VFXTalk: Can you give us a brief run-down on what you did on Constantine and what previous experience you've had?
Dan Cayer: I supervised the compositing of a number of sequences for Tippett Studio on "Constantine", including the Exorcisms and Seplavites Attack sequence. I've been working on features for almost 5 years with my first show being right here at Tippett on "Evolution". Prior to that I had done some 2d and 3d work for an independent film, museum exhibit, and a court case, but was for the most part hired out of school. While here at Tippett I've worked on 9 features and about 8 commercials. "Matrix Revolutions" was the first show that I was given the opportunity to lead on, and "Constantine" was my second.
Matt Jacobs: I was a lead compositor on "Constantine" for the Hell Freeway and Vermin Man sequences. I have been with Tippett Studio for 7 years working on a number of films in many capacities. Coming to film from a background in photography, I initially worked as a rotoscope and 2d paint artist. Having composited on several shows I was given my first opportunity as lead compositor on "Men In Black 2", then later on "Matrix: Revolutions."
VFXTalk: What was by the far the biggest challenge in term of visual effects in this movie, and how did you overcome them? Was there a different treatment? A new techique? Which sequence was the hardest to work on, and why?
Dan: I think several aspects of the movie could get strong consideration for the most challenging. For me, the most challenging shots were the Mammon Exorcisms. In part this was because we were trying to make a demon stick out of the actress' stomach (enough said), but also because the actual look of the effect was a moving target. Over the course of the show, while trying to give the director a look that he felt worked in his movie, we ended up recreating the effect differently on several occasions. And of course, when you change the look in one shot, you have to change them all. Some of the changing aspects were the amount that Mammon would protrude, the transparency of the skin or the amount that we actually saw Mammon underneath, the amount of bruising and damage that should be visible, and how tight the skin should look as he moved underneath.
We created the shots through a series of layers that were combined and balanced in the comp to look like stretching skin. For certain reasons this technique worked well. For others, it made the constantly changing look more difficult. Adjusting for things like skin transparency and visible bruising was easier due to the fact that it was being balanced in the comp. The difficult part came when animation changes were requested such as the amount that Mammon protrudes. At that point many layers were already balanced in relation to depth and proximity of Mammon under the skin. We use special depth renders to calculate the distance relationships between the two. When the animation and the renders changed, most of the comp had to change as well.
It seems that on most movies there's always one shot, or sequence, or effect that ends up getting worked on for most of the show. This effect was just that for me. It was challenging to do, but in the end the director was happy and ultimately that's our job.
This is one of the long shots of Mammon in Angela's belly. It wasn't quite as tricky as some of the close ups, but was still assembled with all the layers mentioned above.
Matt: For me the biggest challenge was the creation of the "Hell Freeway" effects. In each of the shots we had a plate shot against a green screen, various CG layers for the city, CG cars, CG freeway extensions, a roiling, burning sky, scavenger creatures, and multiple layers of airborne particles like detritus and the sheet dust we see peeling off the hard surfaces. Added to that we used a lot dust and smoke elements we have in our library to add more dust and atmospherics to the shots. Between all those various elements and the multi-streams of the CG renders we had a lot of layers to deal with.
VFXTalk: What sort of freedom were you given in creating the looks for the sequences you were in charge of?
Dan: That pretty much depended on the sequence, how much concept art we were given, and how much of a clear picture my visual effects supervisor already had in his head. Some sequences such as Hell Freeway had lots of concept art. In a case like that we must do our best to match the feel of the artwork. Other sequences such as Seplavites Attack or the exorcisms had less or none.
Craig Hayes was the VFX Supervisor at Tippett. He's really good about steering us in a direction that he likes, but keeping it loose enough so that things he may not have thought of can still happen. The less clear everyone is on what the result should look like, the more fun it becomes. In those cases, Craig might give me or other compositors suggestions, but sit back and see what we can come up with. That was pretty common on this show because we were often dealing with layers and layers of renders that were designed to be mixed together. This gave us a lot of creative freedom in the compositing stage.
Matt: Initially we had story boards to work with, and through our visual effects supervisor Craig Hayes we weeded out the elements we would need to create the Hell Freeway effects. When it came time to put all the elements together though I would say I was given a lot of freedom. Many of the elements needed a lot of massaging to make work, that's where our artistic license comes into play. Obviously there is a lot of back and forth at that point as to what is working and what isn't.
