Why beginners don't learn the basics first?

[ChinRey]

22 hours ago, NiranV Dean said:

Textures are relatively cheap to render once they are downloaded and decoded and pushed into GPU memory. It's hard to measure how much exactly...

There are three very simple tests anybody can do and it only takes a few minutes.

1. Find a place with not too much content - prefeably high up in the air away from everything else. Rez a prim, make it big enough to fill your entire screen, texture it with a 512 and check you fps (wait a few second for it to stabilize). Then retexture the prim with a 1024 and see how much your fps drops.
2. Fill you view with multiple prims, all with the same texture. Then give each prim its own unique texture, all with the same resolution, and watch how much your fps drops.
3. Rez an object with .. let's say about 1000 triangles. make a background big enough to fill up your entire screen and texture the background and object with the same texture. Check fps, delete object and watch how much fps changes.

These tests don't give an exact answer of course but they do give an idea what the practical implications of more and larger textures and more geometry really are and I think most people will be surprised.

I once did the test with a friend. I have a fairly modest home computer with a mid range AMD gpu, he is an architect student so he has a powerhouse graphics computer with a NVIDIA 1040. I don't remember the exact numbers but my fps dropped by something between 0.2 and 0.5, hi by 3-5.

 

22 hours ago, NiranV Dean said:

Triangles ARE cheap but don't be fooled. This is THE number one excuse for people spamming them "triangles weight nothing".

The ultimate answer to optimization is of course to save on everything. Use every triangle and every pixel for all they are worth. There are tons of tricks to reduce both textures and geometry without affecting the visual end result negatively. But you won't find many of them in YT videos or elsewhere on the internet, although some of them are mentioned on this forum and I've also found a few gold nuggets at the polycount forum. It takes a bit of thinking, both inside and outside the box.

---

Awww, I can't resist it. I did a fourth test for this post: Go to a really laggy area, check fps, derender objects and see how it changes. It's not nice of me since the obvious location/target was of course Belleseria and those trees:

This is a brand new region, it doesn't even have a name yet, just an SSP number, so no user added content. My fps with the trees was 25.1, without the trees  29.1. That's less of a difference than I expected (probably because I didn't eliminate that many textures from the scene) but still quite significant.

I feel a bit bad about posting this example because even though I think LDPW could do with a few skilled mesh makers, it is probably also a question of budget control. Belleseria had to be made as cheaply as possible or not at all. Even I would have needed a few hours to make an optimized tree of this style and visual quality, a Mole or another lesser skilled mesh maker might well need a day or more. So, launch Tree[d], spend a minute or five tweaking the settings, export and import to SL. Don't bother with LoD or physics models - use LoD above for mid and low, just buthcer lowest LoD away, set tree to phantom and add a physics prim. It saves hours of work and time is money.

Edited February 6, 2020 by ChinRey

[Drayke Newall]

2 hours ago, ChinRey said:

I feel a bit bad about posting this example because even though I think LDPW could do with a few skilled mesh makers, it is probably also a question of budget control. Belleseria had to be made as cheaply as possible or not at all. Even I would have needed a few hours to make an optimized tree of this style and visual quality, a Mole or another lesser skilled mesh maker might well need a day or more. So, launch Tree[d], spend a minute or five tweaking the settings, export and import to SL. Don't bother with LoD or physics models - use LoD above for mid and low, just buthcer lowest LoD away, set tree to phantom and add a physics prim. It saves hours of work and time is money.

The problem is that seeing as this is a LL created area, they should be the ones setting the example mole or not. For a simple tree that is repeated over and over, the time spent creating it correctly, even if it took a little longer, would be far more beneficial in the long run. Add to this the potential of adding the tree to the LL inventory Library for everyone to use and its even better as it provides an example of proper creation and its benefits.

EDIT:

I actually have just then been to Belleseria for the first time and was to be honest surprised at the lack of optimisation in the area. Trees and bushes as you mention using outdated physic tricks using prims (prims have worse optimisation than a mesh object), buildings that use a combination of mesh and prims where prim faces not seen having 512 textures still on them, mesh buildings and rocks that have faces/shape that could have easily used normal maps to achieve the same result, grass that in a small patch has more triangles and vram usage on it than mesh grass 4x the size can achieve, lamp posts that render cost is just to big, bridges that use improper smoothing systems making them look like a cut octagon, prims used on bridges instead of mesh, textures that could easily be combined into one texture, list goes on.

