the marker material uses it to hide those parts
when you animate the marker to shoot away inside the avatar or say, 2m down, there's problems with the interpolation of the movements causing visible jagged endpoint triangles
hmmmm
u can use two gestures to draw, one to activate the marker, 2nd to from -10000 to 0 location, my avatars don't have bounds issue due to marker, as I use two gestures
I'll give that a shot real quick
@safe swift What are your avatars rated at now? I've seen some of them in game before. :p
poor/very poor, because it takes worst as display
any avatar that uses rigid bodies for arms has really been hit hard.
i don't care about system for now, it's complete bs
they fit the world avatar rules, so idc further for now
It's wayyyyyy too strict on dynamic bones imo.
The recommended should be less than 20 bones or something.
Not 0
dynamic bones are awful for perfomance
The rest of the suggested limits are fine.
What do you have a Rigidbody on? 👀
my avatars
Nice, I got that part.
Uh, alright.
That's a lot of rigidbodies.
depends on what you use it for
I suppose rigidbodies are really taxing, they're essentially just gravity/physics components.
Dynamic bones are simpler.
well, dynamic bones are #1 issue
Imo, yeah.
The suggested limit really does need to be like 15-20 bones.
At 0, it means anyone using dynamic bones will be medium or below.
And people will use it.
Optimization matters a lot more for public and friends+
But 15-20 dynamic bones per avatar in a public world really isn't much
maybe a future feature can hide poorly performing avatars in publics for instance
^
i dont use gravity for any of my rigid bodies. just fixed joints. which is why im disappointed. im mad that an avatar that has 1 mat 1 mesh etc and is great in every category is marked as badly optimized because it uses some fixed joints
@real orchid they don't check what the components do, so obviously that won't be an issue for performance, but 8 fully enabled rigidbodies would be.
Same thing with dynamic bones really.
Someone could have 5 bones, but have update rate on like 150
can rigid bodies be disabled when not in use? I'd imagine not...
only if the bone is disabled i believe
but yeah im mad they dont take if gravity is enabled on rigid bodies into account.
If they released something, like some kind of research into the performance cost of stuff like dynamic bones, to show how bad the performance hit really is...
That's help them tbh.
its basically an avatar that has arms that copy the other arms. im not sure that affects performance greatly
they're a small team so it might be too much for them
But... Right now people go "well I don't lag in this world with me and 2 people, so it's not an issue"
i think they need to collab with the dynamic bones authors to make an optimized version of dynamic bones for the game
I'm not sure it can be optimised.
It's just a lot of time consuming calculations to do each frame.
The page on the VRChat docs for avatar optimisation is really useful on how to optimise dynamic bones though.
well i believe that the unity cloth script runs on gpu so there might be a better way. since cpu is the bottleneck in this game
https://i.gyazo.com/7cbb49e0f41117bf349d619e07c2889f.png 'trusted user'
^^^ most trusted users
there are world based avatars that are way over
what the fuck how do you have 329 mats
@real orchid that was my thought tbh, I assumed cloth would perform better.
Thats not mine, just some random guy in Pug
I just saw that
130 skinned meshes?
They probably forgot to combine and then just didn't care.
@last light pretty sure if they found a way to mod cloth for bones it would be better. pretty sure its the same concept
Cloth can perform well if it's set up well.
@real orchid in before Nvidia hair works. 😅
Or actually tested before thrown into live.
To think I'm barely above someone with 563k polys.
just sad about my optimized crab avatar. fucking rigid bodies
and 130 skinne dmeshes
Yup.
I'm actually done spending so much time on this one avatar to try to get it to their standards.
I've spent so much time optimizing one stupid avatar to just get poor.
ive given up on some of my avatars. i just need to make new ones
My avatars were poor to begin with.
woof
I already atlas'd and all that good stuff. AAAAAAA
man i wonder how it must rank my
optimize! or else you make things worse for everyone! its being polite!
What's it getting poor based on? :p @heady smelt
3 skinned mesh 4 material and 70 something tousand poly avatar with cloth
even tho it not lag in game
prob very poor
I'm getting frustrated because I've spent so much time optimizing to get poor.
they should remove rankings until theyre more reasonably accurate
with "lag detection"
You can try removing some verticies from the cloth. @heady smelt
I did.

Oh. ☹
Yup.
Rip
i really think many people will be blocked based off of performance
They suggest 200 in their performance guide but 200 = poor.
It is really hard to get above poor though.
So I'm actually just done.
I think they'll tweak it to be fair.
when vrchat says ur rinky dink avatar is very poor
Yeah im settling in on Poor for now,even tho its noowhere near as bad as others
Anyway I'm dipping from these channels and ignoring performance stats cause I'm getting irritable.
