Sculpty LOD problem - Need Help!

[cutegl08]

I bought some sculpties from market place for my build but they have terrible LOD issues  (even with RenderVolumeLODFactor set to 4). It loses shape completely when I zoom out a little.  The sculpt maps are full perm (copy/mod/trans) Is it possible for me to export these sculpt maps and fix the LOD problem? I am not at ll familiar with Blender. However I will learn blender if there is an easy way to fix this problem. Thank You!

[Rhys Goode]

If they are full perm sculpt maps, then yes, its possible to read them into blender, and edit the resulting 3D model.  There is nothing easy or straightforward about Blender, so be forewarned.

But you won't be able to do much about the LOD shifts, they are built into the system for sculpties.  I'm pretty sure there is nothing you can do to change them. (that is, there is no simple setting you can check)

But some sculpts are much better behaved than others in this regard, because there are tricks you can use when you are building the model. Sculpts from more skilled modelers can be dramatically better than sculpts from beginners.  

Oh, and note that RenderVolumeLODfactor only affects how *you* see the world, it does nothing for anyone else's client.

[Jasmin Helstein]

What might help is clearing your cache. It does happen sometimes that sculpts look weird, maybe not all of them, only a few. But after clearing the cache they all look normal again. It's worth a try ...

[Geldsbard Freck]

Understand what a "sculpty" is. The sculpty image is actually called a "height map" in 3D modeling terms; the RGB values for each pixel are actually the X,Y,Z coordinates for each vertex in the wire mesh for the object relative to the objects center.

When uploading any image to Second Life, Sl turns the image into a TGA ("Targa" ) image. This includes Scuplty maps.  (If you are unsure this is the case, download ("save as") any image in your inventory and it will save as a TGA image file).

TGA, like PNG, JPG, and many others have compression schemes that help to reduce the final file size. Sometimes these compression schemes will, among other things, merge two pixels by finding a visual medium - such as a blue pixel and a red pixel, side-by-side can be merged into a single purple pixel. To our eyes, this is visually acceptable, but when each pixel in the sculpty map is actually representative of the numerical position of a point in the mesh, this can begin to distort the object or even delete points.

When DirectX or OpenGL (the graphics drivers in our computers that actually draw Second Life on our screens) performs a LOD function (Level Of Detail) on an object mesh, it "culls" (removes) vertex at specific intervals - sometimes every other, sometimes every third, sometimes every forth vertex, according to your graphics settings. 

If the height map (Sculpty map) has been compressed in the upload to SL, thereby distorting the mesh to begin with, the LOD function has a much more difficult time deciding which vertex to remove and can (and many times does) removes the wrong one creating an even more heavily distorted object.

Many casual (?careless?) builders either do not know or don't care to check the "losses compression" box in the image upload dialog when they upload their maps to SL. You may not be able to detect any "smudging" with your eyes but an analisis of the numerical data would show missing vertex or out of place vertex as a result of even slight compression of the image.

If this is the case with your purchased object, unfortunately, there isn't much you can do about a disappointing LOD reduction.

If it is your own creation, simply re-upload the map and be sure "use lossless compression" is checked in the upload dialog.

 

[Chosen Few]

---

Geldsbard Freck wrote:

The sculpty image is actually called a "height map" in 3D modeling terms; the RGB values for each pixel are actually the X,Y,Z coordinates for each vertex in the wire mesh for the object relative to the objects center.

---

Technically, it's not a height map, but rather a three dimensional vertex displacement map.  If it were indeed just a height map, then by definition, it would only displace along one axis (the height axis), rather than along all three.

 

---

Geldsbard Freck wrote:

When uploading any image to Second Life, Sl turns the image into a TGA ("Targa" ) image. This includes Scuplty maps.  (

If you are unsure this is the case, download ("save as") any image in your inventory and it will save as a TGA image file

).

---

This is not accurate.  SL converts images to JPEG2000 format, not TGA format, upon upload.  TGA is way too big for quick transfer across networks.

The fact that the viewer can read and write TGA files for import and export doesn't mean that's what the system is using internally.  Inside, it's all JPEG2000.

 

---

Geldsbard Freck wrote:

TGA, like PNG, JPG, and many others have compression schemes that help to reduce the final file size. 

Sometimes these compression schemes will, among other things, merge two pixels by finding a visual medium - such as a blue pixel and a red pixel, side-by-side can be merged into a single purple pixel. To our eyes, this is visually acceptable, but when each pixel in the sculpty map is actually representative of the numerical position of a point in the mesh, this can begin to distort the object or even delete points.

---

Discussion of compression options for TGA is largely irrelevant, for two reasons.  First, as we just discussed, SL does not use TGA internally.  Second, SL cannot even import compressed TGA files at all.  Only uncompressed ones can be used. 

Since you brought it up, though, let's talk about compressed TGA's for a moment.  The compression scheme that can be applied to TGA is called RLE compression.  It does not work very well with continuous-tone (photographic) imagery, which is so common in texturing.  In some cases, it can even end up increasing file size, rather than reducing it.  It only really works well with pallette-based imagery, like line drawings, and such.  Because of this, it's industry standard practice to use uncompressed TGA's in texturing.

 

What SL actually does use is JPEG2000 compression, which is far more sophisticated than RLE.  Among other benefits, it is optionally lossless.  SL allows a losssless option for small images, including sculpt maps.

