Nova Rhyne - FACE EXPRESSIONS VISUAL UPGRADE

Hey guys!

There’s been something bothering me for a long time, and after experimenting with different ways to solve it, I finally found a workflow that I think can make a big difference in production.

So... what was the problem?

Eira’s face. More specifically, her eyes and facial expressions

When I create a completely AI-generated image, I feed the model with all the information I’ve built around the character so it can keep Eira consistent.

Most of the time, it works really well. But AI still has its weird moments.

Sometimes the eyes don't look where I want them to. Sometimes an expression feels slightly off. And sometimes you look at an image and immediately know something isn't right, even if you can't explain exactly what it is.

That's because we're humans. We’re incredibly good at recognizing faces and expressions, so even tiny inconsistencies can stand out.

FIXING THE PROBLEM

Instead of simply accepting those results, I started experimenting with another AI platform using a different model to correct these specific issues.

Then I can take the different results, blend them together, apply digital painting, fix details, match the layers, add effects, and finish everything with post-production.

So the final image isn't simply:

“AI-generated → done.”

There’s actually quite a bit happening between those two steps.

And while working on this, I had a realization:

WAIT... I CAN USE THIS FOR 3D TOO.

If AI can help me correct an expression on an AI-generated image...

Why couldn't I use the same idea on my 3D renders?

Instead of spending a huge amount of time creating every facial expression manually in 3D, I can create the character with a neutral expression, render it, and then use AI to create the expression I actually want.

Want Eira to smile?

Let AI handle the initial expression.

Then I can bring the result back into my workflow, blend it with the original render, correct anything that looks wrong, and finish it with my usual post-production.

Basically, something that started as a problem with AI-generated images became a tool to improve my 3D workflow

WHY THIS IS IMPORTANT

Facial expressions in 3D can be surprisingly time-consuming.

Eyes, eyebrows, eyelids, mouth, cheeks... There are a lot of little things that need to move together to create a convincing expression.

Doing that manually for every shot adds another layer of work to an already massive production pipeline. If I can move some of that work into an AI-assisted workflow, I can save time without sacrificing the control I want over the final result.

And this is exactly why I keep experimenting with these tools.

As an artist, I have a vision of what I want to create, but sometimes the limitation isn't creativity. It's time.

So whenever I find a way to remove some of the repetitive mechanical work, I can use that saved time to create more.

THE BIGGER PICTURE

This is also why I use multiple AI platforms, 3D software, and post-production tools.

Each one has different strengths.

Sometimes I use one model to create the base, another one to fix a specific problem, and then I manually bring everything together.

The AI helps me move faster, but I still need to recognize the problems, understand what needs to change, make the artistic decisions, and integrate everything into the final result.

And because I’m building an entire universe, consistency is extremely important.

Every new workflow I create is another step toward making that process faster and more reliable. So, ironically...

An AI problem ended up becoming a 3D solution.

And I’m pretty happy about that

I’ll keep experimenting with this workflow and, as always, I’ll share more of these improvements as they become part of the Primal Cargo production process