What I like about doing this
rotoscoping and thinking about the layers or the potential 3D construction of the scene is fun
the motion tends to jump out at you from the picture
but in cases where it doesn't, like with some of the landscape perspective animation tricks, adding it is also very satisfying
in a business sense, sometimes its little touches or flourishes like these that go the extra mile when it comes to marketing or internal communications projects, or even the pitching/planning stages thereof
exciting to see/experience some of the different ways that the paradigm shift with AI imagery has played out; these new products make for such amazing starting points/springboards
The following are several micro examples of working through this process with Midjourney prompts quickly and on instinct:
When you consider that, in the original picture, there's nothing "behind" where the cliff clips in from the lower left, you begin to see the inherent challenge in these compositions.
You can't stop at slicing and scaling the layers; you have to backfill the occluded areas to account for the perspective changes.
Although it took me roughly the same time as the Coastal Cliffs animation, I was able to play around much more with the virtual camera settings this time, particularly the focal length (down to 16mm), aperture (30mm), and depth of field/blur.
From the BTS version, you can see how the layer stacking technique works to create a sense of depth.
Much like when I finished the "Great Green Heist" for Inktober, this image felt like it had too much motion to stay static.
Unlike the landscapes, this piece has a few animated elements besides the camera with the fish/bubbles etc.. My personal favorite, however, was using the puppet pin tool to have the child rearing back in surprise as the focus changes.
usually the big problem is separating things out into layers from a single image that you didn't construct from the ground up
as time goes on, different tools/methods are coming onto the scene that are shaking up the pre-existing stability of the image-construction pipeline (image, rotoscoping, animation, recompositing...); it won't be terribly surprising when there are options which generate images as independent, stacked layers (in 3D space or otherwise) or, conversely, systems for taking single images and converting them to such layered files automatically.
We're already a good deal of the way there with the way Adobe Photoshop (Beta for the moment) has its Generative Fill which uses area selections and text prompts akin to Midjourney to contextually fill specified area with new, normalized imagery that goes far beyond the previous capabilities of Content Aware Fill
I'm a big Hocus Pocus fan, and for the release of the sequel, I recut the 2022 sequel trailer with some composition work with a custom BooOoOok rig I made in After Effects.
When I built my new computer in 2022, one of the first performance tests I ran was a 3D camera rigged scene from a Midjourney piece cutout. Built from a single base image!
Ever on the hunt for ways to turn still photos into short animations, I began revisiting the depth map technique for simulating parallax over short offset distances.