BLOG

Master and Apprentice - Zoom

by GymDreams
Builders of the Great Greek Temple
Master and Apprentice - Zoom, with Midjoureny v5, v5.2

Master and Apprentice is a sub-series from the Builders of the Great Greek Temple. It depicts the relationship of an old master and a young apprentice seen throughout history across all cultures. Although not always gay in the nature, many gay men did find such bond to be a meaningful part of their personal self growth, especially when being gay could often lead of losses from the family that they grew up in.

Technical Info

The images are a sequence of 8 images, each at x2 zoome from each other. The video is a smooth transition from the first to the last image, using an improved workflow and technique inside After Effects, which I will explain below.

On Instagram the first and third sets have identical images, which are all x2 zooms for each step. I wanted to maintain the 3 per row on my Instagram, so I simply reverse the order of the sets, so you can see the furthest and the closest image when viewing from the profile page.

The reel is a zoom video of the 8 images stacked together to create one smooth transition, with an improved workflow inside After Effects. The TL;DR is that scale and zoom levels do not have linear relationships, so using scales to do the zoom transformation will result in hiccups. The solution is to use 3D layers with a 3D camera to create a single dolly shot.

Images

One downside with Midjourney’s implementation of this is that there’s no way to change the prompt in subsequent zoom levels, so your fate depends highly on how Midjourney interprets your text prompt, and some RNG.

  • Original painting: text prompt in Midjourney v5
  • Subsequent x2 Zooms: text prompt in Midjourney v5.2

Video

It took many tries until I figured out how to do this zoom in a perfect seamless move. When I made the Mondrian reel, I used scale on a 2D layers, and due to size limitations of the canvas in After Effects, broke that sequence into multiple comps. But even with that and by scaling the layers, something looks off in the final render. That’s because scales don’t correspond to zoom levels in a linear correlation, so if you transform the scale for the zoom, it won’t look linear at all.

Instead of working in 2D, I worked with 3D layers for this video. The upside is that I’m no long limited by the canvas size while being able to keep the needed resolution. I still stack all 8 images together on the same Z plane. Then I added two things: a null object to control all 8 layers’s scale level, and a 3D camera to do the actual zooming by dollying in. The result is a smooth walk through the hallway without any hiccups seen in the Mondrian video. Technically, only the 3D camera is needed, but I ran into the z-depth minimum, at -100000. So I cheated by scaling the Null layer, which served as a parent controler for all 8 images, to achieve the final effect.

Images: Zoom In

Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In
Master and Apprentice - Zoom In

Images: Zoom Out

Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out
Master and Apprentice - Zoom Out