Just wanted to share the latest project I've been working on; a diamond heart cathedral engagement ring.  This project was a model that had been requested of me from a friend on DeviantArt.com, and ended up being one of those models that took a lot longer than I ever would have expected.

I put a lot of research and effort into designing the ring, and optimizing that design to work best with physically-accurate materials.  I must have looked at 30 to 40 tutorials and examples of diamond models, and different methodologies about how best to go about creating diamonds and jewels for rendering in different render engines.  This isn't my first jewelry model, but it was certainly one of my most in-depth jewelry models; everything from the optimization of topology, to the decision of modeling strategy, I learned a lot in the process of creating this particular model.  Here's a few wireframe renders, illustrating the underlying topology I ultimately decided to use for this model:

The object consists of 9,602 polygons, and 9,592 vertices.  There are no textures with the model, but all the parts have been UV unwrapped, and work well with any procedural material. Here are some sample shots of the model I did with the Octane render engine.  They took about 3 hours each, on my Geforce GTX 660.

The object is available for free download on several free resource websites, including here on Renderosity, in case you wanted to take a closer look.

Wow! Amazing detail and fantastic model!

Quote - Wow! Amazing detail and fantastic model!

Thank you very much!  It was an exercise in trial and error to be sure.  I must have modeled about 10 different versions of the heart alone, and tested each one in several render engines with physical diamond materials, just to see how (if at all) the topology affected the render results.

In the end, I decided that trying to model the stones wasn't as easy as I first thought, because the realistic appearance depended heavily on the topological detail when using unbiased rendering, and very close up camera angles.  In a "biased" engine, like Mental Ray, I was able to model the jewels very simply, and get the results I expected, because the render engine used a lot of material "fakes" to achieve a realistic result.  However, in more accurate engines, like Octane, or even Luxrender, each facet and face of the geometry played a surprisingly crucial role in the realism of the refraction, caustics and dispersion effects.  In other words, the more realistic and detailed the geometry, the more realistic the light played with the materials.  I think this was particularly true since there were no control maps used to define the surfaces, and help with realism.

Anyway, I hope I made the right choices in regard to topology, so that the model renders as expected in any render engine, provided the lighting and materials are reasonably accurate.  I know some people might think I could have gone "lighter" with the polycount or edge density, but I wanted to be sure that the important parts of the surfaces here didn't rely too heavily on smoothing groups or edge weights, which could get over-ridden in some applications or render engines.  Adding support loops and edge density is the only way to be sure that the object renders as intended, no matter where it's used.

So, is this the actual topology? I'd always assumed, of all things 3d, that a diamond would not need bevelling and perfectly flat sides a must.

That said, your approach here, as in working backwards from the renderer, has payed dividens. This render is superb 

Quote - So, is this the actual topology? I'd always assumed, of all things 3d, that a diamond would not need bevelling and perfectly flat sides a must. That said, your approach here, as in working backwards from the renderer, has payed dividens. This render is superb

Yes, this is the actual topology of the model.  The wires above, and the renders are the same geo.

99 percent of the time, flat surfaces, and simple geometry would work just fine for such a model.  However, when distributing the mesh for use in whatever application the end user decides to use, some extra caution seems to be required; namely, the use of extra edges and polys, to ensure the flat surfaces do not receive incorrect smoothing by the default import settings in many applications or render engines.  Simply defining "hard edges" doesn't guarantee the results will be as intended.  The only way to be sure the object appears as consitent as possible across the various render packages seems to be strategically inserting support geometry to critical areas where you want to prevent excessive vertex or shader smoothing at all cost.  Of course, I don't want to add arbitrary, unnecessary detail, but adding extra edge loops and segments to certain areas definitely helps eliminate the risk of some users getting unexpected results with the model.

Also, in regard to Path Tracing, the results of caustics and refraction of certain materials, like diamond, seemed much more believable when I actually defined the edges and facets with additional geometry.  This wasn't the case in unbiased render engines, like MentalRay, but it didn't actually hurt the results either.  So best to err on the side of caution.