Wish you all the best of luck in optimizing your avatars.
To be fair, anything about very poor is good. 😅
the exact moment 0 dynamic bones was the recommended setting I lost interest
I only used dynamic bones to begin with because of tupper's tutorial.
I haven't spent a year making models and learning to get told that lmao
OK now I'm gone.
Ideally you should have none, but people really like it, and will use it, so they need to raise that recommendation.
I'm not being good either
being good or medium is ideal
medium is the most I'll shoot for as 'ideal'
medium would be fine if they upped the dyn bone limit to 64 or something more reasonable
none of them is just excellent performance wise
but thats obvious
i dont think they expect most to be excellent
when public models are out there as very poor and poor? no
theres going to be a transition period for sure
those public models are from before they started requiring more optimization when submitting worlds, no?
no
Pretty much all except polygons are off on the sdk/avatar perfomance details for me
bounding boxes can sometimes change in play mode, but I've only really experienced having to fix that when using cloth
their public world standards are a bit different, theyre still tweaking things.
public avatars should at least be medium
i agree
@radiant shadow I just saw that screencap. My god...I cant even.
Its like a game can i find someone worse then that avatar now :p
lmao.
at least that person kept the dynamic bone stuff low
i saw someone in the pug with 6k collision check count
oof
i really look forward to blocking very poor by default
I'm losing my mind cause i always thought i was barely optimized. Then you guys show me these stats and i'm like. WHAT.
6k? on the avatar I went all out on I only got to 1500 christ
Seen so many trusted with 100k+ poly and 40+ materials running around
I never took that ranking system seriously anyways
I mean just because you uploaded a lot of stuff (assuming) doesn't necessarily mean they're good at making it
Yeah i been crashed by all ranks
I'm only a known and none of my stuff is horrendusly bad.
Most my models in the poor rating are mainly for 50 bone transforms or 30 of them. and 1-2 colliders.
actually I need to check something
I could easily redo most my dynamic bone efected riggings and make medium at least
nvm I thought having avatar swaps might be counted as extra meshes but nah they have no excuse
lmao
lol
this is so sad
coat, hair, shirt
you could probably do those with less bones
you are correct
is there like, a bunch of bones in the hair?
no
The new? documentation have some new videos how to merge bones and stuff, first time i done that was today so learning something new, not that im using much usually around 55~ dyn bones but now i managed to get one of mine to 33 i guess thats,.. progress
learning is a good journey, you're probably lucky to be learning all that stuff now when it's a big fad
will be more fun
also, adding exclusions is also a good trick to reduce bones
ew
like, the _end bone of a chain doesnt really need to be dynamic
might be good to note that model is from 4 - 5 months ago
blender auto adds end bones i believe
https://i.gyazo.com/5b8be00831c5c6545bce164bcfaca0c4.png
only thing holding me back atm from good/med.
they can be excluded to save some transforms
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 purely informational at this...
I've done every possible sick experiment there is to do with dynamic bones, they're my favorite fascination
heres the exact stats
Can somebody help me ? should i remove the bones or keep it ? i want this ? Or should i just change parent to legs ? https://www.youtube.com/watch?v=8u1NBstIcy8
[[ENABLE CC ON THIS VIDEO]] Quick rundown of how I set up a skirt for VRchat that looks good and wont make everyone's frames rate tank. Currently there is a ...
Thats one thing i want to learn how to do, but havnt yet
those bones seem fine but you'll get clipping if using dyn bones from what I can see
Yeah i have clipping how do i fix it ? i am new
you can try parenting to the legs but it won't look good, a skirt that short won't look good even if you go for colliders, best to get used to some clipping
if you're going with the big performance party try parenting to the legs
So I keep just put 3 bones to left leg and 3 to right leg or i just put one to each ?
you can try three on each side, make sure you save first, it's all about trial and error
and I highly recommending getting into learning weight painting sometime because you'll eventually need it or at least an understanding of it
it big helps with this stuff
Well i try it if it will not work i go sleep 😄 thank you .. do you know some good tutorial for weight paint ?
hm, can't remember its been too long, but I mostly learned from looking at other models and trying to emulate what they had until I learned what each setting meant
mostly from trial and error
but youtube has a lot of tutorials to get you to that point
I recommend stuff like watching people add clothes to models
I can't talk too much about this but looking at professional made models, not mmd models, is better for learning if you catch my drift
I understand .. 😃 i will check that out when I wake up .. I just uploaded the model i seen clipping so i was like how do I fix that xD well tomorrow is an another day
o yeee boi learning about this stuff gets pretty addicting when you realize the more you know the more power you have to create more etc it's a vicious cycle
speaking of professionally made models
@fluid turret so true
I wonder how much TDA would charge to make one custom 
More than a skilled Japanese guy I'd say
you're better off gettinga custom from the booth sellers
someone know if unity chan have dynamic bones
You mean the vrchat default avatar? Or unity chan the unity package. The latter you can customize however you want, with or without bones.
