I woke up with an idea in my head. Because I am working on the “The Hoof That Wouldn’t Die” movie, I thought it would be pretty hilarious if the head from it would talk or sing something regarding the new year. Being kept alive as just a human head in some strange liquid isn’t the best starting point for a great new year. And then the song “Feeling Good” came to mind. A seed was planted that grew up to be this one-minute video. Kind of strange, the music is not particularly soothing and her behavior, also a little creepy.
For the track I took an acapella of the singer of the
Avicii version of Feeling Good sung by Audra Mae and took some inspiration for the instrumentation and the electric piano arrangement from the version of Muse. No samples were harmed in the making of this. Only the vocal performance, which was immediately recognized by Youtube. That is why you may see ads (so the original artists gets paid).
These are the steps to create the animation.
I picked a scene I had already worked on for the full-length movie. Just because it has no camera move, light disturbances in the liquid and reflections, and a nice framing of the face.
I color keyframes for the scenes mostly by hand, but I do start out with the a.i. based option in Photoshop to color black and white images. When creating a colored version of this old movie, I discovered to have consistent images and style, I am better off choosing my own colors. And of course, Photoshop would not think of using these over-the-top colors. The movie will be in technocolor. With a techno soundtrack. So the colors have to pop. Also, I do work on the images to make them less flat in a later step.
The way I wanted to make the image sing was originally by giving her a new face and matching it as good as I could with the “real” material. I created the animation of the face in Poser. I have used that in the past for different means, for now the Talk Editor was my reason to open it again. That proved to be a tough job, but for a first try, it was ok. Besides the mouth movement, that does influence the rest of the face, I also wanted to emphasise the singing with movement of the brows and eyes. Human eyes are never completely stationary, so I animated them to be constantly moving subtly.
Having the light a little similar to the situation in the movie helps, and I made a high quality render as a basis for the performance.
First I created keyframes for every mouth shape and before and after every blink. Using EbSynth to map those to the rendered file from Poser. The result needed a lot of work and remained very wonky, I remembered seeing face replacement done with EbSynth that looked way better, so I revisited that YT-video. That taught me to use just 1 keyframe, that will follow every movement of the face. In the original footage, the actress never really looks straight into the camera, so that had to be created from a single frame where her eyes were roughly in the right position when moving from right to left. some endless hand painting and a lof of steps back and forth resulted in a passable frame.
With too much layers and some nifty tricks to replace part of the moving background image with static frames, it was possible to merge the output of the video mapping and the original footage. Using the eyes from my first try for when they are closed, and the other material for the rest of the face.
In the software Studio Artist I make the frames look less flat and have an airbrushed feel to them. The latest version of the software allows to do the same series of a.i. based filters and assisted painting to a movie file easily. Though it still takes a long time to do 1900 frames of course. Most scenes in a movie or way shorter than that.
In the same software, I also create a crosshatched version of the movie, that makes the images look more hand-crafted, like somebody made the darker areas with a pen or a thin pencil.
In my mind, this makes it look like an image that could be from a graphic novel, or old comics. When developing this look, that also influenced using less colors, to fit the idea that this is a moving version of that particular medium.
The third layer I generate is a line version of the image, that gets multiplied with the previous two.
The end result is not a photorealistic color version of a 1959 movie, but something different. But it still moves like a movie (which is more visible when you see the full length movie of course).
To convey a feeling of analog old times, a video of old, damaged, white film gets overlaid on this.
Adding animated titles (jittering a little) and the logo of the upcoming movie completed this idea. Five to six days of work for one minute of, well, something.