Jump to content
Kyrah Abattoir

Why beginners don't learn the basics first?

Recommended Posts

11 hours ago, Bitsy Buccaneer said:

It was just a wee joke and observation. Mostly it was a joke.

For those who want to take the observation seriously, is there interest in brainstorming ways to engineer the social aspects as well as the technical?

This came up at Creator User Group last week. I made the point that Project Arctan needs some carrot to go along with the stick. Creators need more automated help with LOD generation. It can be anywhere along the chain. Tools in Blender and Maya. A better mesh optimizer in the uploader. A background task generating impostors for existing objects. A general impostor generator in the viewer. But something.

Share this post


Link to post
Share on other sites
10 hours ago, animats said:

I made the point that Project Arctan needs some carrot to go along with the stick

To me, a less punishing calculation for the low and lowest LoD would work great. The main reason for zeroed out lods is the current penalty on those lods when there are more than a few triangles. I can't stress enough the fact that an impostor is a no go for clothing or rigged content in general, as the default avatar gets impostoreed on a plane that faces the camera, thing that would not really work for dozens of different attachments. For buildings it would be ok, but as a general "cover it all" feature it's a foreseeable waste of time as it would not change the current trend on rigged meshes

Share this post


Link to post
Share on other sites
Posted (edited)
42 minutes ago, OptimoMaximo said:

To me, a less punishing calculation for the low and lowest LoD would work great. The main reason for zeroed out lods is the current penalty on those lods when there are more than a few triangles. I can't stress enough the fact that an impostor is a no go for clothing or rigged content in general, as the default avatar gets impostoreed on a plane that faces the camera, thing that would not really work for dozens of different attachments. For buildings it would be ok, but as a general "cover it all" feature it's a foreseeable waste of time as it would not change the current trend on rigged meshes

You can make reasonable "cover" lods for clothes without exploding your complexity score.

But then your complexity score will  usually be higher than your triangle count (personal experience there).

Now of course if you want to shoot BELOW your tricount, you have to dump lods.

 

Imho a good adjustment to complexity would be to have some kind of penality for objects that are composed of many meshs with a lot of "faces" that have a low triangle count compared to the whole object.

And I'm sure you'll have guessed what I'm targeting with this.

Edited by Kyrah Abattoir

Share this post


Link to post
Share on other sites
On 8/4/2019 at 3:48 PM, Candide LeMay said:

How does one deal with UV mapping for multiple LOD level if you want to create the LOD models yourself instead of using the uploader? You can't have different textures for different LOD levels, right?

I don't know how everyone else deals with it, but what I do is make the highest detail model first - and this is an "insane LOD" that never gets uploaded - it's the source for baking normals. I unwrap that, trying to make sure that I'm sensibly placing seams along edges which will persist in lower LODs.

I then start with a duplicate of that model, complete with its UV map, and remove detail that I will be relying on normal mapping to show in the uploaded model - making sure to not delete any edges which are seams. That gives me the high LOD.

I repeat the process to make the medium, low and lowest LODs, making each from the next-higher model and always keeping the seams of my UV islands.

All my LOD levels have the same UV map and while the fitting of the texture to the model will degrade slightly with lower LODs you never get tearing from having faces in a lower LOD not contained within a single UV island.

  • Thanks 1

Share this post


Link to post
Share on other sites
Posted (edited)
On 8/26/2019 at 8:42 AM, OptimoMaximo said:

To me, a less punishing calculation for the low and lowest LoD would work great.

It may be an improvement and the only way to go but I wouldn't call it great. There is no standard amount for how much LoD models should be reduced. Some items can hardly be reduced at all, others can easily take 90% or more. If the limit is raised enough to eliminate the collapsed LoD model problem entirely, the LoD system will effectively be disabled. If it's kept low enough to really encourage and reward efficient LoD, there won't be much changes from how it is today. In fact, even today's generously low thershold is still sometimes too high to reward really efficient LoD.

An official LoD reduction limit will be taken as the gospel by many creators and we will end up with lots of meshes that are reduced to exactly that level, no more and no less, regardless of how much those meshes can actually take.

 

On 8/25/2019 at 10:24 PM, animats said:

This came up at Creator User Group last week.

I'm almost tempted to start attending those meetings but it's rather painful and seems fruitless so I probably won't.

The LoD issues we have in Second Life are mainly caused by two big blunders of the past (three if we include the fitmesh LoD bugs but those are at least understandable and forgiveable). Both are so integrated into SL by now, I can't really see how they can be fixed.