i think tupper said he added dynamic bones to the default unity chan recently
Yeah she does, you check by looking at her avatar stats. She also has nice shaders now!
i wonder why they pushed out this update that fast. i really dislike the huge gap between poor and medium. if 1 little thingie is considered as poor and everything else is good/excellent your avatar is poor. even tho if your character would be multiple times worse, it would still be ranked the same as the nearly optimal one. that..... really doesnt motivates to try your best. then you see tons of public avatars which are 30x times worse than yours. ranked the same. also i somehow dont get displayed the correct informtions from my avatar before uploading? it shows everything dynamic bone related at 0 and only ingame what it really is. guess there will be a lot adjustments coming. i just wonder why this patch was so rushed^^
Oida fyi they are looking for feedback on the new system so feel free to make a post on the canny about it.
argh yeah x.x i constantly forget about the canny :x
would this be the channel to ask about some unusual shenanigans happening while texture atlasing?
Probably best to share in avatars
green gang
😂
green gang master gang
scrolls up and sees the picture of the avatar with 100 meshes and 300 mats
what in the actual F....
Just wanted to give a shoutout to @sudden jewel and the dev team for the latest update. It's actually really nice to have some solid avatar standards we can build around, and listening to the community to adjust them to be a bit more to our expectations. <3
I'm really hoping this is all moving towards VRchat ending up on the Oculus Quest, because mobile VR is definitely the future and I'm excited to see more folks get involved with the platform.
Thanks again for all the hard work you are doing. 
Nice :>
lol 65 materials, bro wtf 😂
it appears we know now where the lag is comming from
remove this rank system
btw
nice way to add a way to steal private avatars back
new .DLL btw
Turn it off in your settings and move on
There's a new atlas feature for the blender. How good is it, exactly?
And how should one use it?
I know, but in the past when I made an atlas, it has ruined the eyes due to the necessary transparency setting for f example eye shines.
Im wondering if that part has been fixed?
It is said to leave textures that need transparency in their own atlas
mhm, so what textures can/should be merged?
should clothes and skin be separate, should hair be separate and should eye/eye shines be separate?
The way I do it is by separating the mesh and and grouping which one I want to merge together. I have notes I write down for which to merge but like what atlas is saying. hair, face, and the rest atlased. 3 mats
alright, thanks. Im going to need to make it 4 due to the necessary transparency for eye shines, but this helped, thank you!
Now the only challenge is identifying the textures when the names are all confusing mumbo jumbo
rename then
aight, ill give it a go. Test two ways.
doesn't look like its letting me change the settings in... unity..
Now the settings for changing shaders/transparency are greyed out...
Maybe try
Select FBX -> Inspector -> Materials tab -> Switch from Internal to External (legacy)
Please don't torture my bots like this

Okay I take it back. Every single one passes perf checks.
Nobody has any excuse to not optimize if this dude can optimize 743 avatars
lmao
Lol nice
Do particle materials and weapon materials count or is it just unique avatar materials that matter Tupper?
Hey Tupper, is there a detailed explanation we can link to people when the topic of dynamic bones and colliders come up? Because since the performance ranking system, we see a lot of people in chat say things like "they have virtually no affect on performance". Just wondering if there was something we could link people to in these cases.
any materials on the models at all count
particle systems, meshes, whatever
@heady smelt I should do a video on how bad DynBones are. Its somewhat "complicated" and requires showing off the Unity Profiler
Alright ^^
I assume they affect more things than just visual lag, that most of aren't aware of?
The amount of 30+ material avatars i see in a lobby is tremendous
Yep :>
Most ive seen so far was the one i posted above^
i need have update sdk
@heady smelt I've recorded a video. Uploading it now.
Oh awesome! Thank you
New Tupper tutorial incoming?

Explanation on dynamic bones and colliders impact i think
So I went and checked out that world that had 700 avatars that you said should "set an example", it was just a bunch of pokemon with no gestures or anything to them.. of course they're being to be optimized. The models probably only came with a few materials so optimization was piss easy.. but people are adding weapons and particles and optimization becomes difficult because people want to make the best avatars possible, not generic pokemon with an animation to them, so to sit there and say "omg well done for importing generic pokemonz youre an example!!" shows how out of touch you are with avatar worlds lmfao
I had a bunch of bots scan the world. I didn't eyeball them myself. ¯_(ツ)_/¯
If you add more props, animations, particles, effects, etc, you're going to pay for it with optimization costs.
That's how it works.
