Drongle McMahon

Yes. Turn on backface culling in Blender to see one-sided faces as in SL. That makes it easy to spot inverted normals.

Does the same thing happen with a simple cube with two (or more) different materials? If it does, can you make the dae file of that available?

If you have specular (glossiness) but no or low environmental reflection, then whether you can see the effect of the normal map depend on the angle of the sun and the or the camera. You can see the effect quite well with the diffuse lighting where the light is glancing the surface, as on the sides of a cube at midday, but on top, te diffuse reflection is much more even, so that you mainly see only specular effects, and these are very angle dependent. Here, for example, is the set of cubes from that other thread, viewed with the sun set at 3pm, either looking towards the sun (top) of away from it (bottom). The top looks much flatter in the sencond view because there is no speculat reflection from that vcamera angle. There is only a bit of lighting variation from the normals because the sun is still pretty much shining down on the surface. I think this is the sort of effect you are noticing. This is one of the reasons for using the cavity map, as on the middle cube.

That was using geometry, which is what you asked for. However, there is a much easier way. Here are the equivalent pictures using a different generated texture from Blender (Musgrave Hetero Terrain with tweaked parameters). This time I just used a simple plane with no subdivision, UV mapped, set the texture to Influence only the normal, and baked the tangent-space normals onto itself - no second plane required. This gave the normal map. Now, instaed of a AO bake, I used the method I described in this thread to make a cavity map from the normal map in GIMP. Used the cavity map as the diffuse texture with a rusty tint, and the normal map in bumpiness. Blank specular map with a bit of glossiness added (25). The again, depending on what you are looking for, you might want to use diffuse and normal maps that are based on completely different textures. That can provide a different sort od effect.

1. Add plane. plane 2. Subdivide 3 times (or once with 7 cuts). 4. UV map. 5. New material. 6. New texture - leave as clouds, a. Set Mapping:Coordinate to UV. b. Set clouds:basissize to 0.05 c. Uncheck Influence:diffuse:color d. Check Influence:geometry:displacement e. Set Influence:geometry:displacement to 0.05 7. In UV editor, Image:new:256x256 blank 8. Camera:bake ... a. Make sure Selective to active is unchecked. b. Check Normalised. c. Set Bake Mode to Ambient occlusion d. Bake 9. Image:save ... this is your ao map. 10. In object mode, make a new plane and put it eacatly aligned just underneath the first one. Don't subdivide it. (you can align them bu doing transform>origin to geometry on each and then setting their X and Y positions identical). 11. Select the new plane, then shift-select the old one. 12. Camera:bake ... a. Set Bake Mode to Normals b. Set Normal Space to Tangent c. Check Selected to Active. d. Bake 13. Image:save ... this is your normal map 14. Upload to SL and apply aomap as diffuse, normal map as bumpiness, set specular map to blank. You can Set the duffuse colour for variation, or you can combine the ao map with another texture by using multiply mode in GIMP or PS. There are infinite variations of the above with different textures and different and additional parameters. Experiment in Blender by rendering with a couple of lights. Then there are specular maps... This was for a 256x256 texture with one vertex per pixel fot the displacement. You can get better results using more subdivision. Here are the maps made as above, a Blender render, and what they look like in SL, 2x2 on a prim cube, rusty tint, blank specular map with default settings. With higher repeats there is a just discernable seam. If you need to remove that, you have to make the maps seamless, which is a different story.

