The Process

I finally got the entire Tetons region loaded into one giant DEM file. Hooray. The process is so arduous, I feel compelled to write it down, for the cathartic value if nothing else.

SECTION A:
Open Firefox
Go to http://seamless.usgs.gov/
Click to view and download data
Zoom in on your region
Under Downloads, under Elevations, de-select 1” NED and select 1/3” NED.
Click the multiple template download button.
Set the template type to 7.5 Minute CONOS

SECTION B:
Click to select one small square on the map
Wait 3 seconds for the map to update.
Click submit

A new download window appears.
Click Modify Data Request.

Scroll to the bottom of a list of about 150 data types to locate yours.
Change the file type to GeoTIFF.
Click submit.

Click Download.
A new download window appears.
Wait 10-20 seconds.
Save ZIP archive.
Close second download window.
Close the first download window.

SECTION C:
Open the ZIP archive.
Locate and extract the TIFF file.
Delete the ZIP archive.

Open 3Dem.
Choose Open File
Select GeoTIFF
Locate and load the TIFF file
Choose Save As ASCII DEM
Save as a new file.

In 3Dem, choose Open File
Select ASCII DEM
Open all DEM files saved so far
Check landscape for remaining holes.

Back to Firefox
On the main USGS map, Click Clear Selections
Locate the grid cell corresponding to one of the missing sections of landscape.
Return to section B. Repeat. Many Times.

Finally done? Not quite.

SECTION D:
Now load the entire, giant landscape into 3Dem.
Change the landscape colors to run from black to white.
Resize the landscape to sufficient size for my displacement map.
Save the landscape as a BMP file, the only lossless format I can comfortably work with.

Now, fire up Photoshop.
Open the BMP file.
Save it as a compressed TIFF.

Finally, I'm ready to load it into Lightwave and actually do some work.

The Usual

I spent all last night downloading USGS data for the Tetons and the surrounding region. It's a slow, nit-picky and laborious process but I'm happy to have access to the data.

I got 75 out of 80 'grids' downloaded successfully, but that left me with 5 'holes' in the landscape. I identified the missing cells, downloaded them, integrated them into the others, and 3DEM - the app I'm using to work with USGS DEM files - crashed.

So I went to bed. A typical day here, I'm afraid.

3DEM rocks, by the way. I'm sure it's my fault. I'll try again today.

New Texture Samples

I thought I'd give an example showing why I need such ridiculous texture sizes. Here's the first frame of animation with the old textures:


And here's the same frame, using the new textures.


Here's a full earth, rendered with the old textures. At this size, they look just fine. But when the camera comes low, they weren't cutting it.

Cheating on textures

I'm working on one of the opening shots of the animation right now, one that features the classic 'pull out into space shot.' You know the one, the camera starts in looking down on a landscape and then pulls rapidly back to view the planet from outer space, usually passing through some convenient cloud banks on the way, to hide the joins in the footage.

The joins are necessary because any textures of sufficient resolution for the close-up would be ridiculously huge if you expanded them to cover the entire planet.

The pull-back was really suffering from the low-resolution textures I was using so I found a terrifically huge photo composite of the earth's surface. In fact, it was so big I could barely open it in Photoshop and Lightwave just threw up its hands.

I made a 50% version - still 10,000 x 5,000 pixels and rendered some tests and it looks better but, dang, that's a big texture.

My plan now is to cheat. Since only half of the earth is ever seen illuminated in this shot, I'm creating a version of the map that only covers half the earth, then repeating it twice. So if you looked where India should be, you'd find North America, instead. But you can't see that side so I'm gonna go for it.

I've actually done the same thing with the primal earth; to increase the apparent resolution of the texture, I'm just using it twice.

As the camera passes behind the earth, I want the nightside city lights to be visible, then slowly change into glowing volcanoes as the camera comes around the other side, at which point we've gone back in time to primordial earth.

I wasted a lot of time trying to get the nighttime 'light' maps and the daytime texture maps to combine on a single earth before I gave up and decided to do it in post.

I rendered a pass of the earth, then a pass of the earth with no sunlight and the light map applied, then a final pass of the primal earth. All three shots used the same camera setup.

Finally, in After Effects I placed the night time pass over the daylight pass, in Lighten mode, so the lights only show where the daylight pass was in shadow.