It's the beginning of the production that is most interesting where we are fleshing out a lot of the final look in compositing.
All of the lighting ray effects in the Seplavites Attack sequence were done as 2d elements in the comp.
VFXTalk: As lead compositors, how many shots did you do yourselves, and do you get to assign the shots out to other compositors?
Dan: The number of shots that I composited myself varied from sequence to sequence depending on the size of the sequence as well as the amount of R&D that needed to happen in the comp. I led some sequences that I didn't do any shots in, some that I did a few, and a couple that I did all or most of the shots. I met with our production coordinator regularly to assign the shots. We worked out who was best suited for any particular shot and what worked in terms of schedule.
Matt: I really don't know how many shots I did through to completion. Yes, I did get to dole out the shots to the compositors. One of my main considerations was the massive number of elements in these shots, and continuity was extremely critical to me in this highly stylized environment. Therefore, many of the elements were treated like an assembly line where one person would take control of an element, make sure it looked right a shots and then pass off a stack in shake to the comper who was doing the final assembly. It might sound like this approach is taking some of the creativity out of the individual artists hands but I think in the end everyone was satisfied with the challenges each shot presented them.
Here, an animated texture was made in the comp as a reveal to the green demon, Balthazar. From that, renders were once again generated for displacement and shading to be done in the comp.
VFXTalk: How many artists did you lead and did you get a pick of the sequences or are they assigned based on your strengths?
Dan: I think there were 12 different compositors that worked on the show, although not all at the same time. Matt and I were able to decide which sequences we wanted to work on. We both had interest in certain ones and then the others seemed to fall in place due to their similar looks or techniques. He did the Hell Freeway shots so it made sense for him to do the Hell Hydrotherapy shots. I did the first Exorcism sequence so it made sense that I do the Mammon Exorcisms. We've worked together on shows in the past as well as with our Production Coordinator Athena Portillo, so it was easy to divide up the work.
Matt: Throughout the production, Dan and I had 10 artists that we would share between sequences. As far as leading the sequences we were initially assigned to our sequences, then later on in the production when we were awarded more work, Dan and I sat down and figured out which ones we would like to lead. Much of the decision was based on individual interest in a sequence and the schedule of the shots.
VFXTalk: How do you find the crossover from commercials to features and back again. In terms of compositing what are those little areas that you like to pay more attention to when doing a feature composite as opposed to a commercial?
Dan: Although I have worked on almost as many commercials as features, in terms of time spent, I'm mostly a feature compositor. I think the most enjoyable part for me when working on a commercial is that my render times get much faster. In terms of techniques and attention to detail, I've got to say that I treat them the same. The only difference might be when heavily processing something or scaling an element up I may be a little less cautious, but mostly it's the same mind set.
VFXTalk: How often have you busted a gut putting nice subtleties into a shot only to see them disappear in a heavily graded print and how do you go about managing this aspect in your work i.e. when do say "hold on a minute why am I wasting my time doing this, it'll never be seen in the print'?
Dan: Every show is different, but it usually happens in some form whether it be when the film is released or when it goes to DVD. I must say that overall on Constantine I was pretty happy. The best way that we can avoid that is to make sure our shots fit in to the sequence and represent what the director is looking for. We're constantly doing sequence continuity checks to ensure that all of our shots are in line.
The people here who handle the scans and initial color timing do a great job matching color and density with the rest of the movie. At a certain point it is out of our control, but the closer we are to what the director wants the less likely that will happen.
Matt: All the time. One of the things that always strikes me is just how fast our work flashes by on the screen. By the time it's shown in the film, much of it has been left on the cutting room floor. That is one reason that we pay very close attention to our cut lengths and handles. As far as grading the film goes, this can have a huge impact on the final result and yes, much of the subtleties can be lost. When we first started to view the Hell Freeway shots on film they looked very flat and very different from what we were seeing on our monitors. Craig and Mike Fink had several conversations about this subject and it was something we monitored closely. Ultimately the shots went to a digital intermediate for color timing and I would have to say I was happy with the results in the final film.
VFXTalk: What other talents make up a good compositor? What other skills should they be fluent in? Photography? Color theory?
