PBR is coming...

[PekeNL]

35 minutes ago, ChinRey said:

We'd end up with SL scenes containing tens of thousands of 4k textrues and surface maps, more than enough to make even the strongest game computer in the world to crash and burn.

12 minutes ago, Wulfie Reanimator said:

Or more like we'll get one really nice looking object, with the rest of the world textured in glorious 128px downscaled textures because the viewer refuses to keep more in memory.

Unfortunately, this is a much broader issue as texture optimization is a practice not a lot of creators are good at.... You don't need a flippin' 1024x texture on a headset...

I'm excited for this change. I'm very interested in how they're going to incorporate it on worn items. For items placed in world it's a simple drag and drop, but often wearable don't come with the proper permissions to allow that.

Maybe it'll force mod rights on things or is that wishful thinking?

[Wulfie Reanimator]

17 minutes ago, PekeNL said:

Unfortunately, this is a much broader issue as texture optimization is a practice not a lot of creators are good at.... You don't need a flippin' 1024x texture on a headset...

I'm excited for this change. I'm very interested in how they're going to incorporate it on worn items. For items placed in world it's a simple drag and drop, but often wearable don't come with the proper permissions to allow that.

Maybe it'll force mod rights on things or is that wishful thinking?

It hasn't been possible to drag textures (or use the texture picker) on worn attachments for a loooong time, even if you have modify permissions.

Edited January 13, 2023 by Wulfie Reanimator

[Quistess Alpha]

7 hours ago, Wulfie Reanimator said:

(or use the texture picker)

I change the texture on worn attachments all the time?

[Wulfie Reanimator]

1 hour ago, Quistess Alpha said:

I change the texture on worn attachments all the time?

🤷 https://imgur.com/gipxbAc - Most textures are not copy+transfer (for very good reasons) so you end up with this. Dragging a (no transfer) texture onto a modifiable attachment fails to apply it, unless you rez the target object on the ground first. (Dragging the texture onto the slot in the build window also copies the texture into the object's contents for no reason.)

Edited January 13, 2023 by Wulfie Reanimator

[Quistess Alpha]

8 minutes ago, Wulfie Reanimator said:

🤷 https://imgur.com/gipxbAc

Uhm, Why are you trying to select a texture from your inventory window instead of the texture picker window? I just tested it and drag-drop from inventory onto the texture image in the build window works (Kokua viewer) though.

ETA: Oh, I see. The problem is that the textures themselves have restrictive perms. You can add full-perm textures to worn attachments, but for no-mod textures you need to rez the attachment in-world.

Edited January 13, 2023 by Quistess Alpha

[Wulfie Reanimator]

Just now, Quistess Alpha said:

Uhm, Why are you trying to select a texture from your inventory window instead of the texture picker window? I just tested it and drag-drop from inventory onto the texture image in the build window works (Kokua viewer) though.

I know it's a bit confusing to watch since my cursor wasn't visible, I click on the tail texture in the texture picker list at 0:13. It displays it in the texture slot (but not on the attachment) and complains about no-copy/no-transfer. Then I close the picker list and also try to drag textures from my inventory.

That's on Firestorm, and it's been like that for years.

[Quistess Alpha]

11 minutes ago, Wulfie Reanimator said:

Most textures are not copy+transfer (for very good reasons)

my (limited) experience has been that for full-perm templates, the example textures are also full-perms. I guess I don't have much experience with 3rd-party modifications distributed as raw textures though.

[arisaie]

10 hours ago, Wulfie Reanimator said:

Or more like we'll get one really nice looking object, with the rest of the world textured in glorious 128px downscaled textures because the viewer refuses to keep more in memory.

This is the next thing I wanted to ask...does the SL viewer downscale textures with camera distance? 

[Wulfie Reanimator]

13 minutes ago, Quistess Alpha said:

my (limited) experience has been that for full-perm templates, the example textures are also full-perms. I guess I don't have much experience with 3rd-party modifications distributed as raw textures though.

The furry community is pretty mod-friendly. You have lots of different creators creating different avatar components, and then there are creators that create a "mod pack" which is a collection of textures for all of the different components from different creators.

Most furry avatar components are modifiable, so they often don't come with applier scripts, so we end up with loose textures in mods, so we can apply those ourselves. 🙂

5 minutes ago, arisaie said:

This is the next thing I wanted to ask...does the SL viewer downscale textures with camera distance? 