I laid the primal earth pass on top of those two, again in lighten mode.

I clipped the daylight pass while the entire planet was in shadow, so it is hidden before daylight reappears at the end of the shot. And it was as painless as that.

Fix it in post, that's my new motto.

More planet renders

Another day of fussing about and I'm getting pretty close to something I'm happy with. Here's a render of present day earth. That's North America at the upper right.


And here is my latest version of primordial earth. I dunno - maybe it needs more volcanoes. Still, it's at least starting to look like a planet.


I used the same basic setup for both renders. In fact, I used the same models. I just changed the maps and the atmosphere color to change our present Earth into its more primal self.

A new Earth

I know after my last post half of you lost all faith in me. Mark, you said, you seem to be totally lacking in talent and ability. That's a horrible planet!

Yes, yes it was. But I struggle ever onward against my disabilities.

I spent a couple of hours trying to fake up some good maps in Photoshop, but got nowhere at all.

It finally occurred to my to use Vue. I'd like to get some good out of it. So I created a simple landscape in Vue 6, and clipped off the bottom, giving me something like a group of islands sitting in the sea:



I applied a mountain texture to it, rendered it and saved it out for use as a map in Lightwave. Rendered up in Lightwave, it's a definite improvement over yesterday's effort, but the mountains look about 5,000 miles tall. A lot of that comes from the shadows rendered in Vue, but the texture is just too big.


So it was back to Vue to re-render the landscape with all the textures scaled down and with the sun directly overhead, so the mountains didn't cast those ridiculous shadows. Here's my second attempt, with some rough shading of the black oceans and a few spots of color thrown in for good measure:


And here's the resulting Lightwave render. Better. Closer. But still no cigar.


A little further tweaking of the lights, clouds and cameras, and we're actually getting close. That needs work, but it's 100x better than my first try.

Primal Earth, ???

I bob and weave like a savvy prizefighter, trying to stay out of trouble, trying to find a place on this project where I can land a successful punch. But I'm blocked at every turn.

Vue is too slow. I won't have storyboards for weeks. I can't figure out how to make slices of the earth.

So I thought I'd turn to something fun that I know how to do - planet making. I'm pretty sure I'm going to need a primal Earth for this piece, so I decided to build one in LightWave. First I jumped up on the web to do a little research on what the Earth looked like 3.5 billion years ago. (That's when the client wants the animation to begin.) And it turns out ... there's almost no information out there.

Remember Pangaea? That single 'super continent' that existed way back in time, before continental drift had spead the continents out to where we see them today? Turns out, that was only about 250 million years ago. That's only about 1/16th the age of the Earth, and long after life arose. Heck, there were dinosaurs tromping around back then.

That's not so very old at all. What about the other 15/16ths of the Earth's life? What did it look like then? Beats me. I can find vague descriptions of what was happening then, but what did it look like? Were there clouds? What color? Were there oceans? What color? No handy photo guides for me.

Starting from scratch, then, I created the most horrible primordial earth ever. It really sucks. I must start over again today.

Taking a slice

I'm currently struggling with the best method to create cut-away section views of the earth (like this one) which retaining maximum flexibility. I can 'deform' a landscape easily to conform to the actual Tetons, but I can't come up with a good way to mate the deformed top with the 'sides' of the section without leaving any breaks or seems.

I might have to render them separately, and "fix it in post."

Theft

I saw a PBS show on the Everglades tonight, and they did something pretty cool. The used a subtle split-screen, with real video footage on the top and computer animation on the bottom, showing 'above ground' and 'below ground' at the same time, and tying the animation in with the real world footage.

I liked it. I think I'll steal it. :)

Delays, delays

I've about given up on Vue as a major player for this project. It's pretty, but it's just too damn slow. This project is in HDTV 720p resolution: 1280×720.

I figure that high resolution frames will take on the order of an hour (or more) per frame. That means 60 seconds of animation translates into 75 days of render time. I don't have that kind of time.

In other news, I'm still waiting on the producer to get all the information together so I can really get started. He was flying up to Salt Lake City on the 23rd, to meet with the government liaison and the project geologist, but that's now been pushed back to the 25th. It's only two days, but the start date keeps sliding backwards a bit at a time, and the delivery date, mid-July, doesn't move at all.

I hate waiting.

