Load Test 6 October. Some brainstormed ideas on how to improve the experience


I just participated to the event with around 420 people in The Spot. While I was experimenting a very good sound, I hit the problem with the white spheres for 90% of the experience, until I luckily read a board where it was written that “if you see too many white sphere you should try to change domain and coming back”.
After that 99% of the avatars where visible again.

While I was there I noticed that there were (so far as I was able to spot) some activities going on:

  1. A person was hunting the scene trying to interview the people with in depth questions on social VR, this is a very good idea. I suppose that he was from the High Fidelity staff, but am not completely sure.

  2. A quiz was working on some part of the half hour before the event. That was a nice idea, but I found the questions especially for a non native English speaker a bit complex to answer. Also a bit nerdy (some of them were about minecraft crafting).

  3. An avatar context was being held and if I understood correctly ganeesha did win.

  4. At the end Philip did come down to the crowd and handshaking people for friendship. I couldnt remain there after almost two hours of vr I was a bit exhausted.

  5. A lot of check-in helper were available in the scene, but they were almost under-used since the only real check-in information was that people should have logged-in with the login they specified in the eventbrite ticket.

So what in my opinion can be improved in these stress tests? I try to list some ideas and something I had problems with:

A) Important notices should appear in the huds of the participants, like the tip on the white spheres. I would suggest that every participant should have a script loaded where important notices and information should be displayed, (and possibly also the number of people connected that can be quite interesting).

B) Check-in helper can be used better to animate small groups instead of simply staying there without having specific tasks other than saying you are already checked in. Some of the helper were not English speaker, I understood that at least one was Indian and another was Spanish. Maybe not English people can display their nationality so that they can become an attraction pole to begin some interesting discussions? Since sound and audio is really amazing, I think that this should be heavily exploited.

C) I was not able to see the name of the avatar around so it was almost impossible to find my friends or other people from Italy. people button on the hud was completely unusable and quite cumbersome especially in vr mode. Here again I suggest to have a script with avatar tagging enabled by default.

I think that giving a scope to the participating people and avoid having all of them crowded in a unique spot where @philip is, and trying to animate them with professional social extrovert people can be a good idea.

Just brainstorming, but if HF won’t find something different from just “hanging” there (and many people did actually hang in the air), the entire experience risks to become a little boring and people might choose not to come again…

My 2 cents…


I just dropped in and there were 65 people there, all of them spheres. I went home and returned… all spheres. Waited, non of them rezzed.


As I wrote I was for 90% of the event seeing only 1/20 of total avatars, and only after going to another domain and then back the situation slowly but decisely improved.


Claudio has some good ideas.
Mine problems was during the tests, was with the sounds. Sometimes suddenly everything went silent, and i couldn’t hearing what’s going on there, and after a minute the sound came back, but the avatars disappeared then was coming the white spheres, but everything became normal within 1-2 minutes.


I had 1 crash trying to open avatar settings. Only saw the spheres briefly after that
The audio attenuation made chatting difficult at the back away from the stage it was so hard to hear the people right next to me
The event feels a little unplanned greeters were good but large groups of people need structure you are left with a ‘is that it ?’feeling. Everyone leaves rapidly leaving only the same handful as allways.
The load test gives me with the same impression my cat gives me”I love u if you give me treats” no treats then you’re dead to me…


Let’s have a talkative DJ like Tha Phlash perform & maybe host a dance contest or such - either directly in TheSpot or in Rust.
Once we have hundreds of people inworld we should build on their momentum and keep them entertained as long as possible :purple_heart: :woman_singer: :sunglasses:


I think my experience get a bit wasted by the rendering load. What it will be with 600 or 1000 peoples. We will certainly reach a point where this might affect even the best existing machines.

I wonder how better it would be if for such crowded events we would request people to wear very low polygon avatars. As some event request a dressing code in real. (I thing it would be acceptable to have avatars specifically designed for that, if the rule is the same for everyone.)

Just an idea.


It might be prudent to identify what kind of events have large numbers of people and treat the avatars a bit differently. I question the success of this test since so many people saw many orbs instead of avatars. The avatar server seemed to be still a bit overwhelmed afterwards since I was unable to see any of the 65 remaining avatars at the end of the event.


