Scientist Shootout Allabouts.

by - 2012/06/08 22 Comments Artwork

The biggest focus this week was trying to get the shootout scene in the scientist’s tunnel completed. As with everything, it was a crazy balancing act of figuring out the right ratios of feasible render times to sample rates to noise reduction.

Just figuring out how to light realistically in cycles has been one of the biggest challenges. Cycles is incredible for smaller stuff, but once you get up into massively complex environments, it starts to slow way down, and you have to start throwing more and more cheats in there to make things really work. We know what we’re going for, but sometimes it’s a bit like learning a new language to figure out how to make the tools do what we want.

That said, we must be learning or something, because everything is looking better and better.

The big unknown right now is trying to guess how much of the time we’re spending is due to learning/inventing the workflow, and how much of it is indicative of how long it actually takes to finish a shot (in which case, we’re doomed!) The scene we’re working on now is taking 2-3 times longer than we’d originally set aside for it- but it’s also one of the most complex scenes.

The scene is the one where the scientists have to take up arms and defend the bottleneck between the old church and the dome.

1.) The Arminator

(this is from 2-3 weeks ago, and not indicative of any of the recent stuff- but it’s all I got on hand!)

The scene starts with the scientists putting their arms into the ‘arminators’ (we never say this word in the film :P). The shot has been tracked, the arminator built and rigged, the animation matched to the scientist’s movement, the guns tracked onto them, and a dozen other things.

Roman has spent a couple days working on this shot alone, extracting a perfect key and working with the new masking tools.

Kjartan added some liquid nitrogen steamstuff coming out of the holes (man, it’s a sci-fi. you gotta have liquid nitrogen when people pull canisters out of canister holder devices).

2.) The “Horrors of War”

(as with everything else, still WIP. And oh my goodness yes we know there is sun on them and not on the tunnel)

The gratuitous slowmo shot. Dan pulls the gun out and stares at his hand as the gun unfolds from a compact bundle into the final expanded weapon as he nervously walks from the arminator to the makeshift barrier. This one is a doozy- Sebastian had to track both the background (which is barely visible/out of focus), and 3 unique guns. Even after it was tracked, Francesco had to go in and hand tweak all the different axis for the main gun, since the tracking device had some high-frequency wiggle (due to the lack of mass).

The main point of the shot is that the scientists are out of their element, having to put on these absurd weapons to face off against absurd robots, so making the gun as overbearing a presence in the shot as possible was important. It expands outwards over the course of the shot, and at the end he’s holding it right in front of the camera, almost blocking everything.

A few weeks ago I modeled a new scientist gun specifically to cover as many tracking markers as possible. The tracking devices all had balls that extended pretty far away from the actors hands, and obscured faces and such, and to avoid having to painstakingly paint out all the balls (a bit will still be necessary), this gun unfolds to have extended pieces that will occlude the trackers. Here it is unfolding (although probably 8 times faster than it will in the film).

Kjartan did some great shader work, and I went in and tweaked some of the colors and added decals. I think it’s a good looking gun! This is exactly the type of thing cycles is great at rendering.

Additional complications to the shot:

  1. The white costumes! This means greenspill, though not as much as you might think, since the lighting folks were super great about avoiding it. The real problem is just the fact that the costumes are bright and clean and white, whereas the backgrounds are all dark and dirty, so no matter what we do the costumes have a tendency to look pasted on. I suspect once we start keying/compositing the other actors, with tonal ranges more similar to the background, we’ll find it a lot easier.
  2. The tracking rigs! This is a handheld shallow DOF shot in front of a void of green- how do you track that? By putting lots of keying markers in there! Sebastian put several tracking stands and ladders in the background so we could get a good track, but now we gotta take em back out! These masking tools are gonna get a workout. Roman’s been doing super amazing cleaning up shots, but he’s going to need some serious help if we keep going at this rate.
  3. The characters are walking into sunlight that doesn’t totally match what would actually happen in the tunnel, and it’s a bit tricky to reconcile. It’s also super yellow, which will work perfectly for the final grade, but in the meantime means we’re having to make pictures that feel a bit too yellow for their own good.

3.) Onward comes the horde!

Mmm. You can almost smell the hairspray. Obviously we’ll need to tweak some junk with compositing and all that (once we figure out colorspace and the like), but I totally dig this! Nice and kinetic.

 We wanted to have the gun fire something that wasn’t a bullet, and wasn’t a laser. Kjartan put together a great proof of concept a week ago showing a cool compromise. This quick electrical cloud.

Unfortunately, when we put it into context in the scene, we realized it didn’t scale quite as well as we’d like when the distance was more than a few meters, both because it was so heavily simulation based, and because it just plain looked a bit weird (giant electro-cloud for a few frames = awesome. giant electro-cloud flying across a room = a bit wonky). We rethought it a bit and were able to keep all the same elements, but made it a bit smaller/faster, like an electric potato gun.

And what are the scientists shooting at? These guys! These are the robots made by Kjartan and Jeremy a few months ago- we’re finally putting them to use! Jeremy animated a half dozen of em approaching the scientists. I love it- it’s like a tower defence game. You can really feel how the robots are approaching faster than they can be shot down. Now we just have to make sure the timing of the robodeaths and the electropotatos match up nice.

