How the game happened

We started developing The Nameless City 3 years ago. The core idea of the game was a PSX horror aesthetic inspired by Lovecraft’s short story.

At that time, the AI generative tools were an emerging technology. You could type some words, and the AI would output some random images that barely resembled what you typed. The idea was there, but things were oddly displaced or merged with some noise. It was really cool to prompt 'cosmic horror sky' and see how the AI would try to translate that to an image.

So I thought that using AI generated images to convey images (the creatures from the city) and sensations (the contact with a cosmic entity) that can't be processed by a human brain was also a great concept for the game. That’s why we included those vision videos during the demo.

What we think about the AI generative tools

The AI generative tools have come a long way during the past three years. They still suck, but what it was a novelty, now is used to generate marketing material (*cof cof* Wizard of the Coast, *cof cof* Wacom), to imitate the voice of an actor, or to generate scripts. Also, some other people use these tools to call themselves 'artists' (spoiler: they aren't).

Also, some companies are using AI to cut costs and replace people with these tools. When companies prioritise profits over everything else and they don’t value the work of their employees in the first place, this is what happens. The results of AI suck, but that doesn't matter when you can create cheaper and faster content. It's even worse when you know that these AI tools are trained using copyrighted content from the people that they are trying to replace.

Of course, we don't agree about the use of AI to replace people, or to use copyrighted art without consent to train these tools. We are not trying to proclaim ourselves as artists or to replace people (we didn't have money to hire people in the first place anyway), but we are using AI tools because the AI images sucking is what made them a good fit for the game.

We are going to be quite transparent, so we're going to explain the tools that we used, and how we used them. We are also going to attach all the AI generated content with the prompts we used to generate it. So you know that at least we didn't use prompts like 'anime girl in the style of <artist I want to plagiarise>'.

How we used AI generative tools

VQGAN+CLIP

Visions videos when you collect a glyph.

Stable Diffusion and DreamStudio (it uses Midjourney's models)

Frescoes around the end of the game.

Material Stable Diffusion

Frescoes 'sediment'.