Judah Nielsen

One "workaround" I've found is to render multiple files together using POV-Ray, which seems to have slightly better performance. For instance, on a computer that crashes out at around 30,000 pieces in LDD, I can render scenes of at least 56,000 pieces in POV-Ray (though this used something like 8GB of RAM on my 12GB system, whereas I don't think RAM is the limiting issue in LDD. Depending on your machine's specs, it may not buy you as much headroom). http://www.flickr.com/photos/laluneetmoi/7657121986/ - 4 ships which together total 56,000 pieces. The largest of these is actually a hollow shell of around 26,000 pieces, because I couldn't get LDD to save and operate on a file containing the interior. However, this has a few limitations: You can really only use this for the final step (rendering), so if you are rendering a single unified model instead of a multi-model scene like this, you have to build very carefully so that things will all line up, you may have to reposition the camera in POV-Ray, and this is a complicated exercise that you can only be sure of after a fairly intense render (since even making a 300 x 200 render to make sure you placed things correctly will mean sitting through a very long parse period). The basic procedure works something like this: 1.) In a new blank LDD model, place a grid on the ground level that is as long and as wide as your final model will be. 2.) For each piece of the final model, import it into the grid and place it in the correct position. 3.) Save this as a brand new file. 4.) Reload the original grid file, and repeat 2 & 3 until you've created all of the part models. 5.) Now go back through each part file and remove the grid, so it doesn't render in POV-Ray. 6.) Process each lxf file through LDDtoPOV. 7.) In POV-Ray, copy and paste each model together so that your final model is a union of all of the parts. 8.) If you weren't able to keep the same camera angle in the parts, you'll need to reposition the camera in POV-Ray, which is pretty tricky, but some advice on camera positioning has been given in the LDD to POV thread here. 9.) Test render and hope you haven't mis-positioned, mis-copied, or otherwise botched this complicated procedure. LDD performance at large part sizes has long frustrated me, and when I worked out this method (with the help of the ever-resourceful hrontos), I thought it was going to open up my building to bigger and better things, but it turned out to be far too much work to keep my interest. But maybe you have more drive than I do and this will be helpful.

Large models tend to crash for me if I take them much beyond 25,000 pieces, but sometimes repeating the same operation will succeed. I've managed to successfully import a model into a scene with an existing large model to get over 30,000 pieces, but once that's done, almost any action will crash the program, including saving the lxf. However, if your computer has a lot of RAM and and you're willing to render everything in POV-Ray, you can place all of your sub-assemblies so that they would line up, save the individual files, make a union of the converted pov files, manually adjust the camera so that everything is displayed, and render quite large scenes (I've been successful up to 56,000 pieces this way). It is a great deal of work, though.

drone_fighters.lxf Nothing transparent at either end.

A couple of strange artifacts in my latest render: I downscaled the full image, but I put pixel-for-pixel copies of the little meshy artifacts in the corners. Any idea what caused this?

It seems that color 47 - which LDD has as Tr.Floure.Redd.Orange, and which peeron and others list as Trans Neon-Orange - is rendering in POV-Ray as opaque. It appears that it was used extensively in the Ice Planet 2002 theme, so I did a super fast stub of a cockpit to show the issue. LDD: POV-Ray: If I remember these sets correctly, there was something milky or otherwise odd about parts in this color, but they were certainly transparent.

My math suggests this number is a factor of 2 too high (i'm looking at a triangle with camera distance as the "adjacent" but only half the model length as the "opposite", which is where we differ by 2), and my test render does have a lot of whitespace. You need some whitespace, especially if you're rendering a shadow on the baseplane, but I'm going to see if replacing the denominator with 1.5*tan(...) gives me a better result. I also suspect, but haven't done the renders to prove it, that this would be even worse if you had the smallest dimension facing the camera (in this case the front), since the distance is always calculated from the largest dimension, which might be going into the screen.

Sure. The full 9MB png is at http://www.onemoreproject.com/LDD/CL-02.png You mentioned a way to produce a mathematically straight view in POV-Ray. This model (and some of my other large models) is pretty unwieldy in LDD, but I'd like to be able to get identical views for the different ships I've made, so any kind of tutorial on that would be fantastic.

I took the advice of some others in the thread and rendered it at 5 times the "final resolution" of 1920 x 1200 with AA off. I went ahead and left the LEGO off the studs because I knew they would be invisible even at 9600 x 6000. Pixel level render of the rear battery - original size available on flickr on Flickr I walked away during parsing, which took a little under 2 hours on a fairly competent machine. I suspect final memory usage was around 5000MB. But you're right, rendering itself went pretty quickly. I suspect it only took three or four hours, but I went right to sleep.

Once again, amazing work guys. This 560 stud, 23000 piece ship rendered overnight, and looks great, although at this scale it looks less like "real LEGO" and more like "computer graphics" than some of the other renders I've seen here. Light Cruiser - POV Render by JudahMoTron, on Flickr

Well I chose my most complicated file for the test run, so I won't have a real presentable image until sometime tomorrow, it seems, but this is a fantastic development. It seems like the positioning of the model needs to be done in LDD. Is there some easy way that I've overlooked to get perfectly "straight on" views, like a perfect side view or a perfect top view of a model?

I added a visor to the copy of this module I recolored, but you're right, it was easiest to attach it to a fresh, level helmet and then move the whole helmet assembly onto the head.

I've created a few modules and ship using this concept. All modules are linked to from the flickr photos in this set, but they can be found at http://onemoreproject.com/LDD/MODULES as well. Attack Ship [LXF] - New gun pods and smaller wings , deleted forward guns [hard point cover]. The Dragon [LXF] - Deleted tail fin from original model [tail delete module], added gun pods in between heavy weapon module and wings. Interstellar Interceptor [LXF] - Extended fuselage [fuselage]and jump drive engines [jump drive]. Fighter Trainer Variant [LXF] - Extended two-seat cockpit [cockpit], deleted forward guns.

Since bbqqq uploaded an lxf file of 10030 Star Destroyer this week, I thought I'd post one more little link here to show a size comparison for my largest ship, the Cleveland class light cruiser. Larger sizes available on Flickr

Nice looking ship. I've never minded seeing studs.

I've often wished for the ability to somehow paint connection points. Tell LDD exactly which bricks from each subassembly need to connect, and then let it work out whether I've got a legal connection. My least favorite part of assembling a large model is trying to coax one of the imported sections into exactly the right position with respect to the rest.