And there’s another view.  I set up a camera to render an alternate view of the scene for cut-away options. Once we have the lighting/texturing finalized, if we ever have a night where the render farm isn’t busy, I’ll throw it up to give us a few more options (stuff like these are freebies! A couple seconds setting up a camera, a night of rendering, and BAM! More robot action shots.

So all in all, things are going well. TONS of experimenting. Days can go by of just troubleshooting why things aren’t working the way they should. Everything is taking a bit longer than expected- but we’re still starting out, and for the first time I’m really starting to see stuff that show the promise of the final :D.

  1. 3pointedit says:

    Props mass, I think I mentioned that during shoot. Think you may have been busy ;)

    How will you get the arms to stay within the gun rack during pull out?

    • Manuel says:

      They had a green box with 2 holes in them for the actors to put their arms in and get the tracking devices on their arms.
      So since the physically pulled something out of a box with 2 holes, i dont think this will be a problem(given that i understood you question correctly)

  2. Tycho says:

    Waa :) impressive : just one question : why is the gun made in canada ?

  3. 3pointedit says:

    Hey Ian is there anything new in the VSE? I gather you have all the animatic in there to be replaced, does the live footage require much trimming to fit. Is the VSE up to it?

    • ian says:

      Actually, it ended up being the other way around, more or less. I edited the footage without even looking at the animatic, to make sure we got the best edit out of what we shot, then once that worked I started dropping the animatic back into places that were full CG, or whos timing were dependent on CG elements.

      And Campbell made a great new feature! Alt+drag a clip roughly between two others, and the VSE just makes room and plops it in. This was sometimes straightforward before if you already had a gap between the shots, but if you didn’t, it was a whole mess of clicks to make room and get the new shot all in between cleanly.

      • blendercomp says:

        Wow!!!!!
        This single feature will boost cutting by at least a factor of 5! :)))
        Would love to see *more* VSE improvements though!

      • blendercomp says:

        I’ve just tried this and it didn’t work as described here. Has this been committed to trunk or is it only available in one of the custom studio builds? Just wondering…

        • LswaN says:

          I think the mango team has been using the tomato branch from last year’s summer of code as a place for their in progress development, iirc. So maybe the feature is only there for now.

  4. Aaron says:

    With this shot being, as you mentioned, one of the most complex. You’ll probably experience a huge learning curve, and working on the following less complex scenes could probably go faster because of it. So, things might balance out. :)

    It’s really impressive and exciting seeing this come together more and more. Plus, it’s good to see another post from you Ian. I understand you’ll be very busy, though. So don’t take it as a *poke* or *nudge* or anything. :)

    Rock on guys!

    • ian says:

      Yeah! I’m sure hoping so! Everything certainly seems to be going faster. I’ve just become wary of the “Things are only slow because of ______, once we fix that, everything will go fast!” excuse over the years.

      Hahaha, yeahhhh- we’re all supposed to do a blogpost a week, but I’m not sure I’ve been keeping up these past few weeks.

      • Aaron says:

        It’s understandable. So no hard feelings if you miss posts. I’ll just shake and quiver it off, alone in the darkest corner of my garden, while wearing a hat made of twigs and bad-mouthing you to my squirrel friends… o.O

        I felt like a random jaunt into the realm of weird. :P

      • Kirado says:

        You guys could always ask Freddie Wong for some effects help. Maybe he would be up for the whole opensource thing?

  5. Nate Wiebe says:

    Is any work being put into a new keying node? Or are the old ones being used?

    • ton says:

      Several attempts are being made to improve keying. Just wait for experiments to give results we like to share :)

  6. J. says:

    What a lot of work! So no Euro2012 for you guys! :P

    Btw, I though that Brecht had come up with a more or less reasonable color workflow for the time being? (floating point REC709 EXRs)
    You still seem to be using ACES EXRs directly in the trunk version of Blender.

  7. Ben says:

    Kudos for sharing all the problems you guys run into – I learn a lot more from that then from the “look at this cool explosion” posts (nothing against cool explosions, by the way).

  8. Ace Dragon says:

    As I mentioned in the comments of the last article; It seems like now would be a good time to start seeing the first commits from Brecht regarding the performance and sampling improvements as mentioned last month on this blog.

    It is almost a sure thing that he wouldn’t be able to give Arnold-like speeds to the Cycles engine any time soon, but perhaps the type of improvements that can only be done on the CPU-rendering side would give you a major boost in terms of increasing the general performance in terms of the amount of time needed for each sample along with reducing the number of samples needed to get a noiseless result.

  9. Kutato says:

    Imagine also that I’ll make a short (less than budget, of course, because I’m poor: (), and reading these advances fills me with a little terror. Also will base my progress in learning along the way to solve problems and compensate with alternative (but always using open source). mind, and as they say there: the important thing is not the goal, but enjoy the journey!!

  10. Florian says:

    You probably have considered this yourself, but I’m really interested if there’s any thought/work being put into a scalable, distributed, community-based render farm (or rather render network) à la “SETI@home”, where you could have individual frames (or even parts of frames, what with the render boundaries) rendered on volunteer clients all over the net. If this worked properly it could save you lots of render time without additional cost. And once established, such a network could be used for other projects, too. I imagine it could be organizied similar to crowdfunding, so that people could apply with their project (including blendfile), and client owners could vote on which render project to support, or choose individually whom they would like to support with their renderpower…