I get that they are moles and not 'professional' but LL really need to step up their game.

Edited February 6, 2020 by Drayke Newall

[Kyrah Abattoir]

I thought I"d repost an excellent article from Robert Yang https://www.blog.radiator.debacle.us/2012/04/on-joiner-detail-and-greeble.html

And a quote from this article that I believe applies to SL more than ever:

Quote

They model environments and characters as these purely academic exercises detached from how games actually function, framing and composing their environments as flat 2D renders, sculpting intricate nooks into individual bricks: but players will never inspect these bricks, nor will they care about the specular on that trash can model, and shaving 100 polys off that one-use statue prop will almost certainly not impact anything.

Yes, it is understandable -- many are making portfolio pieces and want to demonstrate technique and craft. But as a game maker who makes assets solely to fulfill certain roles in games, part of me also finds it a bit wasteful and perverse, that many of them are making game art that never actually goes into a game, nor fits the real constraints of making a game.

 

[OptimoMaximo]

46 minutes ago, Kyrah Abattoir said:

I thought I"d repost an excellent article from Robert Yang https://www.blog.radiator.debacle.us/2012/04/on-joiner-detail-and-greeble.html

And a quote from this article that I believe applies to SL more than ever:

 

Unfortunately, in SL the average user DOES inspect textures at insane closeup distances, so they will notice the single brick in a wall. More than once, I had encounters in a sandbox, this guy smashed his avatar nose to the wall (not knowing how to cam any better) and pointed out that what I was using was "not a photo texture". Not to mention those complaining about the lack of individual buttons and stuff on garments "if you update your item with actual detail instead of painted on [insert type of miniscule object detail here] I will change my review on the demo(!) and maybe actually buy the item" 

 

So while I fully agree with that article's content and message, by personal experience I find it inapplicable to SL. 

Edited February 8, 2020 by OptimoMaximo

[animats]

6 hours ago, OptimoMaximo said:

Not to mention those complaining about the lack of individual buttons and stuff on garments "if you update your item with actual detail instead of painted on [insert type of miniscule object detail here] I will change my review on the demo(!) and maybe actually buy the item" 

Shoes. SL mesh shoes have insane triangle counts. 20,000 tris per shoe is not unusual. I've seen ones where every eyelet is a full 3D model.

About 2600 triangles for a pair of shoes. Do you need more detail than that?

Made by Duck Girl. A custom job done for my animesh characters. 4 LI with animesh accounting.

What we really need, though, is a level of detail system where the highest level of detail is not loaded unless you get very, very close.

 

[OptimoMaximo]

2 hours ago, animats said:

Shoes. SL mesh shoes have insane triangle counts. 20,000 tris per shoe is not unusual. I've seen ones where every eyelet is a full 3D model.

About 2600 triangles for a pair of shoes. Do you need more detail than that?

Made by Duck Girl. A custom job done for my animesh characters. 4 LI with animesh accounting.

What we really need, though, is a level of detail system where the highest level of detail is not loaded unless you get very, very close.

 

Dude, I agree on this. Try to sell those shoes as regular avatar wearable and see what comments you'd get, how many tons of demos (because from the picture is good, BUT at close... VERY close inspection...) and how few actual sales 

[animats]

 

7 minutes ago, OptimoMaximo said:

BUT at close... VERY close inspection...

Is that close enough?

The stitching detail could be improved with a normal layer.

That's what tooling should do. Take an ultra-high detail version and hammer it down into something like this, with fine detail represented as normals.

[OptimoMaximo]

5 minutes ago, animats said:

 

Is that close enough?

The stitching detail could be improved with a normal layer.

That's what tooling should do. Take an ultra-high detail version and hammer it down into something like this, with fine detail represented as normals.

