Tracking and Rotos

I’ve spent some time refining the roto shapes for this shot as well as trying to fix the stabilization problems.

I’ve added a few more tracking points to try to help keep the dock from shifting. I believe that I’ve solved everything I can in post. The problem roots back to the original camera solve. Thus, in order to completely get rid of the floating problem entirely, I’d have to do a completely new camera solve. This is something I’d like to do if time permits. However, I believe the shot now works as is.

I did notice one problem. As the dock was tracked along the shot, a gap was left between the end of the rendered frame and the edge of the plate.

At first I was going to try to rerender all the footage from Maya will the correct space given to that side of the frame. But then someone recommended I just scale up the the entire shot a bit. It worked nicely.

More Dock

Many things have been done to this comp since last Thursday.

First, I’ve rotoed out the foreground elements at the beginning of the shot:

I’ve also desaturated the dock as a whole. I ended up adding a seventh light to the scene. When looking at the other elements in the shot (the two characters, the real basket, various poles in the sand) it’s possible to see a very harsh highlight on the left side. I felt like there was not enough of a highlight on the dock. The extra light was meant to compensate for this.

I slightly blurred and faded the water reflections. I did the same to the shadow being cast on the water in addition to color correcting it slightly towards blue.

Next, I had to approach the sliding issue. The camera I was given from Boujou was extremely close to what I needed. At first, I was going to simply go back into Boujou and tweak the camera solve and try to get better camera. However, it was suggested to me that I should just try to fix it in post. This method seemed a bit bruit-force to me. However, it worked just fine… In order to do this, I stabilized the dock footage using the highlight on the basket at the end of the dock as the stabilization point. I then tracked the point at the end of the log where one of the characters is sitting. This was the only point I could use because that was the closest trackable point to where the dock meets the shore. All other points I tried caused the dock to slide even worse than the original camera solve. I used the log-point to steady the dock. Since the original camera solve was close, the perspective stayed correct as the shot progressed.

Next, I had to add motion blur. I tried various methods. Last quarter, I tried using a motion vector pass to add motion blur to a previous project with no luck. In the case of this project, I first downloaded a directional blur Shake plugin from highend3d.com. I tried keyframing the direction and blur amount. This quickly proved to be extremely clunky. I next resorted to the motion vector method again. First, I tried using Maya2009’s Render Pass system (preset: Normalized 2D Motion Vector). Again, I had to download a plugin from highend3d.com in order to attempt to get this to work. Apparently, I did something wrong. I was getting a directional blur, but the direction never changed with the camera motion and the blur remained constant. I might as well use the keyframed directional blur method. As a last resort, I found a document explaining how to set up motion vectors using Mental Ray’s production shaders. This method worked.

Last, I had to color correct the plate as a whole. I was originally given the stabilized footage without color correction. Since starting this comp, Glenn and Travis color corrected the plate to match the rest of the film. Glenn gave me the color correction script they used. All I had to do was import it at the end of my script.

Uncolor-corrected:

Color-corrected:

Scene7 Shot1

For my next shot for Mirror, I’m comping a rickety dock onto one of the beach scenes. The purpose for this is to help further the culture of the people in the film.

Again, I was given models to use in the scene.

Model by Rio Harrington

Model by Will Schulthius

While these two models weren’t modeled to look exactly the same, I was still able to fashion a decent looking dock. I used textures Rio created for his model to texture the whole dock. I also modified the models slightly in order to fit the scene.

Glenn also give me some rough tracking data. While the model placed in the scene slid slightly, it was still a good enough track to work with.

I used the tracking markers to correctly place the dock in the scene.

I also created a rough ground plane used to block part of the dock from view. According to Glenn, the beach was fairly steep. The tracking markers extracted from Boujou reflected this. It was easy enough to model a plane following the markers.

I built a quick light rig and broke the scene out into layers. I used six lights in the scene. One for the sun, a bounce light from the water, a fill from the sky, and three other fill lights from various angles.

I put together a rough comp to show Joe, Glenn, and the film crew on Thursday the 23rd:

Despite being littered with dropped frames, the comp gave the idea of what I was going for and received good reviews. The director did not, however, like the hut at the end of the dock. Things needing to be done by Tuesday’s class include fixing the slight sliding going on throughout the sequence, desaturating the dock slightly, tweaking the shadow on the water, adding correct motion blur, and omitting the hut entirely.

Scene3 Shot2b (Shake)

Boujou was used to extract a 3d camera from the shot. This scene was a nightmare to track. With a little help from Professor Pasquale, I was able to extract a camera able to hold up to the motion of the shot with minimal floating.

On to Shake:
Overall 12 render layers were used to composite the shot. Looking back, this was probably extreme overkill. I’m still feeling out which layers are truly necessary for each comp.

Overall, I think the comp works for the purposes needed. However, there are many things I still want to do to it if I have time to go back.