Dan: There are different aspects to being a good compositor including aesthetic, technical, and communication skills. Photography and color theory are certainly great places to start. Knowing what happens through the camera and what kind of decisions might be made on set are great to know. Color is something that you'll think about on a daily basis.
Beyond that, compositors need an understanding of light and shadow and a great sense for composition. Paying attention to details and subtleties in everyday life helps you translate that into your work. Watching the way things move and interact will make you a better compositor.
On a technical level, I think the most important skill is being able to pick things up quickly and to have a good understanding of why things work. There are tons of techniques and programs, most of which are proprietary, that could be learned. Knowing them all isn't the important part. It's more about understanding why they work, which will help you learn more quickly.
One of the most important skills is the ability to work with people. Be willing to compromise but able to communicate your opinion. You can be a great artist or technician, but without the ability to work cooperatively, you won't get much done. A good compositor can take direction and translate it into a successful solution to the problem.
Never become married to anything you do because at some point it probably will have to change.
Matt: A strong background in photography, painting, drawing or any of the arts is going to make you a better compositor. Also observing the real world is key to making better pictures. You can learn a lot from just standing outside and making mental note of the things you see. On the other hand, I always find it interesting when I see a natural phenomenon in light and atmosphere that I know if I were put in a shot it would look completely phony.
VFXTalk: When did you decide to start in the visual effects field? After high school, college, or earlier?
Dan: I decided to pursue effects work after I had started college. I had always been artistic but wasn't sure if that was the direction I would take. After spending a year not enjoying Electrical Engineering as a major, I up and moved from New Hampshire to California. Looking back, I think I had watched a few too many episodes of "Movie Magic," but things all seemed to work out once I got here.
Matt: I've always been interested in photography and have being playing with it since high school. Visual effects was just part of the evolution of my interest in photography, and not my aversion to getting a job where I had to wear a suit.
VFXTalk: For those scenes where Keanu is in Hell, are these motion control shots and was the data translated to the 3d software or are they all matchmoved using software? If so, what is Tippett Studio's preferred matchmoving software?
Matt: We use a combination of Maya Live and some custom solvers.
VFXTalk: Were there any wire removal shots, and did you use any plugins to remove them?
Dan: At Tippett we have a dedicated paint/roto department that handles the paint fixes we know will be a problem ahead of time. There were a few wire removal shots that I can remember, like when they suspended the mirror above the possessed girl's bed. There were certainly many, many other paint fixes above and beyond wires. Our paint/roto department uses Combustion and Commotion for their paint work.
Here the demon and background were comped as usual, then the cracks were added on top. A series of renders were generated to displace and shade the original plate/comp.
VFXTalk: How much has the line between comping and 3D blurred for you? As a lead do you have more control on what the 3d department gives you?
Dan: At Tippett we have lots of useful tools that could be called "pseudo" 3D that help with compositing. We use depth renders for anything from placing photographic elements to actually doing supplementary lighting in the comp. The 3D and 2D departments are definitely separate entities at Tippett, but certain tools and techniques can have a lot of overlap.
As a lead, I have the opportunity to sit down with other departments ahead of time and plan what layers 3D will give the compositors. Once shot production begins, any suggestion from any compositor that makes a good case for a particular element or render certainly is welcome.
Matt: The line has definitely gotten blurred betweent 3D and 2D. We have a lot of custom plug-ins for shake that we use with Maya to get tracking information. Also we have a wide variety of tools that let us do pseudo lighting within the comp. We're constantly looking for ways to mimic things in 3D that we can do in the comp since the turnaround is so much faster.
VFXTalk: As lead compositors, how closely do you work with the lighting and render people?
Dan: All compositors are constantly in communication with their lighting TDs for any given shot. Every show is different, but on Constantine, many looks were created in the comp using numerous layers rendered by the TD. It is up to the compositor to make sure that he/she gets all the layers that are needed from the TD.
As a lead compositor, I work closely with the lead TDs as well as Paint, FX and the CG supervisor to establish what general layers will be needed for a sequence.
Matt: We have to constantly be communicating with all the other parts of our pipeline, especially the TD department. Much of that communication is telling them what is working and what isn't and seeing where best to fix it. If we can we will adjust things in the comp because the turn around is much faster.
VFXTalk: What's the "accuracy" of the TD renders compared to what the shot should look like? What do you guys get in production? How far do they dare to go?