The first blunder was the client side LoD factor. I'm not sure what its original purpose was but from a comment @Grumpity Linden once made to a JIRA I got the impression that the idea was to allow still photos to be taken at full LoD (paging her to see if she can confirm this). Because of it, content creators have no fixed reference point to adjust their LoD models to and the potentially increased object complexities can not be included in the land impact calculation. And of course, strengthening the LoD for an entire scene when only a few of the objects actually need it, is horrendously inefficient.

The second blunder was to not include a LoD factor for individual objects. I have no idea what LL was thinking when they decided to reject that proposal, if they were thinking at all that is. The system we have works well for procedural objects like prims but for value list based objects like sculpts and meshes it is simply not possible to decide good LoD swap distances based on size only. As far as I know, SL and its various open source clones are the only 3D engines that have even contemplated implementing such a harebrained "solution".

I'm sorry about the strong language but unless those two fundamental flaws can be fixed somehow, there is no good solution. No matter what the a standardized reduction amount is set at, it will inevitably lead to both under-reduced and over-reduced LoD models. THe only question is how much we can tolerate of each and anybody's opinion is as valid as anybody else's there.

Edited by ChinRey
Typos
  • Like 2

Share this post


Link to post
Share on other sites
Posted (edited)
6 hours ago, ChinRey said:

It may be an improvement and the only way to go but I wouldn't call it great. There is no standard amount for how much LoD models should be reduced.

This is true, however setting the lowest LoD to be at most a X percentage of the highest LoD should allow a certain degree of flexibility and require more careful attention to the degree of reduction, in my opinion. Perhaps not like animesh items implement it, but still linking the system to the amount of max triangles an items is made of.

Yesterday i went to the Content Creation User Group meeting after a LOOOOOOONG time, and @polysail suggested also a good idea, LoD switch distances to be based off screen space size of the object.

In any case, the current LoD system needs an overhaul, as it is it's ridiculous.

On 8/26/2019 at 9:24 AM, Kyrah Abattoir said:

You can make reasonable "cover" lods for clothes without exploding your complexity score.

But then your complexity score will  usually be higher than your triangle count (personal experience there).

Now of course if you want to shoot BELOW your tricount, you have to dump lods.

All this contributes to the ridiculousness of the whole current LoD system.

On 8/26/2019 at 9:24 AM, Kyrah Abattoir said:

Imho a good adjustment to complexity would be to have some kind of penality for objects that are composed of many meshs with a lot of "faces" that have a low triangle count compared to the whole object.

And I'm sure you'll have guessed what I'm targeting with this.

Yes i guessed it, my UDIM method ;)

Edited by OptimoMaximo

Share this post


Link to post
Share on other sites
19 minutes ago, OptimoMaximo said:

This is true, however setting the lowest LoD to be at most a X percentage of the highest LoD should allow a certain degree of flexibility and require more careful attention to the degree of reduction, in my opinion.

It's not very different from what we have today. There is a threshold for how low you can go and still save LI and it's actually quite a bit higher than most content creators seem to be aware of.

 

21 minutes ago, OptimoMaximo said:

Yesterday i went to the Content Creation User Group meeting after a LOOOOOOONG time, and @polysail suggested also a good idea, LoD switch distances to be based off screen space size of the object.

That again is what we have today. LoD swap distances are determined by object size and distance and that usually translates into screen space size. Such a system would break existing sculpts and meshes with extended bounding boxes but for most objects it wouldn't make much difference. Unless you're talking about taking screen size into account that is, but SL already seems to handle that quite well. Just try to reduce the window size and watch how much your fps increases.

No, the only way to even start fixing SL's LoD issues is to ditch the RenderVolumeLODFactor and replace it with a function to set LoD factor for objects individually. Anything else is jsut patching it up with string and chewing gum.

Share this post


Link to post
Share on other sites

I wonder what would happen if they got rid of the auto generated option for the LOD's and just made the LOD only have "from above" and "file" for choices. 

Share this post


Link to post
Share on other sites
5 minutes ago, ChinRey said:

There is a threshold for how low you can go and still save LI and it's actually quite a bit higher than most content creators seem to be aware of.

You're talking from the perspective of a static object creator. I have to face the problem of a character leg that can't be reduced any further without having to fold over to an impostor. Which can be done, if the problem wasn't the deformation of such impostor and how to make it visible from different angles AND the need to add an extra material. A lot of issues arise from this and the worse part is that it would completely defeat Bakes On Mesh (which i never liked, but i can see the benefits if it was worked correctly, including materials and with a fully working alpha masking on all channels, which it doesn't for now)

11 minutes ago, ChinRey said:

That again is what we have today. LoD swap distances are determined by object size and distance and that usually translates into screen space size.

Not really, if you think about it. Not only it breaks artificially enlarged bounding boxes, but it adapts to screen size and can tell more reliably how distant the observing camera actually is from any object. Again, you think from static objects perspective, but this would be applicable to avatars too and the long timer radius/diameter calculation/bug could be fixed without too much trouble.