Quote - Also, in regard to Path Tracing, the results of caustics and refraction of certain materials, like diamond, seemed much more believable when I actually defined the edges and facets with additional geometry.  This wasn't the case in unbiased render engines, like MentalRay, but it didn't actually hurt the results either.  So best to err on the side of caution.

I'm not totally surprised by this actually.  Unbiased, accurate path tracing techniques used for global light propagation, like the one in Octane, tend to expose low resolution objects for what they are... low resolution objects.  You might get black edges and faceted shadows on glass objects, or reflective surfaces, if you aren't using enough polys on an object.  This happens when the low poly surfaces are smooth shaded.  The common advice in the manuals of most of these render engines is to either define all the edges as hard, or subdivide the object until the prolem goes away.

I think you went a little overboard hardening the edges with extra edge loops in this model, but I can't argue that the render you created is just great, and that the model is very well done for the most part.  Personally, I think you could have just added one extra loop to the edges that you wanted to keep hard, and it would have been good enough.  Seems like you actually added several additional edges that probably weren't needed, even under the most extreme circumstances.  I don't know of any 3D software that would try that hard to smooth out those edges.  Even if the default smoothing was set to 180 degrees, two edge loops should be enough to deter it.

Your renders of the model are just jaw-dropping good, mate.  You nailed the materials perfectly.  Good work on the design too.  My only complaint is that it's a little gawdy for an engagement ring, unless you were marrying a female rapper.

Quote - I think you went a little overboard hardening the edges with extra edge loops in this model, but I can't argue that the render you created is just great, and that the model is very well done for the most part.  Personally, I think you could have just added one extra loop to the edges that you wanted to keep hard, and it would have been good enough.  Seems like you actually added several additional edges that probably weren't needed, even under the most extreme circumstances.  I don't know of any 3D software that would try that hard to smooth out those edges.  Even if the default smoothing was set to 180 degrees, two edge loops should be enough to deter it.

I went through several tests before actually deciding on this particular geometry.  The reason I used the number of edge loops on the hard edges, is because I wanted to "round" out the beveled appearance, and give it a subdivided look, without actually using subD.  This provides a nice highlight on those edges, which wouldn't be there if they were just hardened without extra geo.  It gave me exactly what I wanted... more resolution on the "hard/cut" edges, without increasing the overall density of the mesh.

The geometry behaves as expected when imported in 90% of the applications I've tried.  Daz Studio, an application similar to Poser, seems to do something weird with the smoothing of the chamfered edges, but Blender, Max, and Poser all display it as intended, and even in the viewports, you can see a nice subtle highlight on those hard edges.

Very nice render.

Could use some flattening internally, but hey, it was 13 years ago in Bryce 4. Back then, embedding text and using Poser dolphin models posed around a diamond and exported to Bryce and then textured was a bit of a challenge.

Quote - Bryce Could use some flattening internally, but hey, it was 13 years ago in Bryce 4. Back then, embedding text and using Poser dolphin models posed around a diamond and exported to Bryce and then textured was a bit of a challenge.

Geometry aside, the difference in realism between your render, and that of the OP,  reveals a lot about how far technology has come since then as well.  I'm not insulting your work, just pointing out the obvious.  I can remember the days when Bryce was considered ahead of some far more expensive, and respected, render engines.  But it was such an isolated software, without the ability to play well with other apps, it didn't stand a chance in the professional industry, beyond the occasional matte painting or previz use.  The impractical UI design also helped doom it in professional circles.

I take no insult. It was yesteryear and Bryce was one of the apps that created a cult following.  Many of the cultists, such as myself. bought pro apps like your hated Autodesk Ultimate Suite, max/maya/mud/xsi/ etc.

The very first app that I learned to model in was Bryce.  Even if it didn't export, I learned about booleans and polygons and blobs.

It is kinda fun to look at today's renderings.  I recall the cover of Expose back in 2004.  It was stunning at the time.  Today, any novice can match or exceed that with the default settings in Poser FireFly. 

Lux has some good shaders and procedurals and lighting going there.  I always enjoy his renders.