I think you're being a bit harsh saying I'm "out of touch with avatar [creators]", but hey, that's your call
btw the chairs dont impact perf right?
not really, no
because when a friend sits on me i drop some fps and the chair is positioned in the wrong place for me but not for them
@heady smelt https://www.youtube.com/watch?v=kL1MdV46gdE
So one day in the #avatar-optimization channel Jamie Vatarga asked me this question: "Hey Tupper, is there a detailed explanation we can link to people when ...
im very proud of the face in that thumbnail
very proud
why can't click build & publish
Click the pause button at the top of the Unity screen.
thx
rip
additional note: dynamic bones is written in a way that if your performance is low, it makes performance worse
@heady smelt
it tries to "catch up"
That's the eventual idea
will take time though
either optimize DB itself, or replace it with a drop-in
Hmmmm, guess i'll be using ones with no dynamics for now
I mean, feel free to use fancy avatars with DB. its fun to use DB
just don't use it in worlds where your perf is already low
or populated places
What im doing now is uploading two versions of my avatars. One with my normal amount of dynamic bones and one with them significantly reduced. If the worlds getting a bit laggy I switch to my more optimised version. I love my dynamic tentacles on my avatars but at the same time i dont want them ruining everyones fun if it gets to that point.
yaaay, cake!
Nom nom
Gasp
Lol
Maybe someday, If the servers can handle it, db will be calculated locally and then send to everyone like the IK
Yes, like everything else
Yikes... 
Ok not everything else, but you get the idea :D
Yea. I get it.
So, how are they gonna get the funds for buying their own servers?
They gotta come from somewhere than just investments or merch...

Yea.. somewhere in the millions. I doubt it would last for years, tho
4millon, but that was 2017
I think we should take this to random chat...
I bet they won't let their most lucrative game die


ok thats enough
well even if it was a large sum of money its spread thin until they can find another sponsor 
Hmm.. Maybe Microsoft?
They got a big gig on VR with their MR platform
So is Valve..
Spread thin? They're staff is still pretty small, no?
wow this video is nice
best christmas gift
im in vr so gotta finish it later, but are what other videos are you thining of doing? or did he say in the video
with some messing around with variables and a little bit of mesh splicing i managed to get a nice flow from a cloth component instead of using dynamics, might be able to do that with hair and skirts. can't say much about the jiggle physics though
don't use cloth on hair .... first of all it looks like crap and 2nd it's laggy for hair, unless you have plains and not volume for hair
fuuuuuuu
cloth hair terrifies me
though if you did it how im thinking, thats impressive
Anyone know a way to make hairy physic without using dynamic bones
i dont think you can
than, if dynamic bones have a 30 refresh rate instead of 60
does it make a difference
I'm not 100%, but I've heard changing it only changes it locally and others always see 60, but I'm not 100%. I'm fairly sure the game caps it at 60 though
We cap 60. Feel free to go lower
tupper, could you dev's imagine that it would be possible to use the station for yourself? like i place a chair with a fix joint somewhere and i can sit on it
do you guys cap also cloth? i'm doing from 120 to 10, and there's basically no difference
@static pecan canny was closed asking to be able to sit in your own seats, so no plans for now or at all
were there reasons for the closing?
I think it was along the lines of, it would be too confusing and problematic for the average avatar maker who would stick a chair on their avatar and then accidentally trigger sitting in it themselves
I think that could be solved by having a checkbox on the station component, or maybe even a config tweak like the particle system limiter
Or an option in the options
i see no difference to chairs on avatars where other can sit on, same work how to do it
Chairs in their current state are already slightly confusing because there's no ready-to-go prefabs or anything and it's tough to get them to work correctly
@static pecan the difference is that if the chair is attached to your avatar, you might fly off
But then again, they don't forbid chair loops either so
It only happens if the chair is not on a fixed joint and not at 0,0,0 on your avatar
I hope they will reconsider. I had a lot of ideas for fixed joint setups and vehicle avatars
was it tested earlier ? i mean the issues they would comming with?
We don't know. They did have a closed beta
hm... i would like to test this by my self or get answers from people they were able to test this
Well, I tested it a few months back, the last time chairs were introduced temporarily. It worked pretty well as long as you put it at 0,0,0. Putting it further behind me made me fly off fairly quickly. IMO this should be an "at own risk" thing.
I think it needs a checkbox or something like that. If you put a chair in your hand you don't want to constantly trigger sitting in it all the time
Or maybe self-interaction should be automatically allowed if the chair and its entry point are exactly at position 0,0,0 in your avatar. That would be another way to toggle it.
i see, but the little risk can not be the reason why they dont make it possible, if you just put a collider on your avatar and forget the "as trigger" check in, you will also fly... what i see is it just needs enlightenment how to do it correctly for everyone who wants it
Yeah
Lol, there's a remark on the public world request history about a world with 743 optimized avatars
Message makes it sound like he manually put in the work to optimize it, tbh I am doubtful considering they're game rips that are already optimized
Still though, it's good
Sounds 'bout right
Maybe. Game rips are not necessarily optimized after being pulled out of a game.
