Side-loading? Ayooooo less gooo <3
Also thank you for fixing the scroll bug again, it’s been annoying me for so long <3
Avatars already get unloaded from VRAM after five minutes of not being used.
oh i see, thought for some reason they’re not - good to know now!
Can texture streaming be supported soon?
There should be an option even if it’s not enabled by default.
Related problem solving is slowly pushed by the community.
Because for medium and large worlds and most avatars that are not good on VRAM, this method can effectively solve the problem of RAM and VRAM.
Save many or more than half or even more on personal tests! We don’t have to pay for objects that have no activity and are not in the frustum!
Are yall still experimenting with trying to enable quest hand tracking on PCVR?
If yes, will you still support the quest 1 for that?
Do the existing programs that stream to/from Quest 1 track hands?
It sounds like Quest 1 application might no longer get update via store?
I’ve only seen people talk about hand tracking on Quest 2…
No i was talking about them experimenting with trying to get quest hand tracking working while having the headset connected to a PC.
And yes the Quest 1 supports hand tracking, in fact it supported hand tracking even befor the quest 2 got released.
My suspicion is that VRChat is not working on thier own air link type program, and so if the programs that exist add hand tracking, VRChat would support it.
VRChat probably did reach out to anyone relevant.
I guess Quest can have app side loaded so the PC fun times programs could still technically update after Meta stops accepting updates on the store.
Does this change to server side avatar analysis mean that we avoid the “performance check stutter” that happens once an avatar is done loading in?
2 Likes
Is there any plans on using the server side avatar analysis system to improve the accuracy of performance rankings?
The current system is intuitive but not very comprehensive for advanced creators. Ive heard talks of avatars with very poor ratings outperform medium ranked avatars due to how the performance ranking system categorizes stuff.
Would it be possible long term to make an optional, probably slower, avatar upload that accurately assess the real world performance of an avatar and gives it a rank accordingly? That could be a good incentive for avatar sellers to properly optimize their avatars if its more accurately rated.
2 Likes
Please, please, make avatar theft protection a number 1 priority. I cannot imagine anything else being more important than protecting the works of the community.
The current report system is far, far from enough. Please prevent avatar theft at the source rather than having this clunky, elaborate system to write reports that only English-speakers can use.
As previously noted,
so forgive me if I don’t add too many other replies to your comments 
Eye tracking and server-side avatar blocking!! Very nice work.
Also random question, but are there any plans to add the ability to transfer Group ownership?
1 Like
It’s extremely hard to prevent ripping at the source. They’re doing what they can, but as long as your GPU can render an avatar, it can be ripped. It’s just the way graphics and computers work.
Tupper from the Developer Update - 1 December 2022:
Without a doubt- yeah. It’d be a system replacement.
As tupper’s said- other uses of it are currently unclear, and will be made clear as things progress.
I didn’t get around to replying to all the different people suggesting this in previous weeks but this is essentially impossible due to “performance” being a very arbitrary and system-dependant thing. Your hardware/software combinations will vary, and it’s why most of my benchmarks (outside of the ones including spec-locked platforms like the Steam Deck) are meant to be treated as examples of delta comparisons between 2 sets of data rather than “hey this is exact numbers on how fast or slow different parts of VRChat will run for you!”
People like myself don’t have the throughput (or tooling) to be able to do widescale tests for this, and neither does VRChat- really (especially with how dynamic the content can be; I have to swap out things from my test suite all the time due to significant world changes or them just not being relevant “real-world tests” anymore compared to new content).
It’s something I’d really like to make a publicly contributable database for someday, but I’d need to first figure out how to make the tooling to ensure everything is both automatic and fair (which is daunting).
The real world performance of avatars is also heavily dependent on other things such as the world, and even other avatars. It cannot be meaningfully measured in isolation.
For the performance stats, could do some very loose tweaks, like right now I believe it’s 70k polys and 12 materials. Could allow 80k or 14 materials. Good chance that might cause more complaints though.
I’m excited for this because it implies that you can block avatars or check stats before you load them. This would go perfect alongside granular safety/blocking settings to reduce the amount of stuff that’s loaded in the client and then not used.
I do hope this would include a fix for the mat swap issue where it only accounts for textures that are active. It’d also be cool to have it include mesh VRAM like Thry’s tool does.
And like others have said, an option to calculate the active stats locally would be nice, even if it’s a button the user has to manually click. I’d imagine this is possible since the performance impact of calculating one avatar would be less than it was already doing with 5-20+ avatars at once on initial load.
Tupper please… this is HUGE
I can also second this, as I have actually seen some of the methods put in place to prevent this here and there. Kudos to the team for trying to tackle this problem, as it’s probably my biggest issue with VRC over the years. It seems that a lot of things are done in ways VRChat can’t prevent currently, as is with any UCG game. Security has absolutely improved ten-fold in the past year or so, and I cannot be happier about it.
I’m very happy to see server-side stats becoming a thing. I would love to see a few more blocking methods that I can put in place personally, such as individual limits on categories / stats, however not having to load avatars that are going to get blocked is going to have a huge impact on performance for those struggling with VRAM etc.
Keep it up!
I hope so cause i’m so fucking tired of them