🔨 Forging the Future #1 — Choosing the Right Image Model (Patreon)

🔨 Forging the Future #1 — Choosing the Right Image Model

Welcome to the first post in a new behind-the-scenes series where I pull back the curtain on the tech, the struggles, and the decisions shaping Celina’s Calling (and the future of Morrak Games).

Today: image generation — and why it’s harder than it looks.

When building a game, I need three things from an image model:

Consistency

High Quality

Versatility

Right now? You’re lucky if you get two out of three.

🎭 Consistency vs. Complexity

Some models can produce incredibly realistic renders. With tools like IPAdapter inside ComfyUI, I can keep a character’s face extremely consistent across scenes.

The two attached images show just how consistent Celina’s face could be using that workflow.

But here’s the problem:

What about body consistency?

What about poses?

What about multiple characters interacting?

What about camera angles?

The moment you introduce more than one character, the complexity skyrockets. Keeping one face consistent is manageable. Keeping two bodies, proportions, lighting, style, and composition consistent across dozens (or hundreds) of images? That’s a different beast entirely.

🧌 Versatility (Or Lack Thereof)

Then there’s versatility.

Many modern models (Flux, Pony, SDXL derivatives, Illustrious, etc.) are trained heavily on modern photography and social media–style content. They’re excellent at:

Realistic portraits

Fashion photography

Clean studio renders

But when you ask them for:

Explicit scenes

Orcs

Trolls

Magic effects

Fantasy settings

Things start to break down.

See the attached image of the man with antlers?That was my attempt at generating an orc using a Flux-based workflow.

Technically impressive? Sure.What I needed? Not even close.

That’s the gap between a good model… and the right model for a game like this.

⚖️ The Trade-Off Problem

Could I mix models?

Yes.

But then you introduce:

Style inconsistencies

Lighting differences

Quality variations

Different “feels” between scenes

And while I already wrestle with that somewhat, I don’t want the player subconsciously feeling like they’re jumping between different visual universes every few scenes.

For a story-driven experience, cohesion matters.

🌈 Enter: Chroma

After a lot of experimentation, I’ve landed on Chroma as my base imaging model.

Why?

Because in my experience, it hits the best balance of:

Strong realism

Fantasy capability

Fewer restrictive filters

Better creature generation

More flexible styling

With my new PC setup, I’ve also been able to run the latest versions of Chroma with features I couldn’t properly utilize before. That means:

Better consistency

Faster generation

Stronger LoRA integrations

More reliable workflows

And I’m already seeing improvements.

More sample photos are coming soon. 👀

This is just the beginning of Forging the Future. As always, thank you for supporting this journey. You’re not just funding content — you’re helping build the forge that makes it possible.

— Morrak Games