That does depend on the game yeah
If it's from an MMO it'll most likely be separated into sections, or a game with heavy emphasis on customization, that also applies to textures, even though they may be atlased, most will only be per mesh sections, if it is a game with little to no actual character customization then it'll be most likely relatively optimized
sees the video what about the people running around with craptons of meshes and materials?
yup
Hey guys, is the animagion override consideres an aniamtor? Wanna try snd keep it to 1
I believe that doesn’t change it
unity not best engine for physics stuff
Real-time anything will always be intensive.
Check out my realtime skinned mesh renderer
Available here: https://www.assetstore.unity3d.com/#!/content/117773 Using Shader forge and Unity 3D I created this Hair shader that is simple and effective ...
shader that add physic to hair lol
I do appreciate the video tupper but I do have one issue with it. Last I checked most people that use Dynamic bones don't put them on the entire model. Most people put them on the head and skirt and tail. I kinda feel like this is a rather unrealistic representation of the performance impact per model on an average basis.
what's more realistic is running around with crapton of materials since so many people are too lazy to merge materials and meshes
^^ For sure.
if I would've ran into that jerk with over 300 mats I would've just blocked them completely on the spot
that's just pure laziness
it's easier than ever to reduce your materials and meshes
material have less empac then dynamic bone
I would like to see another video where he does the same thing but on like the head/hair and skirt.
not when you have people with crapton of materials
Materials and shaders can really make a bad day when misused.
20 pass fur shader!
I understand if its like Dynamic bones on little things sleeves or whatever but putting it onto the entire model is a bit unrealistic.
show how to open the peformance window
I already have
But my model is not exactly accurate
My model not only has hardly any bones for the hair and tail but as a whole is really optimised.
My model dynamic bone wise is a rather good if not best case scenario. Kinda can't test it.
My hair to is rather short.
and some of it doesn't even move and no matter the settings its really stiff.
When he puts it onto the entire model legit everything is getting processed like both eyes glasses if he has those in the head and a whole bunch of other stuff people NEVER put on dynamic bones on.
Whens the last time you went into the great Pug and saw a bunch of Jelly people waddling around?
better chances of running into a 50 material monstrousity
Lol or a man with shaders from 5fps hell.
@manic owl what u don't realize that people have more than 100 dynamic bone checks, as per tuppers vid
starting from 200 to 1000+ transform checks
Can I have a player mod for “disable dynamic bone on this world”
better off with an option like that and a "autoblock avatars with such and such materials and meshes" option
Autoblock based on avatar performance is a terrible idea. Second Life did that and it ruined creativity.
They just calculated an overall weighted score and let you autoblock above a given score of your choice
And each client could see what % of people in the world was blocking them
It was... terrible :p
If you want to block people. Do it yourself
Are unused vertex groups still saved by unity? Would it be a micro optimization to delete them before exporting as fbx?
like just stuff made to make selection easier
no relation to rigging
I don't think Blender exports empty groups
And Unity will probably ignore them too
You can delete them but there's little point
well i know it exports them because i can reimport the fbx and they're still there
just trying to make my avatar load faster
Are you embedding the textures? @amber hemlock
Try not doing that, it will make the FBX larger
Unfortunately one of the things that inflates FBX size is embedded media and also shape keys
Separate meshes would actually make it load faster
???
separate meshes instead of shape keys
oh, well VRChat rates that badly
Which would require separate meshes so that isn't worth it
Yes
Just make sure you export without embedded textures and you'll save a lot of time. Otherwise, the texture has to be loaded twice.
Yeah my optimization workflow -> Separate thing to be decimated -> delete it's shape keys -> decimate -> rejoin
I only embed the first time these days to make it easier
What is the point of embedded textuers
and how do i make sure it's not embedded
i potentially have a ton of stuff to reupload again
Do you use cats export?
@amber hemlock
If so, look at the settings in the sidebar for a checkbox that says "embed textures"
Thanks
If you use manual export, make sure the path mode is set to Auto in the bottom left
do the amount of bones really matter even if there are no dynamic bone scripts?
They still have to be rendered and rendered based on their positions, it would be better to bone merge, especially if they never move independently
I hope Unity 2018 LTS comes out soon!
They heavily improved "Optimize Game Objects", which basically makes the bone count nonexistent. In 2018, you can still expose bones that you need, such as the hands to add accessories.
