Cineversity

This is a work in progress that I have a question about. First, let me describe the project.

This is an illustration of a train going eastbound in the San Gorgiono Pass near Palm Springs in early morning. In the background are the Mountains that can be seen. I used the very helpful DEM importing tutorial in Cineversity to download an accurate elevation relief map. Also, I obtained satellite images and stitched them together to create a terrain map. Of course they do not have the resolution for close to the camera but otherwise are used for everywhere else. I was able to use the satelite photos to create a shader effector for the billboard brush plants and also I used the changes in color to create terrain masks so I could apply textures to appear like there are pathways everywhere. Below is my current render in dropbox.

https://www.dropbox.com/s/2s8x3nun7ryjdsq/San-Giorgio-Pass-Illustration_Concept02.jpg?dl=0

My question is how would you approach making the train roadbed, ties and track so that it 1) looks 3d close to the camera, and 2) seamlessly goes on for a long distance (100s of meters). What I created can go on very long but close up looks like a flat photo on it’s side.

Hi vrauckis,

Good question and I have not done it. So, the following is my best idea, going through it the last ten minutes. I go with each step of my assumptions with the fore-mentioned step.

• The light direction comes more from the camera view than the opposite.
• The image will move fast, so “fine detail” noise from the background as well foreground will limited by motion blur
...based on movement and light, the trails itself will not produce specular high-lights.
• The main attention is the locomotive and the landscape, the camera doesn’t go very low to expose the details of the rail-track, e.g., clamps/bolds/nails, etc.
• The gravel and pebbles are not modeled.

Based on that, here is my brainstorming:

You could model a small section, larger than your texture is currently, I see repetition in the “noise”. Then use Render Instances, but do tests, there are always some limitations.

I exclude here any distance replace options as it might be more work than positive effect. Hence my main idea for now

Model it in any way you like, just that part of the rail. Try to have the option to model from low polygon to high polygon models, store the results. In the moment you have the model done. Render the camera move with a slightly wider camera and higher resolution than you need for the delivery.
Then render only the trail, while you work on something else. This camera is then used as a moving projection camera, but on the low resolution model. In this way you keep the options to illuminate the whole track (wood/concrete and gravel/pebbles) and perhaps do it without the two main trails, which will be a Sweep object, to keep the reflections.
The wider camera is used then to camera map (camera projection) the details and rough light situation back on it. The low res. geometry will supply enough details.
This can be rendered long before the deadline and in the color channel. Test if with or without shadow during rendering. This might give you a good head start, and a fast render on the end. While recording the camera projection pass, you might employ MoGraph options to clone parts and have them visible only in the camera “frustum”.

The wider camera allows for smaller main camera adjustments later on.

Alternative: Each track segment has its own camera-projection, with the camera position roughly in the middle of the segment being in main frame. Much more work, but faster in render. The two rails stay as Sweep so it is reflective.

My idea to use here camera projection might be more kind of bake texture than anything else, but any re-render of the final scene will show a positive effect based on it.

Cinematography Hint: our eyes move to changes, not to constant similar changes, but the light reflections on the rails itself might be something. As it has nearly no motion blur, if the camera moves along the trail, this looks sharper than the rest, equally to the locomotive. Anything sharp and contrasty will pull our attention, hence why I would keep it modeled. The Gravel and wood beams are more in danger to get “eaten” by motion blur, and blur is an attention killer, more or less.

Let me know where you see problems or what isn’t fitting to your ideas.
All the best

The MoGraph version.

I have used here only five different detail levels. Detail levels means here changes are large or small. The more you have the smaller the changes, but the more often things will change. It is a balance ...

The main idea is to create a straight track first, and bend it later. I have done it so far in movies like this example: https://youtu.be/psFGdd753Dk

Based on the straight first, the Plain effector, who is in charge to “sort” (modify clones) the details, needs to run along the straight version. But as only the z value needs to be adjusted, this is not a big work. There is a more than one reason why I do not connect the camera with the Plain Effector. But I guess based on the before mentioned reason, anything else is secondary. The size of the falloff is directly connected to the distance when things will swapped out. I have placed colors to the detail levels for an easy exploration. I would do that for a set up always, to be faster.

The low level objects can work with normal and bump maps to keep the difference small. My suggestion would be to keep the rails and wood-beams constantly, and replace only the gravel/pebbles based on distance.

Let me know if there is any question, I’m happy to look into it.

All the best

Preview here:
https://www.amazon.com/clouddrive/share/AEgCmk7eb4HLBIeQCbwuIz75OKsLxPeSZgITK-JEaEU?ref_=cd_share_link_copy

The theme is certainly an important one, as we all have things to do that might stretch our patience too thin, while waiting…

Let me add another though, but more complex to do, but perhaps the fastest and seamless. Just brainstorming here.

You could build the track completely in high resolution and have (via MoGraph) only the parts visible which are crucial to the current image. It has a low res locomotive (invisible to the camera as shadow/reflection dummy)

The rail (spline) is then used to move a card (plane-object) with a gradient in the luminance channel, to create a traveling mask. This mask must be smaller than the visible high res.-models. For creating teh mask-pass, this is the only object in the scene for that.

The low res.track model is then used for everything, no switching. The locomotive has an object buffer to complete the traveling mask.

====

In a nut-shell, the switching of objects is always a critical work, as it might be noticed. With a gradient traveling mask the transition can be more smoothly. Even if it is noticeable for pixel-peeper, it is not drawing a lot of attention anymore, as it flows with the motion. Perhaps covered anyway by shallow depth of field or motion blur. (We work for our audience, not for pixel-peeper anyway, right?)

To sell this effect I certainly would consider to have some depth based atmospheric effects working for you, so any change might be expected anyway.
(Depth-perspective or “one D” perspective [Sfumato] as it is called was used as well in the YouTube link above. Nothing new, but as effective as ever)

Good luck.