Renderbricks

Can someone explain me why I get different and wrong results? Thanks a lot! - I used the latest LDD and load the .lxf +++ result in picture 1 - I export to .ldr and load into LEOCad +++ result in picture 2: parts are missing, wrong placed and rotated - I export to .ldr with the latest beta ldraw.xml and load into LEOCad +++ different result in picture 3 but looks more correct - I set the custom parts library to the original LDRAW to get my modified studs with LEGO +++ result in picture 4: parts of the car wings and the rear mirrors will be replaced by small cubes - I load the .ldr with LDRView or MLCad +++ result in picture 5: car wings parts complete but rear mirrors missing

I did some .obj tests with LEOCad but the results are bad. The model list and naming looks nice but the geometry can't be rendered properly. And it looks like that the parts library of LEOCad is different because when I set the customs part folder to LDRAW I get a different result. Some parts are replaced by cubes.

The textures are there and will be captured to the 3dxml format but unfortunately SimLab Composer can't read and convert them.

Here's another research test with the 4483 - AT-AT set. The capture from LDD via SimLab Composer to MODO will take just a few minutes. If I had a prepared scene with setep materials I just would import, do some adjustements and then I am ready to render. The method is great to get nice renders from LDD models but the limited geometry will not deliver the right material for realistic renderings with all details. Some parts may cause shading issues like the bigger part at the head of the AT-AT.

Use the 1.1 version. I couldn't see a difference in the results.

I am looking forward to your renderings. You have some examples from an actual work in progress?

All I can say that the Printscreen method gives satisfying results. If you want 100% realism you would also fail with LDRAW or other parts because then you need some kind of NURBS models which are res-independent and everything should have the correct bevel scaling because LEGO parts use different bevel sizes. The LDRAW parts are nice but for close-ups they won't work either because you will recognize edges.

Concerning photos of models. If you want to make beautiful shots of real models you need experience as a photographer and knowledge about lighting rigs and methods. In advertisement every packshot of a product is optimized, cleaned up and finally the client wants an idealized version of it what doesn't exist. LEGO might be different because when I watch pictures in a catalog it might be emotionally better to see real models to dream about to have exactly these ones. BUT for the box with the set it can be better to have a perfect visualiszation of the product. If you add some details like bumps, specular maps, random roughness on the edges etc. it can look amazing and a photo might look cheap compared to that. And finally: if you have setup a proper scene where you just load the models and materials and textures will be set automatically you just need to adjust some lighting setups and you can produce the pictures faster, more efficient and in a similar quality. And on the other side: if the 3d isn't really good it's always better you use a real not that good photographed model.

Here's a fast spin-test. I have no idea why VIMEO's compression is so bad on this movie. I need to find out. The spin-test has 200 frames in full HD what took 12 hours overnight on my DELL to render. https://vimeo.com/95264501

I have to disagree. It depends what you like to do. The capture of the ship model till import into MODO took a few minutes of work. Totally easy. If I had a material setup there I just import into my setup scene with an HDR and that's it. Finally the LDRAW parts have issues, too. Not everything is in hi-res and in LDD you catch exactly what you see. E.g. the Bionicles models can be rigged now and animated. Ofc it would be nice to have a better resolution but overall you can do very nice stuff with your LDD creations when using 3Dvia Printscreen, SimLab and MODO. The only problem I need to solve is the tesselation issue. Here's a spin test of one LDD test. https://vimeo.com/95264501

I am in contact with the author(s) of the LDR-Importer for a while. They are rewriting it completely but the actual version 1.1 works very nice. the 1.1.5 version is unfortunately 10 times slower. They will consider instances (linked duplication) in Blender what will be supported by the Collada format, too.

1. "lego-on-studs".....OK (using modified LDraw bricks) The same I do. 2. partial "grain" look for skew parts of some slope bricks.....OK (using Blender editing just the skew face of only single one brick applying separate material for that part and then linking all such bricks to this one as "Object Data", so any change on the first one appears on any brick linked to this one) That's nice. 3. find a way to apply "seam" effect presented in LDView.....OK (using Blender modifier called "SimpleDeform" applied on every brick) Did you check if this won't destroy or mess up some geometry? In MODO I use a shader option called "rounded edge". This will fake a bevel edge what will give you automatically a seam effect. If you look very, very close you'll see that the bevels are faked but it will keep the original more simple geometry. To add bevels by an operator failed. It frakks up the geometry here and also round faces will be beveled. And without instances the geometry will grow insanely. I did some tests with LEGO studs using their own bevel because the bevels are bigger compared to the edges of the bricks. I used a tool called Primitive Generator but it made the bevels too complex and a test showed up that a model like the Millennium Falcon UCS will have so much more geometry what makes no sense then. What is very important. Your converter should take care about instances. A brick used 100 times should use one master and 99 instances. Is this what you talk about in 2.? Linked duplicates? With this you can deal with very complex models. Actually the file size of the Millennium Falcon is 1.6 GB here in MODO. With instances it should be significantly less.