Regarding this… I’d 100% hate it if this was the case. I know people love their nametags and all, but here’s one tiny problem with them: they’re only useful in small crowds. Once you have say a massive amount of people, this completely changes and you just have a blob of tags making it EVEN HARDER to find people.


If anything, improvements to the PAL would be a better step. Having it progressively load/cache the people and connection list would be a good step while also replacing the PAL spheres with something less intensive to draw.

In fact, here’s how I’d approach the connections issue (since I have to scroll down to load more):

  1. At the start of the script, begin checking for connections, if we are signed in.
  2. Pick a threshold number of how many connections we can request for at a time, assuming the API has support for this. Also, pick a threshold time per check as to not tax the client.
  3. Prioritize updates for users who are online with their visibility set properly while doing lower checks for people who are not.
  4. Repeat the same for Nearby.


Apart from the white sphere esperience, when posting this post I was trying to consider how to improve the quality of the experience from a “non technical” but “social” POV. Do you all think that how it is organized now is completely ok? I mean, kudos to the ability and all the energy that the organizers invested in this event, we can all see this, just maybe there can be some easy to be implemented slight modifications to how the people interact, or will simply left it to the user creativity like it is now?

Probably because I myself am deeply shy, I found difficult to really interact with people around, am I the only one who noticed this thing?


I am concerned from social pov, as I stated in my latest comment.
Since there are many people sharing the same AV I would be interested in knowing their tags at least people close to me who are them. If you do this for only the 4/5 people directly around you it would not pollute the experience, and after all it would be just a opt-in thing.


Perhaps one thing that could be done is to make better use of Nearby and Connections, like “Highlight people from my connections in the same domain.” This could also be a setting for any nametag makers out there.


Perhaps a tweak to the radar app might help here. If it had a checkbox that would eliminate everyone but your friends it might at least give you a clue to where they are relative to your position.


Another thing to add is that I feel from this point on, the stress will be less on the servers and more on the clients, specifically rendering.

I’ve said it once, others have said it more times, so here’s one more time:


Not THE CURRENT LOD controls, NEW LOD controls. Any engineer can step in on this, but if I’m not mistaken, the way the current LOD system works is to pick materials from the avatar to hide as the complexity increases.

Okay, so here’s an idea: let us specify the order for you! If no order is given, then use the current LOD method. Everyone wins!

Now before I get the bashing of “That’s not what I call LOD - Volume 100,” I’m well aware. As per a previous meeting, in which desktop echoing was mentioned, there seems to be a love of taking the most complicated route on developing new things. I totally get that introducing a true decimation system would be the best method, including mentioning varying LOD states of the mesh. The thing is that fits into the desktop echo issue, where the solution is to make a complicated algorithm versus just letting users adjust a threshold control… like 90% of other systems with an always-on mic system. I’d rather push for an adjustment for a pre-existing system rather than ask for something that doesn’t exist or did exist and was broken.

On another topic, the LOD animations thing, which is currently mentioned to be a bug, I can also perhaps state was a major helper during the stress test and really eased tensions on the avatar mixer. However, when @philip did the wave, it was painfully obvious that it was on and I had to guess when to put my arms up, since he appeared to be a slideshow just because I was not close enough.

I do think it’s a good step in the right direction, but if it is applied, please interpolate the rotations and perhaps make use a viewpoint priority system. If it’s determined interpolation would be intense on the client, make it an option (which is honestly another issue I’ve seen HighFidelity seemingly not implementing like everyone is afraid of it). In addition, if a camera (keyword there) does not see someone, then their priority isn’t really high. This same application could also apply to the spectator camera, which is why I say “a camera” and not simply the viewpoint of the avatar. Not to mention, most users in desktop mode are the third person.


I was a bit late to the party at around ten minutes, got me thinking if I’d have arrived earlier I may not have had the problem?

I went to many other domains and returned but it didn’t make any difference to me. Everyone was still an orb. Another thing I noticed was the sound wasn’t spacial, everybody’s voices were hitting my ears at the same time, was very overwhelming and dare I say annoying.

Hasn’t put me off though and I’m sure all the problems will get ironed out eventually.