Dan: Accuracy is a difficult way to describe it. In general, the TDs match color, black point, and white point of the photographic plate for all elements. If the shot consists of simply a creature over a background, then you could say they come to us "accurate". However, on Constantine, I don't think there were actually any shots like that.
In Hell for instance, many of the interactive lighting effects as well as practical dust elements were created in the comp. At that point, the once "accurate" renders need to be adjusted. In addition many of those shots had no photographic plates other than the actor, and whenever a change is easiest to make in comp, it usually is.
Another group of shots that were heavily adjusted in comp were all of the exorcisms. The skin effect of both the girl at the beginning of the movie and Angela at the end were achieved by combining many rendered layers in the comp. Each of those layers were rendered to match the photographic plate, but once they are combined in the comp to look like skin, all of that must be readjusted.
When we are heavily altering the layers rendered for us, all cards are on the table. It is something that is predetermined by the leads and the supervisors to be the best way to achieve a desired look, so it isn't a big surprise when the comp looks different than the straight renders. If a shot is pretty much a creature over a background, then that is when the renders should not be tweaked too much. The line of communication is usually good enough so that no one is too surprised with what ends up in the final comp.
Matt: We beat up the TD renders quite a bit to create the final images. Though they do give us a solid foundation to work with. When you're working on such a stylized sequence as the hell freeway sequence you have to expect that 3D is going to take it so far and that comping is going to have to smack the images around to get them to work as a whole.
Although this is a close up of a cg creature, lots of massaging was done in the comp.
VFXTalk: Do you have any input or do you just work with what you get?
Dan: The compositors have a lot of input as to what they need to achieve a shot. As a lead I try to ensure that by the time the shots are in production we know what those layers are. If a compositor has a specific request then he/she will ask their TD. If the request is less than straight forward then we can all talk about it and see if it is really necessary. The key is to get what it takes to make the shot successful and not spend time working with something that won't do that.
Matt: Sometimes we have to just deal with what we get and that can be in any area. A lot of the time the plates aren't ideal for end result. That's one area where you get what you get and you've got to be prepared to make things work.
VFXTalk: The Vermin Man looked cool. How many layers was he rendered out in? What were the main challenges with getting the insects to "read" propperly on the screen?
Matt: The Vermin Man was a sequence where we tried to get the TD renders to be as close to the final elements as possible. The renders were multi-streamed out into their diffuse, specular and reflection passes which gave us the flexibility to change what we had to, but because of continuity I felt it was essential that when we could we should have the TD's make the necessary adjustments then we could dial in the last 5%.
VFXTalk: I read somewhere that the CG artists had to limit themselves to 50,000 models when creating the illusion of millions of bugs on the screen. Were there any compositing tricks used to enhance that illusion, or was he just rendered out in parts of 50,000 insects at a time?
Matt: I don't know exactly what the particle count was on the bugs, but one thing we did to add some meat to the effect was to do some fractal blurring of the bugs then comp them back over themselves. This was a great trick one of the compers came up with that made the cloud of bugs more dense and at the same time broke up the edges with some high frequency noise.
VFXTalk: Did you enjoy working on the film?
Dan: I enjoyed working on the film a lot. Craig is a great Supervisor to work with. He asks for a lot, but puts a lot of trust in the compositors as well. Our in house Producer, Production Manager, and coordinators were great to work with. Most importantly, all of the compositors were very talented and willing to put forth a great effort.
Matt: Yes, very much. We had a great crew working on this show not only in the comping department but in all the departments. The work was challenging and in the end I was very pleased with the results.
VFXTalk: Finally, Red Bull, coffee, or something else?
Dan: Coffee. .....And 2 little kids at home waiting for me to get there.
Matt: I try to get some exercise to get my brain working but I do drink a lot of coffee. Everyone enjoyed it when the producers would get the local coffee shop to come down and make us whatever we wanted at break.
In this shot, goo smear maps and breathe were generated in the comp based on the places the demon touched or came close to on the mirror. Any imperfections or streaks we could get him behind helped to sink him into the mirror.
This shot got the same treatments plus a green screen was replaced with the city out the window.
The WB, Village Roadshow, and Vertigo logos were meticulously detailed and defined in 3D while fx elements were employed in the comp to appropriately add atmosphere and heat distortion as well as particulate matter and dust streams.