17 minutes ago, ChinRey said:

No, the only way to even start fixing SL's LoD issues is to ditch the RenderVolumeLODFactor and replace it with a function to set LoD factor for objects individually.

Which would be abused the same way as the current RenderVolumeLODFactor. There's need for something that can't be manipulated and renders the LoD switches effective on everything, from the rock to the avatar.

Share this post


Link to post
Share on other sites
17 minutes ago, Digit Gears said:

I wonder what would happen if they got rid of the auto generated option for the LOD's and just made the LOD only have "from above" and "file" for choices. 

I'd buy a noise suppression headset for the screaming that it would cause 😂

Share this post


Link to post
Share on other sites
1 hour ago, OptimoMaximo said:

This is true, however setting the lowest LoD to be at most a X percentage of the highest LoD should allow a certain degree of flexibility and require more careful attention to the degree of reduction, in my opinion. Perhaps not like animesh items implement it, but still linking the system to the amount of max triangles an items is made of.

Yesterday i went to the Content Creation User Group meeting after a LOOOOOOONG time, and @polysail suggested also a good idea, LoD switch distances to be based off screen space size of the object.

In any case, the current LoD system needs an overhaul, as it is it's ridiculous.

All this contributes to the ridiculousness of the whole current LoD system.

Yes i guessed it, my UDIM method ;)

UDIM?

Share this post


Link to post
Share on other sites
Just now, Kyrah Abattoir said:

No, i was targeting people who might be tempted to keep making things like alpha slices.

Ooopsie... ok thanks for the clarification!

Share this post


Link to post
Share on other sites
24 minutes ago, OptimoMaximo said:

I'd buy a noise suppression headset for the screaming that it would cause 😂

If people want triangles, they'll have to make their own triangles!

But honestly though, I'd like to think it would at least help give a bit more of a stronger and harder to get around incentive if people couldn't just rely on SL's lackluster auto generate anymore to skimp by.

Share this post


Link to post
Share on other sites
Posted (edited)
47 minutes ago, Digit Gears said:

But honestly though, I'd like to think it would at least help give a bit more of a stronger and harder to get around incentive if people couldn't just rely on SL's lackluster auto generate anymore to skimp by.

I agree. Personally I would not dislike the change, since i just published a script to generate LoDs and rig them to the original high lod 😁 but seriously an automatic lod generator in the upload window is something that LL can't afford to not have, in regard to the spirit of SL. They just need a better one and with more solid rules. Not stricter, solid. Something like if the high lod is close to the limit of vertices it could accept, the lower number of triangles the lowest lod has to be, setting thresholds that trigger different requirements of triangle counts. For sure, people has to stop claiming any object to be 1 LI and LL's uploader has to stop claiming ridiculously low triangle counts for the lowest lods under the penalty of skyrocketing LI

Edited by OptimoMaximo

Share this post


Link to post
Share on other sites
1 hour ago, OptimoMaximo said:

You're talking from the perspective of a static object creator.

Yes, I do you see your point there. We may need a different system for wearables. Right now there's no effective way to control that of course.

 

1 hour ago, OptimoMaximo said:

Which would be abused the same way as the current RenderVolumeLODFactor.

Any feature can be abused of course but LoD factor for objects should be far less vulnerable to it than the RenderVolumeLODFactor:

  • It would only affect that particular object that needs to be strengthened, not everything else in the scene.
  • It wouldn't be possible to cheat on land impact with collapsed LoD models anymore since the LI would be based on the actual swap distances rather than the estimated ones.
  • It would be possible to choose LoD levels according to an object's location. A copy of an item in a secluded place could be set to lower LoD factor than a copy of the same item in an exposed position.
  • It would allow the user to decide what LoD level they want their build to be seen at. If they want to skimp on the LI by collapsing LoD or waste LI by over-strengthening, well, it's their choice.
  • It would make it possible to set swap distances relevant for each object. The hardwider swap distances we have today are really only suitable for a very limited number of items, mostly (but not only) dependent on their size. For objects larger than - say about five meters - the swap distances are usually too large and there's no way to reduce them fast enough. For objects smaller than - let's say 0.5 m - the swap distances are so small you end up only seeing the lowest LoD model at most realistic view distances, making the mid and low LoD models totally irrelevant.

Share this post


Link to post
Share on other sites
3 hours ago, ChinRey said:
  • exposed position.
  • It would allow the user to decide what LoD level they want their build to be seen at. If they want to skimp on the LI by collapsing LoD or waste LI by over-strengthening, well, it's their choice

Ok, if that factor influences the LI realtime, then it's a good idea

Share this post


Link to post
Share on other sites
1 hour ago, OptimoMaximo said:

Ok, if that factor influences the LI realtime, then it's a good idea

Yes, that's essential.