That's what I do, but if the strings are just "painted on" people complain, they want the 3d eyelet for those strings that have to be a long continuous tube, and I've also read and seen people compliment someone because they had modeled the stitchings too, calling that "highest quality". And no, that's not close enough, I've got screenshot of closeups on details much closer to the surface, at the limit of the clipping plane, to point out a pixel that ruined the seamless look from a UV shell to its neighboring one. At my question "can you notice it at normal distance?" the answer was "no but its not lifelike and it ruins the immersion when I zoom up close"... 

Again, I do agree with you and with the article contents posted above. I'm just pointing out what is perceived as "quality" in secondlife from what I gathered along my years of creation and customers. The things that sell 75-80%of the times after a demo is delivered are those where I deliberately modeled useless geometry that a texture would have sufficed for. Where I went for a texture, 5-10% of the times demos turn into an actual sale. 

[ChinRey]

3 hours ago, animats said:

What we really need, though, is a level of detail system where the highest level of detail is not loaded unless you get very, very close.

In theory we do have that. The swap distance between high and mid is supposed to be less than half a meter for an object as small as a shoe. That means you should be able to load as many triangles as you like onto the high model since only people with a serious shoe or foot fetish is ever going to see it. But then there's the LoD factor inflation and, for worn objects (not sure if that applies to animesh), the fitmes LoD bugs.

[Kyrah Abattoir]

On 2/8/2020 at 11:04 PM, OptimoMaximo said:

That's what I do, but if the strings are just "painted on" people complain, they want the 3d eyelet for those strings that have to be a long continuous tube, and I've also read and seen people compliment someone because they had modeled the stitchings too, calling that "highest quality". And no, that's not close enough, I've got screenshot of closeups on details much closer to the surface, at the limit of the clipping plane, to point out a pixel that ruined the seamless look from a UV shell to its neighboring one. At my question "can you notice it at normal distance?" the answer was "no but its not lifelike and it ruins the immersion when I zoom up close"... 

Again, I do agree with you and with the article contents posted above. I'm just pointing out what is perceived as "quality" in secondlife from what I gathered along my years of creation and customers. The things that sell 75-80%of the times after a demo is delivered are those where I deliberately modeled useless geometry that a texture would have sufficed for. Where I went for a texture, 5-10% of the times demos turn into an actual sale. 

I dunno i've never had this kind of complaint from people in my store and i've been in the business just as long as you have.

Over the years i've been pushing my texture usage down with each product.

[OptimoMaximo]

58 minutes ago, Kyrah Abattoir said:

Over the years i've been pushing my texture usage down with each product.

I've been pushing it up in my case, because it was never high res enough eve if I pack UVs extremely close to zero pixel waste. 

1 hour ago, Kyrah Abattoir said:

and i've been in the business just as long as you have.

Nope, you've been in the business for longer than me. It's years for me now, but before I started, I was only teaching. 

[Kyrah Abattoir]

So does no one teaches anymore how to model and texture without going through baking textures?

 

I can't wrap my head around what goes through some creator's heads when they sell you road pieces, but rather than recycling their road texture across all the parts, or hell, providing it in the package alongside the mesh, ... they bake it into a blurry paste and aparently believe that this is just fine?

[Da5id Weatherwax]

2 hours ago, Kyrah Abattoir said:

So does no one teaches anymore how to model and texture without going through baking textures?

 

I can't wrap my head around what goes through some creator's heads when they sell you road pieces, but rather than recycling their road texture across all the parts, or hell, providing it in the package alongside the mesh, ... they bake it into a blurry paste and aparently believe that this is just fine?

The really cray-cray part, Kyrah, is that almost nobody seems to get that there is a prim-era trick that can be made to work on mesh too - a  "face"  (material) of the model that hasn't been unwrapped at all - ie still has every quad on the model UV mapped to the entire texture (that's how blender sets them when creating a new UV) - will usually take a small tileable texture in planar mapping and have the 'repeats per meter" in the SL edit window work....

Seems to me that something like roads would be an ideal time to see if this works with a quick upload to the beta grid...

 

(ETA: IT's real easy to break if your model isnt clean quads, because the default UV mapping of tris is less predictable as to which corner of the texture the vertices are on but if your road surfaces aren;t clean quads, you're probably doing a lot more wrong than the texturing!)