That way you can achieve bone counts like "3" on full humanoids
There is a difference between "unused bones" and "bones that are weighted but you never move" Cats will only remove the first category. the second category should be bone merged
mhmm i can do that i dont use dynamic bone anyways
I swear I learn something new about cats that makes it even better almost every day
guess who accidentally saved a dedimcated version of his model overtop of the undecimated one and has no backup <----

someone tested the hair shader?
annnd i only just now realized the decimate modifier has a vertex group parameter to only decimate verts in that vertex group
so i could have been manually making vertex groups that omit knees and such
(I don't own the model) But feels good 😄 Learnt quite some stuff in optimisation
lmao 😂
Does it actually improve performance in VRChat?
Discussions about it on old unity seems to indicate it didn't really help
It helps a little
It gets rid of excess game objects and also unmapped bones in the model
This is a model with only the left wrist exposed to put accessories in
I don't really understand what it does, surely the gameobjects must still be rendered
What it does is make sure the armature doesn't end up taking a lot of game objects
And will also prevent "unused" bones (such as face bones) from appearing at all
It'll be sorta collapsed into the head
Ruuubick used this a while back but stopped when he discovered it disallowed usage of shape keys in animations
do blink and visemes still work?
If you expose the Body then yes
Visemes will probably work fine regardless, but animations may not
S'fine i don't use animations
Visemes have a hard reference to the shape keys whereas animations do not
The built in blink will probably work as well
Side note, this will make it incredibly tedious to use dynamic bone unless used in tiny amounts
Since you'll have to expose every dynbone transform individually and the UI is probably worse than the one used for Animation properties
When I joined two armatures in cats, it made a ton of new vert groups:
https://i.imgur.com/GseDPpL.png
And now, the hands no longer move with the arms in pose mode. any way to fix without redoing it?
basically i copied my undecimated hands from another blender file and then merged them with cats
but it didn't seem to weight them to the new armature
You could rename every vertex group (based on the bones present on the main model) and attempt to re-merge it, manually would probably work best in that case imo
So I noticed something really weird with the poiyomi shader. It seems to double the draw calls and triple the amount of polys being rendered.
o.o
As you can see out of the three shaders poiyomi gives a different result
This both happens with poiyomi set to double sided texture and single sided
I've heard often that the poyomi shader was bad for that reason
Yeah, it's sad because it gives me such a good effect on my Broly Racer
How do I bring that menu up, I use poi?
This is big oof, as I use some of the lower settings in there : <
Indeed, I'll probably start migrating to noenoe or xiexe
Hey ruuubick, i've just tried to test the profiler, apparently the only lagspikes that occur whilst having the model selected happens with the inspector in-view
Without the Inspector Visible
https://gyazo.com/cad82edd511538c9e7c261485f185b0d
With the Inspector Visible
https://gyazo.com/f6f27c080296983b253f07efe4b454ef
Well, in poiyomis, outline is in a different draw call
And yes in sdk UI drawing has massive lag spikes when avatar root is selected
@marsh trail So you hide the inspector window ? Is it because it also has to display all the transforms numbers in there ?
Don;t know if its from the last one or since 2017
My dynamic bones are all separated on a per-root basis, none are listed under standard root
Sure but the transforms are still updating if you move the entire skeleton, so they had to update and display I think
If there's any other reason I'm interested
That's all that's inside the inspector
It could be the transforms but the lag is certainly something else
You haven't tried with the profiler like I suggested ?
Can you send me the ID so that I can try for myself ?
Uuuh, cats is warning me that I have 6 materials (atlaser says i only have 2) and 6 meshes (I only have 1)
Oh this happens on every model ruuubick
I am using the latest SDK and the correct unity version
Then the problem is most likely on your side then ? You tried resetting settings in unity or looking it up on Google ? Haven't heard of that issue before and there's monstrous avatars out there
Sorry for intruding, but does one know how to prevent hair from clipping through the model? I thought adding colliders would fix it, but it still persists.
Change the colliders / bone values, its made for that
From what i'm seeing it's definitively from the inspector and i really have nothing that messes with it o-o
Even though this makes no sense
or ignore it as no one really cares much 
I guess i'll have to look it up myself but it's really unusual behavior that never occurred prior
I've also found that the inspector is sluggish when selecting the root of my avatar
Profiler says I have 7 calls, but I can only find 6 :<
Prior to the 2017 update that is
For me it started with unity 2017
I don't have DB in my model and Inspector tab drops my frames too lol
Aye, it never occurred prior for me
It is bearable now that i know i can just hide the inspector, but it's horrible for play-testing things whilst changing values
I've been seeing people with pens have super huge bounds. Shrinking down the mesh the marker stuff is on would fix that wouldnt it?