Maybe by distance (doesn't seem like it, I get max resolution textures on objects across the sim), and not by screen space. The smallest object on your screen can have a fully loaded 1024px texture.

Edited January 13, 2023 by Wulfie Reanimator

[arton Rotaru]

1 hour ago, arisaie said:

This is the next thing I wanted to ask...does the SL viewer downscale textures with camera distance? 

The GLTF viewer does. It has an all new texture streaming pipeline. The viewer downloads the lowest resolution needed to display a 1:1 ratio between texels and pixels. It keeps textures in memory until the viewer runs out of memory. So in theory, even 4k textures will be possible in the future. But not with this initial PBR release.

In the log (Ctrl+Shift+4) you can see something like this:

Budget: 5212
CR: 0
CU: 868
EU: 720 (-0.17)
TU: 3164
AM: 2296

Budget is the VRAM budget.
CU is currently used VRAM.
EU is estimated usage.
TU is target usage.
AM is available memory.

If AM runs near zero, the viever will start to remove textures from VRAM. The numbers shown are with a 6 GB VRAM GPU.

Edited January 13, 2023 by arton Rotaru

[animats]

1 hour ago, arton Rotaru said:

The GLTF viewer does. It has an all new texture streaming pipeline. The viewer downloads the lowest resolution needed to display a 1:1 ratio between texels and pixels. It keeps textures in memory until the viewer runs out of memory.

Yes.

I'm amused that, after I did this in my own experimental viewer, it suddenly became possible for LL to do it. It looks like all the textures are at high resolution, but they're not. Multiple threads are frantically loading and unloading textures to try to maintain a 1:1 ratio between texels and pixels. If you move around slowly, the illusion is perfect.

Fast motion is harder. Here's my stress test for my own experimental viewer.  It's a speed run through Port Babbage. I've posted this before. If you watch closely, you'll see moments when a texture is blurry because the loader fell behind. In a second or two it caches up. The fast tour through the Savory Street shops shows this. For that to work, the texture loading order has to change on the fly as the viewpoint moves. Otherwise, too much time is going into loading textures for stuff you were near, but are not near to now.

Here's a early test of mine from a cold cache. This is a shopping area full of high-resolution images. This is more sluggish and has some bugs. You'll see the textures coming in. Get close to something and it shows up quickly. This resolves the SL shopping dilemma - is it worth waiting for this store to load?

Key concept: it doesn't matter how much resolution the image has, because it will never be loaded at full resolution until you get really close.

(Avatar clothing, though... There are avatars with more texture surface area than a house. All that has to be loaded. But there are tricks not used yet, and that problem can be overcome.)

[Yseult UwU]

this is super exciting. will we be able to wear environment probes for metallic clothing?

[Jenna Huntsman]

8 hours ago, Yseult UwU said:

wear environment probes

No - this is specifically blocked, and besides, isn't needed.

 

8 hours ago, Yseult UwU said:

metallic clothing

This will work; as your clothing will use the nearest probe volume for reflections (Usually, that means a probe in the same room)

[Fluffy Sharkfin]

2 hours ago, Jenna Huntsman said:

No - this is specifically blocked, and besides, isn't needed.

Outrageous!  LL should double the attachment points so that we can wear a probe for each attachment!  What if I were standing with one foot in one room and the other foot in another room while holding my hands out of a window and sticking my head in the refrigerator, if I can't have a probe for every shiny thing I'm wearing it's going to ruin "muh immershunz"! 😭

[OptimoMaximo]

On 1/13/2023 at 6:53 PM, animats said:

I'm amused that, after I did this in my own experimental viewer, it suddenly became possible for LL to do it. It looks like all the textures are at high resolution, but they're not. Multiple threads are frantically loading and unloading textures to try to maintain a 1:1 ratio between texels and pixels. If you move around slowly, the illusion is perfect

LL needs to be proved wrong before they admit, with their actions, that what they claimed as not possible is indeed possible, just putting the necessary work (and will to work... )

[Love Zhaoying]

On 1/13/2023 at 1:53 PM, animats said:

I'm amused that, after I did this in my own experimental viewer, it suddenly became possible for LL to do it. 

My guru often talked about the concept that once a thing is done (either science or otherwise), it becomes more possible for the thing to be done again.  Anyway, the scientific community initially was interested in the idea (especially in more open-minded countries like India), but it fizzled out in the 70's. Rupert Sheldrake.