Render Times

I've complained (at length) about Vue's speed. I thought I'd post a concrete example. Vue exports scenes in Lightwave format, so I created a simple mountain, exported it to Lightwave, and rendered it in both programs from more or less the same position.

Vue: (1 minute, 20 seconds)
Now, I'll be the first to admit that the Vue render looks better. But ... a lot of that is in the texture mapping and those texture maps were generated in Vue for import into Lightwave.

There are also some flaws in the exported Lightwave model. Again, that model was generated by Vue.

But look at the render times. 80 seconds vs 2.2 seconds. Lightwave is 36 times faster. There's a lot of room to tweak and improve the render. Lighwave will certainly do that kind of work. I just brought this object in, threw in a background sky, and hit render, without adding many niceties. What's going on in the Vue renderer that makes it so slow?


Lightwave (2.2 seconds)


e-on has a version of Vue which integrates into Lightwave (or 3DS Max). I went with the stand-alone on the assumption that a stand-alone rendered would be faster than a program running under the hood of another program. Looks like I was wrong. I can 'sidegrade' to the integrated version, but that's another $400, money I'm reluctant to spend given how unhappy I am with this software.

Vue makes me so mad!

Vue continues to drive me NUTS. It is piggishly slow on my P4 2.8GHz machine with 1.5GB of RAM and Radeon 9800 video card. But that's not the worst of it. Oh no.

Vue allows you to place all kinds of plants and trees and rocks and houses and what have you in your scenes. Click on the TREE button and you'll see all the species of trees show in this screen shot. And that seems great, but ... it turns out you don't actually have all those trees, or most of those trees, or even half of those trees. Only the top two rows of trees are actually accessible; click on anything below that and you're taken to the 'online store' where you get the opportunity to buy that species of tree.

What the fuck is that about? This is outrageous! The 'for sale' trees aren't even visually identifiable. You have to click on them before Vue tells you that you can't have them.

Some categories are worse; some don't have any accessible items at all, just row after row of objects for sale. I have no objection to offering additional online content, but the way e-on is doing it in Vue is unmitigated horse shit! And the actual selection of objects actually included with the software is insultingly small.

Glad I got that off my chest. I'll end with a pretty picture of a maple, one of the few trees Vue actually ships with. Enjoy!

First Light!

My first render with Vue 6. It ain't much, but it's something.

I'm still feeling my way around in the program. To try and make the process a little easier on myself, I ordered training DVDs from Safe Harbor, but they STILL HAVEN'T SHIPPED THEM TO ME!!! ARGH!!!!!

So it looks like it's a slog through the ol' manual for me. :(

Today I received a new book, Windows into the Earth, co-written by Bob Smith (yeah, I know) who is the consulting geologist on this project. The book tells "The geologic story of Yellowstone and Grand Teton National Parks," so it's a perfect resource for me. Plus, it's filled with those little sciency illustrations of rock layers and whatnot which I plan to steal from liberally use for inspiration for my animations.

Renders with a Vue

After a lot of research, I settled on Vue 6 Infinite as what seemed like the best 3D software for such a landscape-intensive project.

The software arrived today and I quickly loaded it up and got to playing ... I mean, learning.

And now I'm really scared, because this puppy is sloooooooooooowww. I've been doing most of my 3D work in Lightwave 3D and its render times are an order of magnitude or better than Vue's. Even moving objects around in a relatively simple environment gets all laggy on my venerable 2.8GHz P4 machine. Fortunately, I'm getting a new workstation soon, which should help a lot.

Granted, Lightwave doesn't have nearly so many features for this kind of project - trees, grass, USGS import, what have you - but I'm pretty disappointed with the program's performance so far. I'll have to figure something out or I'm going to be in big trouble when it comes to final render time.

Fortunately, I should be getting a new workstation in for the project in the next week or two. Hopefully, that'll help out a bit.

The Tetons Project

I am embarking on my biggest 3D project in quite some time, an animated history of the geology of the Grand Teton National Park.

I've always enjoyed reading 'behind the scenes' journals about a range of projects. I thought it was time I tried one myself.

It will give me a place to vent my frustration at the 1,000 things that will try to thwart me in the next few months and, hopefully, end in a proud success and lots of pretty pictures.

I'll attempt to post regularly, but I'm terrible at that kind of thing so no promises!