I have struggled a bit with this dirty vertex thingy. Let's see if I have understood any of it. It is supposed to darken the cavities. I's not the same as AO, but vaguely similar in some ways. The differences are (?) because this is supposed to mimic accumulation of grime rather than purely lihjting effects. With the right settings, it can darken flat surfaces too, leaving only protruding edges, from where the dirt gets rubbed off, un-darkened. So while it can mimic AO, it has a wider repertoire of effects. However, as far as I can see it only works with real geometry. Maybe I was missing something? I couldn't get it to work with just an active subsurf modifier. I had to apply the subsurf and then add a displacement modifier to get the required geometry. In contrast, the AO could be done with a un-applied subsurf and using the displacement map as input to the material. It didn't have to be a displacement modfier. In any case, neither AO nor the dirty virtex things can be done without real geometry in one form or another. So that's quite distinct from the methods for making a map directly from a normal map. If you are making the normal map from geometry anyway, then needing it is not a problem, but in some of my examples, the normal mps are made from a fractal in GIMP; no geometry anywhere. Of course I could have made a displacement map from the same fractal (in fact that;s essentially what you do on the way to making the normal map) and used that as a displacement map in Blender to generate the AO or cavity map, but that is much more time consuming, and would need very high subdivision to get the same resolution. The effects are certainly not the same. Thw direct method produces highlighted edges as well as shadowed areas. So it is useful to have these different methods avalible. Anyway, just for fun, here is a comparison of an AO map (top) and a dirty-vertex-cavity map (bottom) made from the same displacement map in Blender. The displacement map was a procedural texture.

...keeping the Level of Detail on Low keeps it down to 1 prim as well. You mean for the physics shape? Yes, it depends on the model. Sometimes the low LOD is too distorted to produce a satisfactory physics shapel sometimes it's OK.

That is the default convex hull physics shape. You need to set the physics shape type on the Features tab of the edit dialog to "Prim". That will use the shape set up on the physics tab of the upload dialog. You may need to work on that, especially to keep the phyics weight (ang thence LI) low. For your railing, it could be a new mesh file (dae) containing just two non-overlapping cubes adjusted to fit the two parts of the railing - then click "Analyze". If you still have two objects, then the physics file needs to be two objects as well, and they need to be in the same order as in the dae files. If you play around with the settings on the physics tab, you might be able to get the same effect using one of the LODs for the starting point for the physics shape, but once you know how to do it, it's usually better to make a physics model.

Here's a better example. Top with ALM, bottom without. Middle has cavity map. others don't. The cavity ,ap doesn't improve the appearance with a detailed diffuse texture as at the right, even after shifting it 1 pixel to get proper registration. Sunlight only, 3pm.

Here's an illustration of the value of the cavity map, using a simple normal map so that it's clear (I hope). Small picture shows the original pattern (height map, traced from the real Chipendale catalogue!), the normal map made from it with GIMP Norma;map plugin, and the cavity map made from that the way I described. The big picture compares two simple prim boxes. Both have the same normal map and use the cavity map as the specular map (no alpha channels used). Shininess is 100 and environment is 0. The nearer one (in 3/4 views) has the cavity map as the diffuse texture. The one at the back has the blank texture. The blue is from the diffuse coulour, which is darker and less saturated for the blank texture to compensate for the greyness of the cavity map. Top left is with ALM and AO on. Top right is AO turned off. I was wrong - the inworld AO is apparently affected by the normal map detail, although it's too coarse to be really appropriate for this size of detail. Lower left is with ALM turned off. Lower right is looking from the opposite direction where reflection from the sun is maximum (3pm). In this case, the difference is minimal. So, as expected, the cavity map does greatly improve the appearance of the details when the sun is not at the optimal angle. It also allows some detail to remain apparent when ALM is off. Of course in normal use, the map would be combined with a more complex diffuse texture, and the differences would be less priminent, but still worth having. Importantly, this can be used for tiled textures as well as custom textures with baked AO.

I think I have worked out a way to get a virtually identical result in GIMP... If normal map has alpha channel, remove it ... 1. Colors/decompose/to layers/{RGBA|RGBA}* With resulting new image... 2. Delete blue layer (and alpha layer) 3. Select Green layer 4. Filters/Edge-detect/Sobel/{keep sign, horizontal only} 5. Select Red layer 6. Filters/Edge-detect/Sobel/{keep sign. vertical only} 7. Make read and green layers visible 8. Set Red layer mode to Grain Extract 9. Merge Red down 10. Export ,png. * RGBA if normal map has alpha channel, otherwise RGB Here's an example (done at 512x512, this image reduced). Didn't put the map in your link here because of copyright, but the result was just about identical to that shown in the link.