[animats]

9 hours ago, OptimoMaximo said:

LL needs to be proved wrong before they admit, with their actions, that what they claimed as not possible is indeed possible, just putting the necessary work (and will to work... )

I've now made the point I was stating a few years back: Second Life can look as good and run as well as an AAA title video game. Look back in the forums a few years, and you'll see many people saying that's impossible with all the un-optimized user-created content. I said that it was quite possible with a specific set of improvements to the technology. Today's' viewers do some of that, and SL looks and runs much better. On good content, it looks better than many video games.

It just takes work.

We don't yet have the smooth, seamless experience. Working on it.

[Yseult UwU]

On 1/14/2023 at 7:16 AM, Fluffy Sharkfin said:

Outrageous!  LL should double the attachment points so that we can wear a probe for each attachment!  What if I were standing with one foot in one room and the other foot in another room while holding my hands out of a window and sticking my head in the refrigerator, if I can't have a probe for every shiny thing I'm wearing it's going to ruin "muh immershunz"! 😭

oh this is definitely not something i would like. i was hoping a max of 1 only attachable to avatar center.

Edit:

and it would only update when standing still

Edit 2:

oh and if it detects it's within 10m of another probe, it'll refuse to update (unless forced to (which would be a clientside thingy))

Edited January 15, 2023 by Yseult UwU
even more perspective

[Fluffy Sharkfin]

24 minutes ago, Yseult UwU said:

oh this is definitely not something i would like. i was hoping a max of 1 only attachable to avatar center.

Edit:

and it would only update when standing still

Edit 2:

oh and if it detects it's within 10m of another probe, it'll refuse to update (unless forced to (which would be a clientside thingy))

Yeah... I was kidding!

Reflection probes will be automatically placed in the region by default so essentially your avatar should always be in range of a reflection probe which is why, as Jenna pointed out above, wearable reflection probes aren't needed.

[Yseult UwU]

21 minutes ago, Fluffy Sharkfin said:

Yeah... I was kidding!

Reflection probes will be automatically placed in the region by default so essentially your avatar should always be in range of a reflection probe which is why, as Jenna pointed out above, wearable reflection probes aren't needed.

oh i see! automatic placement helps a ton. I wonder what the parameters for placement are, can't imagine them being placed every 10 cubic meters lol

[Yseult UwU]

would someone mind explaining what an AO map does as opposed to AO the post-process effect? the best i can make out of it is that it simply darkens the diffuse texture.

[Wulfie Reanimator]

20 minutes ago, Yseult UwU said:

would someone mind explaining what an AO map does as opposed to AO the post-process effect? the best i can make out of it is that it simply darkens the diffuse texture.

It's a similar effect, but the ambient occlusion map is a kind of "local, static ambient occlusion" for only that object. If the object has no gaps, scratches, or any concave surfaces, the AO map should be blank. (Thing to look into: "cavity map")

The other (screen space) ambient occlusion is a dynamic effect, in the sense that separate objects can interact with each other. They'll get that darkening effect when they get close together, which cannot be done with AO maps.

Edited February 2, 2023 by Wulfie Reanimator

[OptimoMaximo]

17 hours ago, Yseult UwU said:

would someone mind explaining what an AO map does as opposed to AO the post-process effect? the best i can make out of it is that it simply darkens the diffuse texture.

In a pbr environment, the AO maps mainly influence the strength of the specular reflection coming off the surface in the darker areas, because it is basically receiving less light to begin with. Then it also darkens the diffuse color.

[Wulfie Reanimator]

26 minutes ago, OptimoMaximo said:

In a pbr environment, the AO maps mainly influence the strength of the specular reflection coming off the surface in the darker areas, because it is basically receiving less light to begin with. Then it also darkens the diffuse color.

Do you happen to know if that also means the AO map can be (completely) masked away by intense lighting in SL's implementation? I haven't used the PBR viewer at all myself, but that is one big difference between the AO map vs AO baked into the diffuse(/albedo) layer.

Edited February 2, 2023 by Wulfie Reanimator

[arton Rotaru]

1 hour ago, Wulfie Reanimator said:

Do you happen to know if that also means the AO map can be (completely) masked away by intense lighting in SL's implementation? I haven't used the PBR viewer at all myself, but that is one big difference between the AO map vs AO baked into the diffuse(/albedo) layer.

Yes. The AO map won't show up where direct light hits.

Edited February 2, 2023 by arton Rotaru