Building the Razor crest: An ILM Inspired Project

Printing the Razor Crest

Everything was grown on a Formlabs Form 2 resin printer. All the parts were post-processed with several grits of sandpaper and filler primer. I designed an internal clamping assembly with shaft collars positioned both parallel and perpendicular to the ship, allowing me to pick the ship from different spots depending on the shot.

To access the shaft collars, a portion of the chassis was separated out in cad prior to growing the parts. Two side panels and the rear cargo door were separated and attached with magnets.

LED rings were installed in the engines, lighting them with their signature, orangy-yellow glow.

The ship was airbrushed with Alumaluster and orange acrylic orange paint for the decals. A final weathering pass was applied with burnt umber and sienna oils.

Thank you Misumi-USA and IGUS for sponsoring this project.

Camera Rig

I began analyzing the ILM Razor Crest video frame by frame, specifically on the camera rig. I started by identifying a couple of parts and worked from there to make a very basic CAD model. The linear motion along an 80/20 rail is achieved using a 15:1 stepper motor, which self-actuates the whole camera rig with a 3-meter belt.

I built off the linear axis and designed the camera rig’s yaw and pitch with 7:1 stepper motors. The majority of the parts were grown on a 3d printer, accompanied by OpenBuilds slider mounting kit.

The camera I am using is a Canon EOS 5D MIII with a manual Nikkor prime 28mm f/2.8 lens.

miniature Rig

  1. The structure primarily consists of the 30 series 80/20 extrusion. The roll mechanism uses an IGUS quarter-circle rail and provides radial movement for the ship. A belt spans either end of the rail and a motor actuates along it, carrying the main chassis and pitch mechanism.

  2. The model is picked nodally to the rig from the pitch mechanism.

  3. The yaw mechanism is a staged sprocket chain that spins the entire carriage above.

The Shoot

I reserved a space at my college to setup up the rigs, bluescreen, and lighting. I did around four, six-hour shoots and filmed six shots.

There were a few issues that arose during the shoot. For one, the linear rail for the camera rig was too shaky from the overall lack of structural stability from the rail columns.

Furthermore, the grown parts from the 3D printer often warped, only worsening the shakiness of the rig. Many of the issues could be improved with metal parts.

The software I used to create the gcode files was Blender. In Blender, there is an animation window where I can record keyframes. Each keyframe for the motor axes can be exported through a Python script into a Gcode file. It took me around an hour to animate a full-length shot.

The Shots: What went wrong

I did run into a few issues with the post-production of each shot. One issue was the lack of ample lighting.

  1. In order to get the Razor Crest in full focus, I set the aperture to f/22 to get the widest depth of field. However, less light leaks into the sensor because the iris of the lens is very small at f/22. Being a complete newcomer to photography, I thought the solution was increasing the iso. I learned after that increasing the iso also increases the noise in the shot.

    • Longer exposures and ample lighting are needed to compensate for a wide depth of field.

  2. My second mistake was shooting in video. I should have instead exposed photos. I learned much later that this was how Go-Motion worked.

    • Exposing each frame for at least 1 second at f/22 in a properly lit environment, coupled with both rigs moving at a slow speed, creates the desired effect.

I will take my learned lessons from this version of the project to design a more robust and refined motion control system.