The Lenticular Process Using 3D Rendering

It is is an animation and VFX teacher who was kind enough to share his insights into the process of creating a lenticular image.

This article was originally written for students, but it would be helpful for any CG artist so we reached out to share it with our readers, and we’re really glad to do it! Check out his helpful breakdown of the lenticular process using 3D Rendering below:

3D Lenticular Approach

CG artist always on the lookout for cool and new ways of showing off their works. The popularity of stereoscopy in cinema productions has given rise to a lot of great technologies, and an expectation from audiences that images should not be restricted to two dimensions. It seems strange that this has become the norm for movies projected at anywhere up to 48 frames a second yet printed media is still dominated by flat 2d ink on paper.

This case study details some of the processes involved with the creation of the source material for the lenticular print of the monkey you see below:

The Setup / The Composition of 3D Rendering

The output to lenticular required a composition that fit nicely into a 50 X 60 cm (the desired image size). This is a portrait orientation and at the recommended print resolution of 300dpi works out to being 6000 x 7200 pixels.

So the first thing We did was create our scene to take this framing into account (the dotted lines are composition guides in the 3d software Blender):

Also, We took a bit of advice from some stereoscopic film makers. Mostly this entailed pulling the audiences eye to a really specific focal point.

To this end We used a few simple techniques:

The color of the monkey contrasting with the environmental elements.
The heavy use of depth of field to blur the other foreground/background elements.
A composition that draws the eye to the monkeys face.

The Camera Setup

We require at least for 18 images in sequence. 3D Rendering has three options for the 3d camera setup.

Tracking: The camera simply tracking side to side (number of positions simplified, and range exaggerated for illustration purposes):

Tracking & Rotating: The camera tracking, but also aiming the camera at the subject:

Rotating: The camera rotating around the subject while aiming at it:

We ended up going with the third option (rotating around the subject).

Our decision was partly based on the parallax idea stating that the foreground objects should move from left to right. This implies the background objects would move in the other direction, and this wouldn’t be the case with a simple tracking shot sequence.

The reason We chose the rotating setup as opposed to the tracking & rotating setup was due to the fact that:

The rotating method kept the subject at the same distance from the camera from shot to shot.
The shot We were dealing with didn’t have objects way off in the distance or super close to the camera. This means the change from one frame to the next was small enough to maintain the illusion on stereoscopy.

For perfection you may need the angle of view of the lenticular lens. We took a stab in the dark and chose to make it a 15 degree sweep between frame 1 and 18 (about a 0.822 degree sweep between frames). This works pretty well for most of the lenticular pattern, but in hindsight, We advise you, have just got in contact with 3D Adworks and asked for the lens specifications.

To get the positioning of the camera right for each frame We made a simple setup that parented the camera to locator/null object that was placed at the focal point. The null object was then animated to rotate 15 degrees over 18 frames:

To get the spacing correct between frames we set the rotation animation curve to represent a constant speed (with no acceleration or deceleration):

Post Production

It’s worth noting that when it came to rendering the shot, We used the same general techniques used for rendering CG for feature film. Mostly this meant rendering multiple layers to EXR files for use with a compositing software.

Blender has a great compositor, but for this I opted to use the Foundry’s Nuke. Usually, we would separate out a great many passes to access the specular component, diffuse component, indirect illumination etc. This time however, we had pretty long render times due to the use of hair in my scene. So I opted for simply separating out the foreground, mid ground and background elements, the atmospheric elements and the different light passes.

In Nuke these were composited into the final image that was then taken across to Adobe Photoshop.

In Photoshop We increased the canvas size by 0.2 inches to allow for the bleed required by us. We also converted the colour space to RGB. We recorded this into an action and used the file->automate->batch command to apply this action to all 18 images.