Suiting Up: Revisited

As a class project a couple years ago we created an "homage" to effects like Iron Man and Star Craft 2 where a mechanical suit is being assembled over a person. I recently reshot the footage with our school's Sony HXRNX5U and have retracked it. This is a simple key and track test as a proof that the track is of a workable quality. I left the 2x4s in that I step onto so that it wouldn't look like my feet were slipping when I step up.

This is all part of a larger, semester-long project. In its first iteration, it was a single shot with a single design. This time around I'm wanting to give the students some variety. Along with the shot posted here, I (we, really. Forde Oliver and I filmed it) shot 4 additional optional shots. My hope is that those that want a larger challenge will edit together a sequence that shows more assembly animations and more complex mechanical modeling.

Along with more shots, I'm wanting to do more designs. In its first version, Tom had done the sole concept art. This time around, I'm hoping to have 2 or 3 versions for students to choose from. Should they choose, I'll even consider allowing the mixing and matching of concepts for more variety in the final work.

Stu Maschwitz's DV Rebel's Guide is possibly the greatest book ever. Smooth segue, right? Well it might not be a smooth segue, but it does get me right into something I want to talk about; shooting video. I've read many blog posts and forums on how to shoot quality video, but nothing is as thorough and concise as Stu's book. The biggest revelation for me so far has been about how to access settings in Adobe Premier for bringing your over-bright whites back into the usable colour spectrum and rendering it out to retain all that needed information. After doing a slight colour correct to my footage I was able to put detail back into areas that were too flat either in their whites or blacks. Wonderful.

I'm going to come back to the camera for a moment... the Sony HXRNX5U is a good camera. Not great in my initial shooting, but very good. Certainly better than my consumer level Canon HV30. I just don't understand the movement towards CMOS over CCD based cameras. Rolling shutter alone is a big enough reason to never use them. Tracking footage with rolling shutter issues isn't an impossibility, but it does limit how crazy you can go with your camera motions. Even on slight moves, if you're zoomed in, it becomes more pronounced. Shaky hands plus zoomed lens equals jello effect. No good.

Now the fact that this Sony pro camera has 3 CMOS sensors has seemingly given it the ability to shoot better in low light and the colours themselves seem to really pop, however looking at the final footage in Fusion channel by channel reveals a very noisy blue channel. It could be due to the fact that we shot on greenscreen and the solid colour green was messing with the camera's ability to cleanly "capture" other channels, but it seems to be really bad. I'm going to have to bring the camera home to do further testing on more "natural" environments to see what the results are.

Slight Update

Here's a slight update on the motion and modeling. I've shortened the length of the shot from 719 to 460 frames. I've also rescaled all the geometry to be more "real world" scale. The motion is better but now the end portion could be smoothed out a bit in 3D. This will all have to take a back seat for a few days as I have some school obligations this week. Until then, I'll post this video and get back to it after this weekend.

One thing I should mention... the image in the background is also temporary. I was just playing around with some 3D comping in Fusion. Seems to work well, and I'll be shooting my own HDRI for the background image when I find the time for that too.

Thoughts and comments are always appreciated.

World's Edge Stabilize

So after running into a friend and colleague, it was suggested that I stabilize the footage and re-track it. I've now done that and added more terrain. The stabilize is good, but far from perfect. I'll have to revisit that.

The reason I don't like stabilizing footage and camera tracking is purely a technical one. Artistically, it can look much better because any shaky hands can be smoothed out. The down side is twofold:

One, it can look pan-and-scan-esque if the stabilization is very large where you're reintroducing a move into something that's static or conversely, stabilizing something that's moving.

Two, by sliding the footage around, you're essentially "lying" to your camera tracking software. When you film something, your footage has a center of interest which is the center of your footage. If you stabilize that and shift the footage around, you're forcing your camera tracking software to move the camera all over the place, either rotationally or positionally, all in the name of "stable" 2d footage.

Thirdly (I know, I said two reasons), when you stabilize, if your footage has motion blur and you’re telling that footage, "no, actually I want you to be still," you'll have a perfectly smooth piece of footage that suddenly has weird blur going in different directions. A good example of this is that VFX shot in Gladiator where the tiger is leaping at Russel Crowe. I'm positive that the motion of the tiger was shot strangely and the motion blur that it had was unremovable and therefore it looks very strange when compared with the motion blur that he has.

If any of you readers have any comments on how it's progressing, I'd be interested to hear them. I'm not 100% sold on the composition of the land, but I do hope to finish that aspect in the next week so that I can start playing around in Mudbox and detailing them up.

World's Edge Proof of Concept

Here's a first pass of setting up the "World's Edge" project. Now that I've finished the camera track, I wanted to have a temp version of the synthetic camera move coming towards the real footage. The transition is wonky and now that I've put in all the work, I'm not entirely stoked on the take that I chose. Why couldn't I film it with a level horizon? Maybe I should choose a different version.

I've also decided to start posting video with Vimeo since it gives me much better quality than what Blogger has to offer.

World's Edge Proof from Matt Broeska on Vimeo.

Footage

Alright. Here's what all 3 of you have been waiting for: the footage. I've uploaded an image to show the full resolution of the footage that we shot along with a nice, low res video to show the camera move. Hopefully I'll have a camera track of this in the next few days, as time permits. The start of the camera move lingers on the ground for about 8 seconds before booming up to show the full reveal. This is to have something as we do a synthetic camera move rushing along the ground before seeing our post-apocalyptic-Tom standing on the edge of a cliff. Our key light was placed for some strong rim light from a low sun. In the next week or so I'll head out to Garbage hill to shoot an HDRI of the setting sun and comp that in.

This is going to be so much better than Terminator Salvation.

New Project

Okay I know. I haven't finished my face tracking project and I'm starting a new one. There are a few reasons. Regarding the face tracking project, I don't have final concepts and I still haven't shot the new footage. Both are pretty crucial for me to even consider moving forward on that project. In the meantime, I have at least finished my face model (as shown in an earlier post). At the same time, that face model can't be done since I need to know how I'm going to break it apart and open it up.

Regarding the picture that I've posted here, those are tracking markers I made the other day for this new project. Tracking markers are SUPER neccessary in a greenscreen shoot to help tell tracking software like Boujou how the camera is moving. It knows how the camera is moving by analysing the paralax between foreground and background objects. When you shoot greenscreen though, all the software sees is solid colour. Detailed markers on the greenscreen and on stands or posts can help the software see specific areas of contrast and corners in space. We as viewers can look at that footage and know how the camera is moving, but imagine you're a piece of software that only sees a series of images made up of per-pixel colour information. How does it know that the camera is moving up and slightly forward? How does it know when the camera stops? The answer...

Maths.

I'm calling this new project "World's Edge". Was that a Pirate's of the Caribbean movie? Anyway, the idea is to have a person standing on a greenscreen stage and film a camera move, starting from ground level, until we're looking over the person's shoulder. The CG work will be to have the person standing on the edge of a cliff, looking over a canyon-esque, alien landscape. This project will end up being a large assignment for our 3D students next year. Of course, before they start it, I have to have it at least mostly done. My colleague Tom is going to be working on this with me too. We should be filming it in the next few days, perhaps early next week. Once we have it shot, I'll post the greenscreen footage and update this blog as I track and modify the footage.

That's all for now. Hopefully I'll have more regular posts starting next week.