Not “output” — artwork

There’s a common idea that AI art is like a vending machine:

type in a prompt, get out a finished image.

If that were true, you’d mostly get shallow, generic pictures.

Verse Zine Royalty Free doesn’t work that way.

Every image in the library is created by me, Nathan Roberton—a human artist—using AI tools as part of a larger process. This isn’t a “type your own prompt” generator. It’s a growing library of finished artwork, built gallery by gallery with specific moods and use cases in mind.

‍

It starts with a vision

Each gallery begins before any AI is turned on:

• a mood or feeling I want to capture,
• a visual metaphor or material (glass, dunes, architecture, fog, etc.),
• sometimes seeds from my own photography, 3D renders, or earlier sketches.

I treat it like any serious design project:

• What is this series for?
• Where could someone realistically use these images?
• What emotion do I want them to carry?

Once that direction is clear, I start working with models.

‍

How AI actually fits into the work

Behind each gallery is a lot of technical and editorial effort. At a high level:

1. Choosing and shaping models

Over years of experimentation I’ve developed and continue to refine:

• personalization codes and moodboards,
• my own style references and custom model tweaks (like LoRAs),
• a curated library of discovered codes, models, and techniques from other artists,
• carefully refined image prompts and text prompts.

That lets me aim for a particular look instead of hoping for random happy accidents.

2. Iterating and steering

I don't write one prompt and hope for magic. I:

• iterate on prompts across hundreds of generations,
• use variations (Vary Strong in Midjourney) to explore promising directions,
• combine personalization codes and style references strategically,
• sometimes use outputs from other AI tools (like Nano Banana or specialized models) as visual prompts,
• collect and curate image references that help steer future generations.

This is the unglamorous part: a lot of trial, error, and judgment.

3. Refining the images

When a piece is close but not there yet, I’ll:

• generate many subtle variations and select the strongest results,
• use Midjourney's editing tools to selectively regenerate regions that need refinement,
• bring images into tools like Pixelmator Pro for manual touch-up and polish.

The goal is to reach something I’d be comfortable using in a real client project.

4. Selecting and sequencing the gallery

For every image that ends up in a gallery, many are discarded.

I’m looking for:

• a cohesive visual language,
• enough variation to be usable across layouts,
• and a set that feels like it truly belongs together.

That selection step—what stays, what gets cut—is where most of the art direction happens.

The “gap” where art actually happens

Like any serious medium, there’s always a gap between the image in my head and what the tools can produce.

Painters deal with paint. Photographers deal with light and weather.

With AI, I deal with how the model interprets language, reference images, and training.

I’m in control of the direction, but I’m also responding to unexpected results—keeping the ones that carry the feeling I was chasing, and rejecting the ones that don’t. The “soul” in the work lives in that ongoing decision-making, not in the raw model output.

Why my background still matters

AI didn’t erase the need to learn art. It builds on it.

I've been working with AI art since 2018—starting with GANBreeder and early experiments with GAN+CLIP notebooks—years before generative AI became mainstream.

Before Verse Zine Royalty Free, I spent decades:

• drawing and painting in traditional media,
• modeling, lighting, and rendering in 3D tools,
• designing interfaces and visual systems for real-world products,
• experimenting with early generative models long before tools like Midjourney existed.

That history shapes how I see composition, light, and rhythm—and how I decide when a piece is good enough to ship.

AI makes some parts faster, but it doesn’t decide what’s worth keeping. That’s my job.

The process keeps evolving

The images in the current library are primarily created in Midjourney using the methods above. But I'm constantly exploring new techniques—integrating hand-drawn and digitally-painted elements (via tools like Krita AI Diffusion), training custom models, and developing new ways to maintain aesthetic control while expanding what's possible.

The goal isn't to chase every new tool. It's to find what serves the vision.

What this means for you

When you license an image from Verse Zine Royalty Free, you’re getting:

• original artwork born from a specific human vision,
• shaped over many iterations rather than one prompt,
• selected to work in real projects—presentations, websites, brand systems, and more,
• priced to reflect the efficiency that AI brings to the production process.

You’re not getting:

• random one-shot outputs with no artistic direction,
• AI 'slop'—the thoughtless mass-generation critics rightfully call out,
• images scraped from someone else's finished work.

These are images I’ve created, evaluated, and stand behind.

Why we’re clear about using AI

Some people love AI art. Some dislike it on principle. Most just want honesty about what they’re using.

So here it is:

• Yes, these images are made with AI tools.
• Yes, they’re directed, revised, and selected by a human artist.
• No, the tool does not generate this library on its own.

If that combination of human vision and modern tools appeals to you, you’re exactly who I’m building Verse Zine Royalty Free for. If AI-assisted art isn’t for you, that’s okay too—I’d rather you know now than feel misled later.

—

Nathan Roberton

Artist & Founder, Verse Zine Royalty Free