Making a Fast 3D CG Workflow for Promos
Impossible Pictures Relies on Softimage XSI to Build a Twister for Discovery Channel
Joel Pilger: It’s standard procedure when we’re doing projects that have a heavy reliance on 3D and CGI and have short time frames. This is the workflow and pipeline that we’ve almost unknowingly developed over the years.
Describe your concept-development process.
Obviously, it’s fruitless to brainstorm and come up with concepts that you go present to the client when none of them are achievable within their budget or timeframe. It’s a waste of energy, and ultimately it will just frustrate the client on the other end. We immediately do the initial brainstorming and concepting and let our imaginations go as far and wide as we can, and then we rein it in and say, OK, how would we tweak this, modify this idea to make it doable. What would that look like? What are the steps to get it there, what efficiencies or processes can we take advantage of?
One of our animators came up with an idea: let’s do one of these pieces with a tornado. And we all went, “Wow that’s great. But there’s no frickin’ way to do a tornado in two weeks with full CG.” But then the animator says, “Well, but, you remember there’s this tool we haven’t exploited yet, and if we do this and limit the camera and limit the environment, placing certain parameters on them.” And then we present that to the client. And the client says, “Wow, a tornado. I wasn’t expecting that – but I love it.”
At this point, is anyone sitting at the computer doing any work at all?
It’s sketching little thumbnails and [communicating] verbally.
It’s a real awareness of what your limits are.
It’s very much familiarity with myriads of problems we’ve seen over the years, so we know what it takes. We haven’t done that specifically, but we’ve done things that let us extrapolate and predict how realistically some of these challenges might be overcome.
And that concept comes back as a simple sketched storyboard?
Yes, it’s not even electronic at that point. It’s a thumbnail that’s done in pen or pencil to get something in front of the client before we spend 10 more minutes entertaining an idea that may be a complete waste of time. “Now I see the gist of it and my reaction is yes – I still like it even though our camera is going to be somewhat fixed, traveling in a certain way. I like all that. Let’s go for it.”
And then you start the previs. How elaborate is your previs?
It varies, but the biggest questions that we’re answering are what’s the camera doing, and what are the major compositional elements. In the example of the twister, we built a gigantic cone and had it twisting so we could see the speed of the tornado. How far away is it going to feel? How fast are the telephone poles going by? A vehicle that’s careening out of control almost collides with us – at what point in the story are we going to see that?
It’s all about making it as cool as they can possibly be, but it’s a promo. It’s a banner for the Discovery Channel. We can’t get too far off track. It’s ultimately got to get people’s attention and drive them to digest the tune-in information for the shows they’re promoting. We don’t want to get so fancy that we’re losing track of that, and the previs lets us establish the amount of action we’re going to tell.
And this is all happening in Softimage XSI, so you’re already in the tool you’re going to use throughout the 3D process.
Yup. Sometimes we’ll switch gears. Even if it’s really a Flame project, the previs will be in XSI. But in this case, since these assignments were predominantly 3D, we were able to take the previs that was there and then continue [in XSI] with the work we had already done, leverage it and build on it and add detail.
What would make you go to Flame earlier in the process? Lots of live-action elements?
Yes. When we’re doing a full green-screen shoot and have a person in multiple shots, we would previs all the shots and show what the person is doing. But that [XSI previs] would drop into a timeline in Smoke or Flame and that would become the bottom layer that we will place shots on so we know we’re on track with our timing and our audio.
Is there a certain class of artist at Impossible that you would put on this kind of project because of their speed?
Yeah. We’ve learned to be really, really fast, and we have pretty darned good ideas. It makes a huge difference when working with a CGI artist who is very confident and has jumped through lots of different types of hoops. Lots of times we feel like we’re throwing that kind of artist under the bus. “We’ve got about two weeks. We can do this – right? You can pull this off?” And away that person goes. It’s a high expectation, and it is very demanding, but if you have somebody who’s very fast and has a fast machine and a really good tool, like I think XSI is, all those things come together and make it possible.
When that previs is approved, in this case, the same artist went back to the same XSI system and started adding the detail – but used the original previs instead of starting from scratch.
Yes, it makes a big difference. You already have the basics. We have Maya, but we prefer XSI because of the integration with Mental Ray. If we can be previewing renders as we’re working – constantly, constantly, constantly – we identify problems and exploit opportunities much earlier. You could be just a few steps into refining the previs, and very quickly you realize, “Hmm, we’re going to have to light things a different way.” That’s better than getting way downstream and starting to do test renders and saying, “Oh, crap. We have to start from scratch because the lighting isn’t going to work.” With Softimage you’re seeing a full render as you adjust and tweak and move. It’s constantly updating. The artist does not have to pause and stop what he’s doing at all.
You’re also staying in XSI as long as you can.
In these instances it proves to be advantageous. Normally, the artist would get to a certain point and render passes and hand those passes off to the Flame guy or the Smoke artist. But we have the XSI artist pre-comp and make what I call “render chunks.” He’ll render five or six or eight layers and pre-composite them in the XSI compositor to see, once this goes over to Flame and we stack it all up, what it’s actually going to look like. Rather than just feeding all those things to the Flame guy, the XSI artist can say, “Hmm, not quite right. I need more opacity here, and that shadow is not coming through.” He’s fixing a lot of those problems as he’s working. He’s got a compositing tree right in his project to test and test. And then he’ll render those multiple layers out as a chunk.
And the quality is good enough that you don’t have to redo it in Flame.
Absolutely. “Here’s the eight layers that make up the twister, here’s the environment, here’s the background sky, here’s the road, here’s the car when it does its little thing, but I already precomposited the dust that the car kicks up,” things like that. When the Flame guy gets going and brings it all in, it’s pretty darned fast. We do the compositing in half a day on a lot of these things.