The MODO tests are basic atm. That means: I use the models, materials and a HDR. To make it more realistic I have to add slightly bumps onto the surfaces and give some random roughness to the edges. This will happen when I have finished my test and created a basic material library. The random micro-unligment of the parts would be possible by a script but then the parts start to intersect. It would be very complicated to consider collisions to prevent intersections. Animalogic - the creators of the LEGO movie - did a great job with this. They added all the necessary detail to the bricks like specular maps to make them looking used, scratches, bumps etc. It's a huge amount of manual work and clever scripts, complicated setups and finally they developed their own renderer because the commercial ones failed to render the insane complexity. To me the LEGO look in the movie is using a bit too much subsurface scattering. But overall the whole movie is a visual blast what makes you think that everything is build in real and animated by stop motion. 3Dvia Printscreen just can catch what we see in LDD. The parts might be low-res but you can render them with very nice results. The 3dxml format can be read and converted with SimLab Composer. Unfortunately the textures will be lost and strangely some parts using the same colors are merged. I can't say if this is 3DVia Printscreen or the importer. I will contact the developer and ask if they can fix that in the next release. And there's one issue with the geometry. The geometry is tesselated what will cause wrong looking reflections in MODO. E.g. on a tube the reflections will follow the diagonal lines and they look slightly rotated. I tried to fix that by quadrangulating the models but the operator might destroy important parts. CHECK OUT THESE TESTS WITH BIOLICLES. Or this rough test with a more complex model. I deleted the diagonal edges and when you look at the shark you realize that the surface is damaged by this. Click on picture for details in 8K. And keep in mind that these picture render extremely fast. I tried LDD2POVRAY and dropped it because a picture like this in a not that good look would render hours in 8K. Here we talk about 30 minutes. But there's still a lot of grain. To compensate this the rendering would take double time easily. I have to check the stats at work again.

MODO is also available for Mac.

Here's a test. I spotted some problems with the export to LDR. Some parts are missing, some are wrong placed and some are rotated. I also tried the latest LDR.xml file. Usually the missing parts are listed in lDRView and I google for them and try to fix it by renaming in the LDR-file. Some parts have different numbers like 12345a.dat instead of 12345.dat. I didn't tested the capture method with 3dvia Printscreen yet. This should work but then I get low-res models and no studs with LEGO logo what isn't an issue here with the Technic parts. Click on picture for 4K.

They are pretty awesome except some glitches with different resolutions in one model. Overall it's a gift that LDRAW exists.

The bevels are created by a rounded shader using normals.

Might be the rubber material.

I want to render those cars. If there's an LXF or LDR just let me know. *:-)

Update on a former test. Fixed some missing parts but some bended hose is missing here. Here's a wallpaper in 3840 x 1080. Click on HD picture for the full 4K version.

Nice one. If you like just send me the file and I show you my result on this *:-)

LDD2POVRAY is indeed a simple for free method and every other method needs more steps and knowledge. LDD2POVRAY is the only tool what can render .lxf files directly with POV-RAY. Every other method needs more steps and knowledge. I am using MODO atm with really nice results. I am rendering with a 12 core DELL and 32 GR RAM and the rendertimes for HD pictures are under 10 minutes even with complex models like the Millennium Falcon UCS.

Phantastic. You have a LDD file of that I can render? *:-)

This website is really impressive.

HDR and SubSurfaceScattering is important here. And the rendertimes are nothing. I did a spin-test with the first camera over 200 frames in 1280 x 720 what took 1,5 minutes for each frame. I need to organize my mess now with all the tests I did. The LDD capture works fine but takes some manual steps. The 3DXML format contains the textures what is shown in a 3DXML player but SimLab - what fortunately can read this format - unfortunately can't convert the textures. After importing to MODO by fbx I recognized little shading issues caused by tesselated objects. Quadrangulation helps here to fix this. I am not sure what ware the plans of LEGO for the future versions of LDD. Hi-res models and and an export into fbx or Collada would be totally awesome but I am sure that this won't happen for a consumer tool like LDD. Probably LEGO won't anyway give access to their data. 3Dvia Printscreen is more a hack. *:-)