Scene3 Shot 2b (Maya)

Since finishing my studio work on Trouble Brewing (www.troublebrewingfilm.com), I haven’t really done a good job keeping an updated blog on my progress. Glenn was kind enough to let me use the Mirror blog for all my Mirror-related work I’m doing this quarter. Over the next few posts, I’ll provide an overall update to this point.

The first thing I worked on for Mirror was Scene3 Shot2b. This was ment to be an extremely quick comp where I would have to cover some shoes accidently left in frame during shooting.

I was given a few models to use (all modeled by Will Schulthius): an axe, a basket, a tray, and some wood.

At first, I added the items all around the scene (not just over the shoes). It was soon pointed out to me that this was a bit pointless so I cut the entire comp back to only placing a few items over the shoe. The wood was omitted entirely.

While waiting for textures, I quickly developed a light rig, and split everything into render layers. After a while, it was soon evident that everyone’s schedules were packed and I wasn’t going to get the textures anytime soon. Each model was already UVed. I decided to texture everything procedurally. They did not have to hold up to intense scrutiny… I just had to give the basic idea of each material.

Naming Convention

<!– @page { size: 8.5in 11in; margin: 0.79in } P { margin-bottom: 0.08in } –>

In order to keep this mess as manageable as possible, please always follow the naming convention. The naming convention will also be based on the footage given to us. Starting with the scene, then the shot, your intitals, a description, the version, and finally the file type:

-Scene_7_S11_GS_Particles_V1.hip

Scene 7, Shot 2, by Glenn Snyder, Particle Effects

-Scene_7_S1_WS_Basket_V4.mb

Scene 7, Shot 1, by Will Schulthius, Basket model

-Scene_7_S3_MK_Track_V1.pt

Scene 7, Shot 3, by Matt Kiefer, Tracked camera

-Scene_7_S2_DG_Comp_V3.shk

Scene 7, Shot 2, by Dorian Gunnels, Composite script

File Structure

<!– @page { size: 8.5in 11in; margin: 0.79in } P { margin-bottom: 0.08in } –>

We will be basing the file structure around the footage, grouped by scene and shot. Inside each scene folder will be a folder named Footage that will contain all of the footage for the entire scene. Inside each shot folder you will find another Footage folder, this time only containing the .tif files for that exact shot.

Along with the footage, each shot directory will also contain a folder for each element necessary to complete that particular shot. For instance:

Scene_7 Shot_10 is just before the mirror breaks, the shot needs to

be tracked, then stabilized, and finally color corrected,

also there is footage because it is live action.

In Scene_7 Shot_11, the mirror breaks underwater,

we have Effects and Compositing.

Since it is a full CG shot, there is no need for a

Footage, Color Correction, or Tracking folder.

Objects, such as models, textures, and camera rigs that will be used repeatedly by multiple people will all be kept together in an Asset folder for easy organization. Inside the Asset directory you will find separate directories for Models, Textured Models, Houdini OTLs, Tracked Camera, Light Rigs, and Color Tables. Import the assets needed from this folder into the project folder of the scene you are building.

Located in the VSFX directory with the Scenes are two other directories, Reference and Documentation. The Reference folder contains the “Look Book” given to us by the Director and DP for overall direction. The Documentation folder contains all word documents and .pdfs, such as the shot list, script, and due dates.

Project Breakdown

For the past year, Writer/Director Stephen Cellar, Director of Photography Jarad Kleinburg, Producers Aram Kaplan and Max Crandall, and Editor Francis Repass of the Film Department at the Savannah College of Art and Design have collaborated with students from the Fashion, Set Design, Media and Performing Arts, and Industrial Design to create beautiful 1080p HD footage. From here it is up to the Post-Production teams from the Visual Effects and Sound Design Departments to create the flow, rhythm, and feel of the film. The Visual Effects team will have control of the final look and atmosphere of the piece as a whole. That means adding objects, set extensions, and effects to create a more natural feel, as well as using stabilization and color correction to control the movement and mood.
We will be working with fifteen minutes of footage, which is just over 21,600 frames or 124gb. Those 21,600 frames are then nicely cut into twelve scenes, and finally placed ever so elegantly into about 150-170 shots. All of these shots need color correction, 75% need stabilization, and about 35% have CG elements. There will be a total of fifteen students working collectively as a team for six weeks.

Software we will be using:
Maya – Set Extension
Houdini – Effects
Shake – Compositing/Stabilizing
PFTrack – Tracking
Luster – Color Correction

Effects

The Effects team will be shattering a mirror in slow-mo, cracking a mirror underwater, recreating a small wave on the shore, splashing in a puddle, replaing fire, and adding smoke. The team consists of Ryan Bowden, Joseph Burnette, Chris Brown, Glenn Snyder, and TJ Zoltner. Here are the stills from the shots they will be working on:

2D Set Extension

The 2D Set Extension Team is Tyler Rodenburg and Robert Gilkes. They will be compositing a mix of live footage and matte paintings to add a little more depth to the village and beach.