Do werewolfs wear clothes? As far as I recall, in all the werewolf films I saw, the clothes disintegrated and fell away as the change happened. :matte-motes-smile:

In case you wantb to make it one object next time, here is how to do that. Each mesh can have up to eight materials. These can each be textured with a different texture inworld. The mesh faces sharing each material behave like the faces on a standard prim. You can select a face with the button above the edit linked checkbox and then clicking on the face. Then you can adjust all the parameters and select a texture for that face independently of the others. You can also drop a texture on it and it will just go onto that face. One material can be spread over disconnected parts of the mesh surface, like the posts in your railing. The picture shows the simplest steps for defining and applying different materials in Blender... 1. Open the materials tab of the properties panel. 2. Click + to add a new material slot. 3. Click + to add the new material into the slot. 4. Choose a distinct colour (so you can see what you are doing). 5. Give it a suitable name. 6. Select all the faces to receive the material and assign the material to them with this button. (Note; the first material it will be automatically assigned to the whole mesh, but for subsequent materials you need to assign them.). Now when you import the mesh, it will have the colours you assigned to the materials showing which faces are which. You can select them with "Select face" and then turn the colour to white and assign the desired texture. This is just the beginning. There is tons more to learn about materials and UV mapping. Have fun.

After a bit of reading, I think I can suggest the following definition: Ambient occlusion is the variation in diffusely reflected light resulting from the restriction of diffuse environmental light by nearby geometry. The typical AO bake approximates this effect of the geometry physically present in the mesh it is baked from. However the tangent-space normal map represents the presence of finer geometry that is not present in the physical mesh and is therefore ignored by an AO bake. The normal map affects shading due to the variation of the surface angle resulting from these finer geometric details, but it doesn't produce their AO effects, The purpose of the cavity map is to approximate the AO resulting from these finer details, which are represented in the normal map but are missing from the actual geometry. That's why it is derived from the normal map. I assume that the inworld AO does only uses real geometry, ignoring the normal map. This may be part of the reasons why so many people complaing about the minimal effects of normal maps on mesh lit only by sunlight. Adding this effect using the cavity maps as described may be effective in alleviating this lack of effect. Thanks for bringing them to our attention.

"That's 6 objects in Blender total" Made that way, yes, although the LI would be 3*, because eacj would have the minimum 0.5 (server weight). If byou make it all one mesh, then it will be 1. You might need to learn about materials so that you can pu a different texture on the posts and the handrail. *as long as you don't waste polygons.

Lovely. That's the sort of thing parametric mesh would make possible for inworld building with mesh, as has been discussed here ages ago. It would work with local rendering while you build. followed by the usual upload process when you have it finished. Almost enoughy to make me want to set up to compile my own viewer to put this in.

... so the larger it get the more degraded it becomes? In the sense that the flat surfaces are larger, yes. The high LOD* prim cylinder is always 24 sided, whatever its size (the one in my picture was 50m diameter). Then of course, the smaller a cylinder is, the sooner it will switch to lower LODs as the camera moves away. It goes from 24 to 15 to 9 to 6 sides, as shown in the picture (taken inworld in wireframe view with surface patch, water and sky turner off, adjusting renderVolumeLODFactor to switch LODs). Sothe accuracy becomes less again, but it doesn't matter because you are too far away to see the difference when the LOD switches. * level of detail.