 

---

Geldsbard Freck wrote:

When DirectX or OpenGL (the graphics drivers in our computers that actually draw Second Life on our screens) performs a LOD function (Level Of Detail) on an object mesh, it "culls" (removes) vertex at specific intervals - sometimes every other, sometimes every third, sometimes every forth vertex, according to your graphics settings.

---

A few things here:

First, SL is an OpenGL app, not a DirectX app, so mention of DirectX is pretty much irrelevant. 

Second, and more importantly for this discussion, the vertex culling for sculpties is built directly into the viewer software itself.  The culling happens according to a very strict, entirely preset, reduction scheme.  In other words, which vertices get culled, and which ones remain, are entirely consistent, and entirely predictable.  By knowing which ones are which, it's possible to create "LOD-proof" sculpties, which maintain their basic shape, as the LOD drops from level to level.  The effectiveness of this depends on the basic shape itself, of course.  If it's a suitable shape, it can work really well.  If it's not, then there's not much to be done.  Generally speaking, it works best with organic forms, and with relatively simplistic geometric shapes.

Third, neither DirectX, nor OpenGL, are drivers.  They are API's.  There's quite a difference there.

 

---

Geldsbard Freck wrote:

If the height map (Sculpty map) has been compressed in the upload to SL, thereby distorting the mesh to begin with, the LOD function has a much more difficult time deciding which vertex to remove and can (and many times does) removes the wrong one creating an even more heavily distorted object.

---

You are coorrect that if the sculpt map were uploaded with lossy compression, then the model would be distorted, right from the start.  However, you're a bit off the mark when you say the LOD function would have a harder time deciding which vertices to remove. That's just not how it works.

The sculpty LOD system removes the same vertices, each and every time.  It doesn't "decide" anything, ever.  It's not capable of making any decisions, or of "removing the wrong ones".  It removes exactly the oens it's been pre-programmed to remove, always.  It has no other choice.

If the sculpt map ends up artifacted, due to lossy compression, then all that happens is that the corresponding vertices will end up at undesirable coordinates, causing the model to look funny.  The actual topological structure of the model does not change.  It's still the same grid of quads, no matter what.  And the exact same vertices get removed from that grid, at the various LOD levels, no matter what.  Whether lossless or lossy, the sculpt map has no power to change this. 

With that in mind, the question is not whether the right or wrong vertices will get culled, but rather whether or not the artist has structured the model in harmony with how the LOD culling happens.  For example, if the artist made a corner from a vertex slated to get culled at medium LOD,, then the corner wouldn't maintian its shape when the camera moves out to medium distance.  If, however, that same corner were made with a different vertex, one that doesn't get culled until lowest LOD, then the corner would keep its shape until the camera gets much further away.  And if its made from a vertex that doesn't get culled at all, then the corner would keep its shape, no matter what.

Being an effective sculpty artist means being aware of which vertices are which, and making intelligent decisions about which ones should form the most vital parts of the overall object shape.

 

---

Geldsbard Freck wrote:

Many casual (?careless?) builders either do not know or don't care to check the "losses compression" box in the image upload dialog when they upload their maps to SL. You may not be able to detect any "smudging" with your eyes but an analisis of the numerical data would show missing vertex or out of place vertex as a result of even slight compression of the image.

If this is the case with your purchased object, unfortunately, there isn't much you can do about a disappointing LOD reduction.

If it is your own creation, simply re-upload the map and be sure "use lossless compression" is checked in the upload dialog.

---

True, but this does not seem to describe the problem the OP mentioned.  I understood the post to mean that the high LOD looks good, but then it quickly fails as the camera moves away.  If the sculpt map had not been uploaded losslessly, chances are the high LOD wouldn't look so good.  Lossy sculpt maps tend to produce lumpy or "melted" looking sculpties.  When the shapes are nice and clean, that's a strong indicator that the map was indeed uploaded losslessly. 

Neither scenario says anything about what's going to happen when the LOD drops.  That has everything to do with the way the artist structured the source model.  As I said above, if the model was structured so that those vertices most directly responsible for the overall shape of the model are the ones that won't be culled very early, then the model will maintain its shape pretty well.  If the model was structured more carelessly, then it will likely fall apart fast.

 

From the OP's description, there are two possibilities for what's going on.  One is that the model is not structured all that well, meaning the vertices that get culled early are being used to form the basic shape, instead of to form less important details.  The other is that the model shape itself is too complex to be made from just a single sculpty.  Both of these things happen all the time, and both are 100% due to poor decision making by the creator.

The former tends to happen when the creator either doesn't know or doesn't care how sculpties actually work.  The latter happens when people try to push sculpties beyond the limits of what they were designed to do.  People make things like one-prim ladders or  one-prim picket fences out of them, and then wonder why they fall apart.  That's just not the kind of thing sculpties were ever meant for.

In any case, now that we can use proper mesh models in SL, most of these sculpty issues really don't matter anymore. 

 

My suggestion to the OP:  Since you're going to be learning Blender anyway, do yourself a big favor.  Ditch the sculpties, and make plain old ordinary mesh models instead.  Do it well, and they'll look better, use less resources, and behave a thousand times better with respect to LOD. Plus, you'll be able to use them in other places besides just SL.  That's a total win-win-win-win.