And I suppose you now see why I'm not going to attend the Gontent Creator meetings. You've worked with Unity so you are famiiar with the concept of adjustable LoD swap distances. The people at those meetings are not familiar with it, they're not the kind of people who grasps new ideas easily and they're not much into thinking outside the box. And you only have 15 sedonds to sell the idea before somebody changes the topic - an elevator pitch is way too long winded at those meetings.

Even if people did catch on to the idea, it's probably too late because it would require changes to code buried deep in the spaghetti. It should have been implemented back in 2011 with mesh and the new land impact system. It was actually proposed back then but LL decided it wasn't important. That decision is the reason why we have all those LoD issues today.

Share this post


Link to post
Share on other sites

Hello,

Second life is not Unity, Unity has one main goal, despite we can use it for more than game. Second life has infinity of possible use.

Second life is significantly more easy to use than Unity. LoD is important for game, of course, but on Second life it musts stay an option for the users who need good LoD. For this users, I agree a better LoD system is necessary to have better FpS on their sims.

But a lot of peoples on Second life use this platform as a platform as Bryce was used in the past, and so LoD is less important than details and very detailed objects, even avatars and outfits of course. Because they like to reproduce environment with lot of details. An other main point is Second life is for everybody, beginners and more experimented users. I am in favor about better LoD management as an option to help create better objects for sims with game / role play action goals, for many other use, LoD is not a top priority, very detailed objects are more important.

And I forget an other point peoples have yet the right to use their sims or lands, to create and use what they create, they pay for this.

Share this post


Link to post
Share on other sites

Why do people keep bringing up this idea that detail and optimized content are at two opposite ends of a spectrum?

 

37 minutes ago, Motoko Oanomochi said:

And I forget an other point peoples have yet the right to use their sims or lands, to create and use what they create, they pay for this.

Within the constraints of what Linden Lab allows, this is still their system, we just rent it for a while.

  • Like 1

Share this post


Link to post
Share on other sites
31 minutes ago, Motoko Oanomochi said:

Second life is not Unity, Unity has one main goal, despite we can use it for more than game. Second life has infinity of possible use.

I wasn't trying to compare SL to Unity as such, only mentioned that Unity is one of the game engines that use adjustable LoD swap distances.

Beyond that I'm a bit confused about what you mean. I get the overall impression that you want to criticise the idea of adjustable LoD in SL but all your arguments are actully for it.

  • Like 1

Share this post


Link to post
Share on other sites
19 minutes ago, ChinRey said:

I wasn't trying to compare SL to Unity as such, only mentioned that Unity is one of the game engines that use adjustable LoD swap distances.

Beyond that I'm a bit confused about what you mean. I get the overall impression that you want to criticise the idea of adjustable LoD in SL but all your arguments are actully for it.

I said, I am in favor about better LoD management as an option to help create better objects for sims with game / role play action goals.

But as an option, not mandatory. It is why I think deeply the best way is to create a new type of mesh, let say mesh V2 or arctan mesh, as you wish with LoD management as we already have for animesh.

 

Share this post


Link to post
Share on other sites
29 minutes ago, Kyrah Abattoir said:

Why do people keep bringing up this idea that detail and optimized content are at two opposite ends of a spectrum?

 

Within the constraints of what Linden Lab allows, this is still their system, we just rent it for a while.

We can see this from a lot of point of views, the fact is with the mandatory respect for TOS, a sim owner is renting a server with a large freedom to use it for a lot of purposes. During the rent, the owner is a customer who pays and so deserves the service. As it is the case actually, and really Second life is working well.

Details and optimization can go together of course, but a designer is not always an engineer or an uvmap specialist or a LoD specialist. And for some use of Second life, as make art or nice pictures , LoD is not a top priority. A lot of peoples use Second life as Bryce was used, as a "photorealistic" render engine for beginner, because Second life is versatile enough. I am not saying Second life can be compared with Bryce directly or photorealistic render engine as V-Ray, I am saying lot of users are using Second life instead of using photorealistic render engine, and for them FpS can go at 8 or even less, if the camera moves well and they see all details they want, all is good. They have this right.

 

Share this post


Link to post
Share on other sites
59 minutes ago, Motoko Oanomochi said:

I said, I am in favor about better LoD management as an option to help create better objects for sims with game / role play action goals.

But as an option, not mandatory.

And that's exactly my point - or at least one of my points. With adjustable LoD it's you as the enduser who ultimately decide how much resources you want to spend on improved LoD.

  • Like 1

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...