Apologies in advance, but I'm going to be ridiculously pedantic. "Cylinders" in SL are not actually cylinders. They are 24-sided polygonal prisms. The consequence of this, and of the way the path cut works, is that the thickness across a radial cut depends upon the position of the cut relative to the vertices of the polygon. The top picture is two inworld hollow cylinders, identical except that the blue one has a pathcut as nearly as possible half way between polygonal vertices. Note that the pathcut code does not calculate a new cut face at the proper radius, but simply cuts across the existing trapezoidal face. It is simple then to calculate that the ratio of the lengths of the cut at this point, with that of a cut through the vertex is cos(7.5 degrees). This is seen more easily in the construction velow, made in Blender. Thus the optimal hollow to fit exactly against certain wall thickness depends on the position of the pathcut. The most extreme difference is with the cut half way between vertices, as shown. Then it is about 1%. (Not that anyone needs to care about that! :matte-motes-smile:)

The problem is that the sparkle of real diamonds comes from refraction of light, and there is no refraction in SL. So it has to be faked. Maybe Googling "diamond cuts" and selecting images will give you some inspitation. You'll see that cut diamonds ar less smooth, with fewer facets, than your mesh.

Here's another example with a filled frame, illustrating some other mistakes/options. Four stages: 1. Left everything from loop cuts made to insert the walls without any hidden surfaces. 2. Remove redundant vertices, as Aquila said. 3. Delete faces at the bottom that will be underground and pointing down. 4. Sacrifice perfect joints by making the cross beam and the walls unconnected pieces (like the bottoms of Aquila's pillars). This saves the extra loops in the main pillars. Note that these pieces are not complete cubes - the edges of the walls and the ends of the cross beams are deleted because they are buried inside the main beams. This works, but you can run into problems with baked textures where one piece penetrates another unless you are very accurate. It's not as aesthetically satisfying as the properly jointed mesh ... but nearly halves the triangle count. (Tr = triangles; Vt = vertices).

I thought the same about LI. So I just made a similar model, a bit more detailed, with 12 step staircase inside, and at about 38x35x21 it's LI was 18 (download weight, default CH physics=1.7, but no good for real use, of course)). That's all in one piece,default medium LOD (you won't ever see anything lower). So that would be a good maximum target. More important, I was impressed by the inworld AO, This is the latest viewer. It seems much stronger than I remember last time I made a big building. It alsi scales with the ogject. Anyone know if there have been changes to this? If it's like this, thgen there is indeed no need for baked AO, so that tiling textures can be used without sacrificing AO. That will mean a huge saving in texture download and gpu memory consumption. Here's a picture, inside and out, with blank white texture (I haven't UV mapped it, and I'm not going to, as it was just an experiment). Of course, only those with good enough GPUs to activate AO will see this. So maybe this is for the future.

You can't avoid having a physics shape in the mesh asset, If you don't give it one, the uploader still creates a default convex hull from thr low LOD visible mesh. That can sometimes still have a high physics weight, If you set the physics shape type inworld to "None", that lets the physics engine ignore the object completely, whatever its physics shape, and reduces the physics weight to zero. However, you can't set the physics shape type of an unlinked object, or the root of a linkset, to "None". So to use this type the object has to be linked to at least one other, which will still have a physics weight. If the other is a (tiny invisible) undistorted cube prim, it will have physics weight of only 0.1.

How about flat vs smooth shading? Flat (sharp-edged) shading will increase the vertex count to double or more than with smooth shading. You don't want flat shading with cloth anyway, and yoy might find you can do away with the subsurf as well. Your triangle count is past the 21844 where strange thing happen with materials. I don't know if this has any effects on rigging, but it's possible. Try reducing the triangle count (by dissolving edge loops if you really want to keep the subsurf) and see if that helps.

Roligf's cishions make the point, but here is the picture I promised anyway; LOD & physics meshes; inworld with test texture and nice material (0.5m square); upload triangle counts.

The default physics shape is the convex hull of the low LOD mesh. This can be quite expensive for a convex object where lots of the detail atays in the hull. You need to specify a simpler physics shape. The easiest way to do that here would be to select the lowest LOD instead, but a costom mesh wou;d be better (pics to follow, maybe). However, your download weight is also too high, so reducing the physics only will just reduce the weight to that and no further, The real anwer then is that you have far too many polys in your high LOD cushion. Cut that right down, and you will solve both problems at once. There is no good reason a little cushion like this needs to have a LI more than 1 (or even more than 0.5 when it's part of a linkset).