@shut sequoia nope
The problem is that the marker has to be pretty far away in order to use it properly
To make the writing stay, the trail renderer is moved very far off rather than disabled
And the shader makes sure that trails made at very high speeds aren't rendered either
So by default it's off in nowhere, and it shoots to your hand when activated
It's actually not on the marker mesh to begin with (most people don't use markers)
@grim basalt this does not appear to be an optimization question
Hit the Play button and retry? Never seen that happen
I might just disable the marker on my main avatar until there’s a proper fix for this
There probably won't be one
Unless they removed the bounds check which is unlikely
@shut sequoia you can't really shrink the marker stuff, it breaks the invisible lines and adds a lot of jaggy endsegments to each line
for now it will make any avatar rate as very poor just by using it once 😐
You can use a vrc inventory system emote toggle to turn the pen on then you can draw with it
I usually do that, though you have to be careful to reactivate it for late joiners or maybe make the pen gesture trigger the enable animator instead of using an emote
@manic owl I shoved it on the hips to simulate more players than were actually present. That's why I phrased all of my references to the total number of DB transforms in scene rather than the number of avatars.
" There already is one. If you're farther than 20m away from someone, it does not animate. " What about while looking away? and what's 20 meters ingame look like, any examples? I'm already terrible with eyeballing distances
@marsh trail Lag/FPS spikes in editor literally don't matter. What matters is frame timing on things that aren't marked as "Editor Overhead", which you figured out. You can also look at timings for render time and the DynamicBone scripts.
@shut sequoia 20 meters is 20 meters. /shrug
They also don't animate if the user is culled (as in, you're not looking at them)
oof, bounds RIP
My avatar is about 1.6m tall, so 12.5 of my avatars.
Yes, that's why bounds are important haha
Oh that's really far, damn
It is, yes
Might be cool to do a gradient thing. Cap DB UPS at 15 if you're 20-15m away, cap at 30 if you're 5-15, cap at 60 if you're within 5m
You can't even see that very well when in VR, would reducing that help at all? I'm guessing only on large maps without lots of walls, like outdoor ones
🤔
yeah it'd help a bit. we'll likely do something like that once we do a pass on DB perf.
Its next in line after we get networked IK sorted. 😄
Also need to fix the behavior where it tries to catch up over lost frames.
Right now that basically causes a positive feedback loop of frame timing.
🤤 We'll probably never get rid of excessive DB but at least stuff like that will help
DB says "FPS is lower than my update rate, so I'll do 2 updates next frame." DB goes next frame: "FPS is lower than my update rate, so I'll do 2 updates next frame. Oh, and I still have 2 updates queued from last frame." etc etc
I'm working on getting my world ready for resubmit with over 100 avatars. I think i'm about there but could use a rescan to see what i missed. Is there a way to get that done without submitting, failing, then waiting 24 hrs?
sure, give me the ID here or via DM
Running scan
thanks a ton.
i'll basically delete the ones that won't pass and get it submitted for live.
After uploading to the main ID
i wish i could optimize avatars
What's preventing that?
my knowledge of avatars
@distant forge yeah but I don't want the whole inventory system for just a marker ... -__-
@halcyon vessel unfortunately it looks like some of that asset requires scripts
Also Tupper I plan to replace IDs over the next few days on my world. But I am optimizing and removing colliders on everything for example. Will this cause issues ?
It shouldn't as long as you're ensuring you're below limits.
100% am only uploading after it's optimized.
I've caught a handful of people uploading bad avatars after being approved. They were suspended from world submission for a month. 
ya that's why I ask cause I don't wanna have that happen to me.
oof 
All that effort to work around the avatar test and they could've just, you know, optimized
hopefully it never happens to knuckle avatars 😎
example i just atlased a 20 material into 1 and got bones down to 60
yeah that gaming the system and they know it
@sudden jewel oh ok
yea knuckle made best avatar maker in vrchat
after***
@tired badger Anything orange is over limits.
I wish I could be the one to push the "ONE MONTH SUSPENSION" button
I'd pay to have that job

gotcha most the ones i expected from what i see.
Is that the whole list? Feel free to DM full list and/or excel doc of it? I am labeling and tracking everything in my Google documents.
It is the full list. I will DM you a CSV.
You can keep that there as an example I don't care if that info is public for others.
thanks
damn just 15-ish more 👌
oh wait that is a full list of just the offending ones right?
cause i have wayyy more than that in there was confused
dynamic bones 2.0
That's all avatars in the world with any duplicate IDs removed.
