The Technical Magic Behind the Scenes

People ask: how do you make one new piece per day?

The answer: we built a system.

But I want to be clear: the system doesn't make the art. It amplifies our taste. It lets us explore more ideas without getting bogged down in the repetitive parts.

The Concept Library

We have about 150 building blocks: camera angles (centered behind, elevated, wide shot), themes (zen, chaos, cyberpunk, floaty), props (keyboard, headphones, energy cans, cat), body types (hood up, long hair, short hair), art styles (watercolor, anime, concept art, 3D), moods, colors.

When we want to explore a new idea, we pick concepts and combine them.

The Prompt

Then we write a very precise prompt to an AI image generator. We've learned that specificity beats vagueness.

Instead of "draw a gamer," we say:

- "Person sitting on a stool, back of hands facing camera, left hand on keyboard, right hand on mouse" - "No furniture except the stool. Pure white background" - "Watercolor style with tight expressive brushstrokes" - "Emphasize the mood, not the equipment"

A single word wrong and the entire image changes. We've spent months learning this language.

Generation

The AI generates 4-8 variations based on that prompt. We look at them. We ask: does this feel right? Does it capture the mood? Is the composition clean? Is the color palette working?

Curation

Here's the hard part: we reject most of them. We're ruthless. If an image is good but not excellent, we delete it and regenerate.

We'll run 50 generations to get 5 that are worth keeping.

Upscaling

The good ones get upscaled 4x (from ~1000px to ~4000px) so they print beautifully at A3 size. This is the difference between a poster and wall art. The resolution matters.

Listing

We write a title and description that describes what the image feels like. Not what it is (gaming wall art), but what you feel when you see it. The colors, the mood, the moment.

Then it goes live on Etsy.

Why This Takes Time

People think AI means instant. It doesn't. The generation is fast. The curation is slow. The decision-making is slow. We're not trying to maximize output — we're trying to maintain quality.

One good piece per day is more valuable than five mediocre pieces per day.

What Gets Automated

- Image generation (fal.ai handles this) - Upscaling (SeedVR handles this) - Listing title/description (Claude AI drafts based on our templates) - Publishing to Etsy (Gelato handles print-on-demand)

What Doesn't Get Automated

- Deciding whether the image is good - Deciding if the mood is right - Deciding if it belongs in this collection - Deciding what color variants to create - Deciding if it's worth publishing

Those are human decisions. That's where the taste lives.

The Philosophy

We built this system because we wanted to explore more ideas without burning out on the repetitive parts. But the hard part — deciding what's good — that's still all us.

It's not magic. It's work. The system just lets us do more of it without getting tired.