Edited February 28, 2020 by Da5id Weatherwax

[OptimoMaximo]

8 hours ago, Kyrah Abattoir said:

So does no one teaches anymore how to model and texture without going through baking textures?

 

I can't wrap my head around what goes through some creator's heads when they sell you road pieces, but rather than recycling their road texture across all the parts, or hell, providing it in the package alongside the mesh, ... they bake it into a blurry paste and aparently believe that this is just fine?

It's a matter of perception: baked lighting = more realistic = better. In a sales driven mind, knowing this brings to the obvious conclusion that it's just faster to bake individual texture rather than make sure the various models adapt to a single one texture, if that shines enough to the buyer they don't care of its more resource intensive. Why is it faster? See

6 hours ago, Da5id Weatherwax said:

ETA: IT's real easy to break if your model isnt clean quads, because the default UV mapping of tris is less predictable as to which corner of the texture the vertices are on but if your road surfaces aren;t clean quads, you're probably doing a lot more wrong than the texturing!)

David pinpointed it. 

[Bitsy Buccaneer]

45 minutes ago, OptimoMaximo said:

It's a matter of perception: baked lighting = more realistic = better. In a sales driven mind

may i insert a wee rant about how frickin boring a lot of today's render baked textures are, especially in clothes. shadows and not much else. it's like half the grid's never seen fabric before. grrrr.

[ChinRey]

8 hours ago, Da5id Weatherwax said:

The really cray-cray part, Kyrah, is that almost nobody seems to get that there is a prim-era trick that can be made to work on mesh too - a  "face"  (material) of the model that hasn't been unwrapped at all - ie still has every quad on the model UV mapped to the entire texture (that's how blender sets them when creating a new UV) - will usually take a small tileable texture in planar mapping and have the 'repeats per meter" in the SL edit window work....

Seems to me that something like roads would be an ideal time to see if this works with a quick upload to the beta grid...

Lots'a mesh roads, streets and paths, all using tiled textures and default mapping, no need to go planar.

[Da5id Weatherwax]

1 hour ago, Bitsy Buccaneer said:

may i insert a wee rant about how frickin boring a lot of today's render baked textures are, especially in clothes. shadows and not much else. it's like half the grid's never seen fabric before. grrrr.

be fair - getting fabrics right in SL rendering, particularly since you have to account for folks not having advanced lighting on and therefore not seeing materials to their full effect actually takes a modicum of sneakiness. You wouldn't want the poor dears to actually have to think, would you?

 

But it IS worth it, though - judicious normal mapping and tinting of the specular map is, as far as I know, the ONLY way you'll make a dupioni look anywhere near right in SL

[Kyrah Abattoir]

3 minutes ago, Da5id Weatherwax said:

particularly since you have to account for folks not having advanced lighting on and therefore not seeing materials to their full effect

AParently this will soon be a thing of the past, I believe LL is gutting out the legacy renderer.

[Da5id Weatherwax]

1 minute ago, Kyrah Abattoir said:

AParently this will soon be a thing of the past, I believe LL is gutting out the legacy renderer.

Heaven speed the day!

[Bitsy Buccaneer]

14 minutes ago, Da5id Weatherwax said:

be fair - getting fabrics right in SL rendering, particularly since you have to account for folks not having advanced lighting on and therefore not seeing materials to their full effect actually takes a modicum of sneakiness. You wouldn't want the poor dears to actually have to think, would you?

 

But it IS worth it, though - judicious normal mapping and tinting of the specular map is, as far as I know, the ONLY way you'll make a dupioni look anywhere near right in SL

I'm seeing more and more where they're not even trying to get a hint of weave or knit in there. Not everything is best served by looking like latex.

Tongue in cheekiness aside , I think there might be a bit of over-reliance on materials to save the day, more tech and less artistry. Maybe it's just what I've been happening into recently, but it feels like the bell curve has shifted position in the last year or so especially. It's been a long time since I've run into so many should-be-cottons that are just flat colours. Highlights and shadows aren't enough to mask that.

