Inspired by the influx of awesome 3D street art illusions in recent years, I wanted to see if this clever technique could be transferred across to the CGI/motion graphics world.
To do this, I needed a basic understanding of how this method works, specifically how it tricks the eye into believing a false perspective. What I learned is that firstly, it only works when viewed from a specific angle, and secondly, the effect is most convincing when viewed with one eye or through a camera lens. This is because as humans with two eyes, our depth perception immediately breaks the illusion allowing us to realise that the image is in fact completely flat.
My goal was to achieve this same effect, but through a TV screen or monitor, and ANIMATED. Working backwards from the final result, you can see in the image (right), the artwork must be severely distorted in its flattened state, in order for it look correct when viewed from a chosen angle. However, creating this sort of distortion in 3D is incredibly difficult. I realised the best way to achieve this effect, was to model the scene in 3D as normal, and distort the render. Then just as in the street art examples, that distortion would be 'un-done' by viewing the piece from the correct angle.
To better explain this method, I've created a simplified 3-step example (below), displaying the modelling, rendering, and stretching steps needed to achieve this effect.
In the example (above), a simple red sphere is intersecting a 16:9 grey plane. The plane represents the screen that the animation will eventually be displayed on, and as such, anything that goes beyond the boundaries of this area will be cut off, ruining the illusion. It's important that the animation is rendered from the angle it is intended to be viewed from for the effect to work. Once rendered, the animation is then stretched so that the grey plane is restored to it's original dimensions, in this case 16:9.
Below is my second animation test of this technique. It's a simple abstract animation, but I really just wanted to see how far the effect could be pushed. By extruding the surface backwards, it gives the effect that you can see 'inside' the screen.
I then explored if this effect could be displayed on two screens simultaneously. The principle is exactly the same, only each animation is rendered from angles that match the final viewing angles of their corresponding monitors.
To give an idea of the level of distortion these animations need to be in order for this effect to work, I took a photo of an early test from wrong side of the screen (below).
It's an interesting technique that I've yet to find an actual use for, but as a motion graphics designer, the process of figuring these things out is often the most enjoyable part of the job for me.