Thanks
wonder if hair animation shader be less cpu then
bone
if somone could make hair shader has animation less cpu usage think people would use that
where u get that list oof
is it from all reported users or just a survey of peoples avatars
That's the list the Tupbot9000 spits out when it scans a world
they use that to approve/decline world, he just wanted a list so he can fix his avatars before uploading so it doesnt get declined
@halcyon vessel The problem with GPU based physics is that the most efficient way to apply the deformation is in the vertex shader. However this requires custom shaders and anyone writing their own shaders needs to include the deformation code in their shader's vertex shader.
I remember people talking about the idea of making something like gpu dynamic bones, but if i remember correctly they mentioned getting the calculations from the gpu to the cpu would probably not be any faster than dynamic bones currently
There's also the problem that without blit we'd have to use cameras
You could do the calculations on the GPU, then read them back to the CPU and apply the transforms but reading data back from the GPU is slow.
yeah
Hello TCL how are you doing, havent seen you in ages
@uncut maple Pretty good, I'm on vacation in Japan right now.
sounds fun, hope you like it
ok
@ancient crystal Yes, there's compute shaders.
avatars that are extremely very poor
@ancient crystal define GPU access?
access to run instructions on the gpu in a script running on cpu
@empty pier not bad just chilling, i have been chilling, i dont touch vrchat files anymore
That's basically a compute shader.
@ancient crystal ah okay, shaders and rendertextures pretty much so far
Compute shaders are flexible enough to run arbitrary code, within the realm of what GPUs can do at least.
we can't use them in vrc without scripting though right?
Correct
but the question was unity, so yeah
You need scripts to allocate and populate buffers, as well as dispatch the compute shader.
a wavy hair shader could be fun. I think you might still want a tiny dynamic bone chain to learn about velocity and acceleration but apply that chain as offsets to all the hair vertices.
but the issue is that first of all using dynamic bones like cloth would be much much slower
@distant forge but ... my 90 frames ...
the cloth calculations are already a lot faster than each dynamic bone transform
so imagine 900 dynamic bone transforms
true but people put cloth on high poly meshes @ancient crystal
Might be possible to do without render textures by making each hair polygon a tri with the actual position and the two bones in the dynbone chain and some vertex color or uv information to include how far down the hair it is (how much to amplify the waviness)
as opposed to 30 cloth verts
yea but they said
"put a dynamic bone for each vertex"
No it would just be 2 or 3 dynbones and the shader uses a single dynbone to skin all the hair like wind information in a grass shader
um thats already possible
Is Unity's cloth CPU based?
its called weight painting
u literally just use smoother weight painting
and use like elasticity curve or something
cloth hair also works pretty well if done the right way
ive done it pretty well with long hair but it takes a while to make it less buggy
^
Hmm interesting maybe that's sufficient. People have like 5 front hair strands and two back hairs chains and they all have their own dynamic bones. Should only need a few for the whole avatar for simple waving
yea but also
dynamic bones is not cloth
its a major distinction
no matter how u set em up they won't act the same
so cloth hair and dynamic bone hair are very different
@empty pier I think unity cloth is the physx one, maybe you can confirm
cloth hair can be brushed and picked up like actual hair and has more realistic gravity
If it's physx shouldnt it be able to run on the gpu
as a fallback it might lag more, be a bit more buggy, and collapse in on itself
@heady smelt Seems plausible, which would mean it's CPU based.
@distant forge depends on how they iplemented it
but if vrchat could get good shader programer to make them one might be less people use dysamic bone for hair
then just use shader
dynamic bones are for more cartoonish hairs that are just wavy and wiggly not really interactively realistic
I think they store all the points in a texture, just like the vanilla particles
yea but its still a shader
but solve for colliders on the cpu
unless they make a "one shader fits all"
what if i want toon shading with a custom ramp
or standard shading with rimlights
etc
if the cloth did i think everything else would have to also
unless it did cloth simulation separately
which it already does but like
oh, GPU physics ... I get ya now >.<
If the deformations were applied in the vertex shader then VRChat would need to provide an include file for shader writers to use to add the same functionality to their shaders.
yea but like nobody would do that
yeah it's hybrid, the shaders read from textures and do some work, but the heavy lifting and rigid body colliders is CPU afaik
unless ppl givin em out
like portal shader for example
ive never seen people with portal shader use any other materials than raymarchers theyve made (or gotten online) or toon shaders
even tho i expected to see other materials like standard or lux
the portal and the mirror shader should be part of the SDK if you ask me, but it's up to vrc ... you can get both if you look around a little though
I don't think people should mess with portals though, they should all look the same so the language is clear
It wouldn't be too much effort on the part of the shader writer to add the include and call the appropriate function to calculate the deformation.
yea but for people who just re upload crap
I think they mean stencil portals
VRChat would have to provide the include and instructions for shader writers.
or like the bunch of people who hate or don't even want to use blender for some reason