I do want to see your dupioni though

 

[Da5id Weatherwax]

2 hours ago, Bitsy Buccaneer said:

I do want to see your dupioni though

Quick test with an indigo/crimson warp and weft on the beta grid. Slubs normal mapped to raise them slightly, darkened slightly in the diffuse map and blackened completely in the specular map. Base color of the diffuse map is the deep indigo, with the crimson tint on the specular map. Glossiness 51, Environment 0. Official viewer set to ultra graphics. The object is just a test piece created by using blender's cloth sim to drape a circle over a small sphere

 

Edited February 29, 2020 by Da5id Weatherwax

[Bitsy Buccaneer]

5 hours ago, Da5id Weatherwax said:

Quick test with an indigo/crimson warp and weft on the beta grid.

oooo you've done well with the iridescence . that's very cool

[Da5id Weatherwax]

3 hours ago, Bitsy Buccaneer said:

oooo you've done well with the iridescence . that's very cool

Something I have believed for a long time is that the capabilities of SL rendering with full use of materials have been seriously unappreciated on the "mainstream market" - Modellers use normal mapping for detail but so many people are still baking shadows, when at most they would need a tiny amount of AO, they are still baking stuff into their diffuse map that the render pipeline can handle if you let it and it can easily make the final result look worse. Sometimes what you can achieve with just the right normal and specular maps, including the capabilities built in to utilizing their alpha channels, while leaving the diffuse map as a flat color is surprisingly effective.

I'm sure a major driver for this has been LL's reluctance to finally nuke the legacy renderer - for as long as it continues to exist there will be people running a low enough graphic setting to use it, and to cater to those people I fully understand that creators have to keep the "baked everything" in the diffuse map... but then you run head-first into the disconnect where they say "turn up your LOD factor" but don't say "turn on advanced lighting" when the latter is less taxing on the GPU than the former!

If LL are going to finally bite the bullet and dispose of the legacy renderer that will be huge - folks will have to light their places properly, which in turn will mean that things made for "realistic" lighting will be rendered how they should be and when the incredible creativity of SL-folks really gets focused there I'm sure there are dozens, if not hundreds, of people out there who will push its limits in ways I've never even imagined.

As for the fabric test.. I have to hold my hand up and say that it was a real quick-and-dirty recreation of an experiment I performed over a year ago and there are a lot of flaws in it. If I'd been being serious I'd have cleaned up the artefacts in the photo I sourced the fabric structure from much more aggressively, left in they are far too noticeable when making the slubs tileable. The normal map needs to be stronger, the slubs don't "pop" from the fabric surface as much as they should and the texture of the underlying weave is not as strong as it should be. The iridescence of something like a dupioni fabric, where the bicolor effect is entirely due to using two colors of thread in the weave is much easier to create than a true iridescence or opalescence - for this the specular map is just a higher contrast version of the diffuse map tinted a different color. True opalescence or iridescence is something I've only had limited success in simulating with SL materials but I'm a horrible tinkerer and each time I come back to it I get closer - I might even achieve a result I'm prepared to let somebody else see one day

[Bitsy Buccaneer]

2 hours ago, Da5id Weatherwax said:

 Sometimes what you can achieve with just the right normal and specular maps, including the capabilities built in to utilizing their alpha channels, while leaving the diffuse map as a flat color is surprisingly effective.

Can it convey knit or woven fabrics? I've not seen either done effectively with a flat diffuse.

[Da5id Weatherwax]

25 minutes ago, Bitsy Buccaneer said:

Can it convey knit or woven fabrics? I've not seen either done effectively with a flat diffuse.

Not knitted - that really needs the AO on the yarn in addition to the normal map. Depending on the weave a woven fabric is possible with a flat diffuse and a normal map alone but the coarser the weave or the more "featured" it is - such as overshot weaves, for example - the more you get to needing the AO for its structure baked in. A tightly woven fabric made of medium to fine yarn, though, normal mapping is enough - you see "just enough" of it at normal viewing distances to know it's there but that's all that is required (you may need to make the normal mapping a tad stronger than you would if the fabric texture was baked into the diffuse). The effect breaks down if you've got the fabric right up against your camera's near clipping plane but you would design a texture/material intended to be viewed from that close differently.