One thing most Character Creator users never have enough of is new faces for their characters. Now, with AI image generators, we can conjure up any number of facial images for use in the Reallusion plugin that can produce some stunning results. We can generate completely random images, images of celebrities, images that almost look like celebrities, plus lots of plain old Joes and Janes to create a wide and never-ending variety of characters.
As to the legality of using celebrity or celebrity lookalikes I cannot offer advice. The same goes for all AI-generated images in general. We are on new ground here. These are things that lawyers, courts, lawyers, judges, and lawyers will be sorting out for quite some time to come. All I can say is to make sure you are on firm ground with the AI generator you are using in terms of licensing. No use adding more mines to the field.
On the left is the original AI-Generated image. On the right is the AI-Colorized image used in the Headshot plugin.
I decided to throw together a Vampire-Dracula character to sharpen up the old clothing skills since I hadn’t opened Marvelous Designer in a while. When it came to the facial image for Headshot I went the Stable Diffusion route using the NMKD GUI which was on version 1.6.0 as of this writing.
For this batch of renders, I used a simple prompt along the lines of:
Bela Lugosi in a tux, dark, evil, Vray, Unreal, Photorealistic.
No long-winded prompt. In fact, it is an ugly prompt, one only your Momma would love but it was all I needed. I kept it simple with straightforward visual guidelines that work well with Stable Diffusion. I ran several images, thirteen to be exact, and chose the one I like the best.
AI image used in Headshot to Create a face for the 3D Character.
The only real problem with my choice was the color. It barely had any, so I jumped over to Photoshop and ran the image through the AI-driven Colorize neural filter. Yes, I realize the initial image was near perfect color for a vampire but I wanted some flesh tones.
FYI, out of the thirteen generated images, eight met the criteria of working in Headshot (facing forward in a decently lit environment with the mouth closed). It was just a few minutes of work, mostly waiting for the renders to finish. As this is a local Stable Diffusion install, I can run all the images needed burning only storage space for the most part.
At this point, I was using AI to improve AI to use in an AI-driven plugin.
The Final Character. Hair was Generated by Headshot.
I took the finished image over to the Headshot plugin in Character Creator 4, using the Automatic selection, letting CC4 do all the work. I didn’t need to make the character look like the image in facial shape, so I skipped that tedious part and used the default output which came out a lot better than I expected. There is some shading on the face that could be corrected in Photoshop before using the image with Stable Diffusion, but this will be used in a darkened environment anyway.
Overall, the process is as easy as it is straightforward. The only hang-up could be getting a proper, portrait-type pose, and I would have most likely fared even better had I added the word “portrait” to the prompt. This usually generates a fair number of headshots without wasting a lot of frame space on the body or background.
The next time you need a new face… fire up your friendly neighborhood AI image generator which just might be enough to get you inspired to make cool characters.
M.D. McCallum, aka WarLord, is an international award-winning commercial graphics artist, 3D animator, published author, project director, and webmaster with a freelance career that spans over 20 years. Now retired, M.D. is currently working part-time on writing and select character development projects. You can learn more about MD on his website.