A Successful Kickstarter, and What It Taught Me About AI and Art

“2,001 backers chose this project. And the voices that questioned us because of AI made me rethink what kind of publisher we want to be.”

On my computer, there is a folder called “Failed Projects.”

Inside are seven tarot design projects. Some never got past the initial concept; others were in progress for a long time. In the end, each was abandoned for its own reasons.

Over the past eight years, I have worked with many illustrators and tarot readers, and explored all kinds of art styles and creative directions. Gradually I came to see that making a tarot deck with polished artwork and creating one with real depth are two very different things. Drawing skill can be trained. But how an artist understands the world, expresses emotion, and gives a work meaning is bound up with her upbringing, experiences, and values. Those things are hard to replicate.

Creating a tarot deck also takes time. Most illustrators I have worked with need eight months to a year and a half to complete a 78-card deck. After that come the guidebook, refinements to the artwork, prototypes, and production. Not every project that receives significant time and money makes it into collectors’ hands.

Sometimes a project stops not because it isn’t good enough, but because we don’t have the resources to keep it going.

After learning about its creative vision, its storytelling framework, and the tools used in its creation, I knew one thing clearly: given the project’s scope and the resources available at the time, if the project had relied entirely on traditional hand illustration, it would most likely have become the eighth project in that folder.

So I decided to support this deck and its AI-assisted creative process, and we were transparent with our backers about the use of AI from the very first day of the Kickstarter campaign.

Throughout the process, the designer used the Dunhuang murals and traditional Chinese mythological art as visual references, and did not feed the work of contemporary illustrators into the generation tools as references. The designer extensively redrew and refined each card’s composition, visual style, and details, and made further adjustments for printing.

“This was not a project where someone typed in a few prompts and waited for 78 images to appear. AI was part of the process, but the cultural content, aesthetic judgment, and artistic expression in the work still came from people.”

In the end, ACCG was supported by 2,001 backers on Kickstarter and raised over HK$1.56 million. A deck that might never have been finished made it into collectors’ hands. I am glad of that, and grateful to everyone who supported the project. Yet the experience also left me with questions I hadn’t expected.

While promoting ACCG, we received a great deal of positive feedback, and also some very direct opposition. Some people do not accept any form of AI-assisted art, and some customers came to view our brand negatively because of it.

What concerned me more was that some artists we might otherwise have worked with strongly disagreed with the project. In their view, supporting AI-generated images means supporting a technology that may have been trained on human artists’ work without their consent, which amounts to taking the fruits of artists’ labor. We lost potential collaborations because of this.

At first, I wanted to explain how the project was actually made. But later I realized that such explanations might not address their real concerns. What they care about is not only how ACCG was made, but where AI training data comes from, whether artists consented to their work being used, and how original illustrators will continue to make a living as generative technology becomes more widespread.

I understand this unease. I have worked with illustrators for years, and I know how much time, energy, and money it takes to complete an original deck. At the same time, I’ve also seen projects that might never have been completed without AI assistance. Both of these realities exist at once.

I cannot know how many potential collectors chose not to support ACCG because of AI. But commercial success cannot answer every question about creative ethics for us.

As a publisher, I used to judge whether a deck was worth publishing mainly by its artwork, theme, story, production costs, and market potential. ACCG made me think about another question: when people buy a tarot deck, what are they really paying for?

Is it 78 beautiful images? The stories, ideas, and culture behind the work? Or the skill an artist has built over many years, and the time and care she has poured into every card?

Both choices deserve respect.

I don’t believe artistic value can be measured simply by the hours spent, nor that a work made with technological tools lacks the creator’s thought and judgment. But I have come to see that a work’s value does not lie in the final image alone. How it was made, who made it, where it came from, and the trust between creator and collector all shape how people see a piece of art.

This is especially true of tarot. A tarot deck isn’t just something to look at. It may stay with its owner for many years. People use its stories and symbols to explore their own lives, and they form a personal, emotional bond with it. Collectors have every reason to care where these images came from, who created them, and how they were made.

So now, when I think about how to bring a work to completion and how to manage costs, I also ask one more question: how will the creative approach we choose affect other artists, and the trust of our customers?

I don’t regret publishing ACCG. It let us explore a new way of creating, and it introduced more collectors around the world to ancient Chinese mythology and traditional art through tarot. That means a lot to me. But commercial success doesn’t make concerns about AI any less valid.

This experience did not lead me to conclude that “AI is better than hand illustration,” or that “art should never use AI.” Instead, it made three principles clearer to me:

Looking back, ACCG is still a deck we are proud of. Its 2,001 backers showed me that it found the people who love it. And those who disagreed with our approach made me rethink what kind of publisher we want to become.