R Reynolds

Nice work!

The archway to the concourse.

The main waiting room.

A view of the arcade from the floor of the waiting room.

Keep in mind that this is a WIP and I have a long way to go (including getting the lighting right but I'm still pleased with the result. Here's two views of my arcade taken from the street entrance end and showing the concourse archway in the distance.

The ALT snap bias aim feature is such a pleasure to use, I've regained my enthusiasm for an architectural model I set aside in 2009. It's my version of a Hollywood set of the grand old Pennsylvania Station built in New York in 1910 and demolished in 1963. I'm calling it a Hollywood version for three reasons. 1. Since my reference images from books and the Net only have a few readable dimensions I have to make a lot of "looks good to me" guesses. 2. The amount of detailing required for a 100% accurate copy is too staggering so I've simplified it while trying maintain a similar look. 3. The section of the building where people board a train is called the Concourse. While it was a spectacular work of architecture, it was essentially a large room where passengers would congregate before descending down stairs to tunnels in the lowest level to board coaches pulled by electric locomotives. To me that's a subway station not a rail station. So my Concourse will be a complete fiction but it will have steam locomotives. Here's four views of the real thing.

Here's a re-rendering with better behaved sheet metal seams. Not only is the seam now a bump map but it's been reduced from 100% "intensity" to 23%. Being produced at the Instagram worthy ratio of 720 X 480 also makes the noisiness in the reflections more tolerable. piggyback_walkthru_insta_h264.mp4

Excellent suggestion! Attached are three sets of four images with descriptive names for clarity. 9x9_noblur are the originals; 9x9 oversampling with no motion blur 9x9_blur are the same oversampling but with 100% motion blur 16x16_blur are sampled at 16x16 with 100% motion blur It certainly helps the chaos in the reflections but ultimately these grazing angled reflections of closely spaced vertical lines in a very wavy surface will be pretty noisy. That chaos is a good simulation of reality. If I really want less noise, I need to reduce the bumps in the sheet metal material. As for replacing the normal map with a bump map, also a great suggestion since it's trivial to generate and replace with no re-modeling required. I'll just have to get over my obsession with the improved (IMO) look of a normal map. Thanks, Robert. piggyback_walkaround_9x9_noblur_00.tga piggyback_walkaround_9x9_noblur_01.tga piggyback_walkaround_9x9_noblur_02.tga piggyback_walkaround_9x9_noblur_03.tga piggyback_walkaround_9x9_blur_00.tga piggyback_walkaround_9x9_blur_01.tga piggyback_walkaround_9x9_blur_02.tga piggyback_walkaround_9x9_blur_03.tga piggyback_walkaround_16x16_blur_00.tga piggyback_walkaround_16x16_blur_01.tga piggyback_walkaround_16x16_blur_02.tga piggyback_walkaround_16x16_blur_03.tga

I see two issues. First look at the following four frames and note the vertical sheet metal seam that runs between the last two letters of Canadian. To me it flashes like a fine, straight geometric detail that hasn't been sampled enough. piggyback_walkaround_a077.tga piggyback_walkaround_a078.tga piggyback_walkaround_a079.tga piggyback_walkaround_a080.tga What I find more puzzling can be seen in four later frames. piggyback_walkaround_a803.tga piggyback_walkaround_a804.tga piggyback_walkaround_a805.tga piggyback_walkaround_a806.tga I think the noise in the reflection of the factory's vertical siding in the wavy sheet metal would be improved by higher oversampling. But the sheet metal seam (actually created by the normal decal below) that comes in from the right seems way too reflective with no underlying green colour. This is why it's so obvious in the previous four frames. Getting the camera this close was an experiment on my part. After all you can see the pixels in the decal of the word SERVICE as it passes. I figured a decal for the seams would be an adequate way to reduce patch count but I may have to bite that bullet with these chromed seams. But for generating Instagram content for the phones of friends and family, it'll do.

They're not compression artifacts. They're just as obvious in the rendered TGAs and the uncompressed AVI. I suspect it's an anti-aliasing issue so going to 4x4 or 5x5 oversampling would be the first test I'd run.

Try it again Robert, I replaced the video with one that uses the H264 format.

Usually I do a walk around of my models but the piggy-back setup allowed me to do a climb over and between. piggyback_walkthru_h264.mp4

Dang! I've used that technique before and I still didn't think of it for this case. Thanks for the reminder.

I'm building a formed steel frame, a part of which is shown below, to fit inside an archway . I wanted the vertical column to intersect nicely with the curved section, so I built the curved beam first and made a copy of it as a background template. After completing the orthogonal Tee section with all the corners and five pointers, the final step was splicing the Tee section into the curve and massaging it's cp's, mags and bias values so it's shape matches that of the original curve. My question is based on the fact that when you lathe a circle in A:M, no matter how many points make up the circle, the correct mag is always calculated to get the splines to lie on a circle. A four point circle gets mags of 167.39 while a twelve point circle uses mags of 105.78. So does a mathematical formula exist (at least on paper) that says if you want a spline that matches a mathematically defined shape, an ellipse for instance, input the coords. of a CP anywhere on the line and the output is the correct mag and bias to get the best match? I'm guessing the answer is no but I thought it worth asking.