Skip to main content
The Keyword

AI

How four artists used AI to endlessly reimagine “Alice’s Adventures in Wonderland”

a title card saying "Eric Hu," "Shawna X," "Erik Carter" and "Haruko Hayakawa" and "x Imagen" with four animated characters
10:25

From the printing press to cameras to synthesizers, new technology has always raised new questions about the role of art and artists in our society. And today, generative AI has creatives questioning how the nature and process of their work might change again. So, for our latest Lab Session, we’ve partnered with four visual artists — Eric Hu, Shawna X, Erik Carter and Haruko Hayakawa — to explore the question: “Can AI be a tool that expands an artist’s vision and accelerates human imagination?”

Together, we created Infinite Wonderland, a first-of-its-kind experience where the artists fine-tuned an AI model to endlessly reimagine the visual world of the timeless novel “Alice's Adventures in Wonderland.”

How “Infinite Wonderland” was made

Inspired by John Tenniel’s original illustrations, each artist created a set of new images visualizing the novel through their own eyes — designing key scenes and iconic characters like Alice, the Mad Hatter, the White Rabbit or the Queen of Hearts.

Each artist used their original images to fine-tune Google DeepMind’s AI image generation model Imagen 2 to generate infinite images in their own, unique styles.

A row of four images depicting each artist's stylistic interpretation of the character Alice from "Alice's Adventures in Wonderland", with the artist names listed under each: Eric Hu, Shawna X, Erik Carter, and Haruko Hayakwa.

Traditionally, style transfer techniques required hundreds or even thousands of similarly stylized reference images to successfully reproduce a style. But by using a technique called StyleDrop, the artists were able to fine-tune the model on their style with no more than a dozen of their own images.

Through an iterative process, each artist could see how their original images shaped the outputs from Imagen 2. They could then make creative alterations until the model achieved the aesthetic and composition that felt most authentic to their style.

This fine-tuning technique offered an entirely new opportunity for the artists to create work at a previously unachievable scale. Readers of Infinite Wonderland can generate seemingly infinite images for each one of the 1,200 sentences in the book. Every single reader will have their own unique experience of the story.

Human imagination x Artificial intelligence

By using StyleDrop on Imagen 2 to invent a new open-ended storytelling format, these artists can co-create with their audience in a unique way. It also stimulates their own growth as artists, helping them to explore new facets of their own styles. They can quickly generate new interpretations of characters, alternative compositions to complex scenes, and make creative discoveries they may not have found on their own.

Without that human touch, I wouldn’t have been able to create this fantastical world. The role artists play in crafting these AI models is vital and impacts the whole creative process. Haruko Hayakawa

Infinite Wonderland offers a glimpse of what’s possible when we co-create AI with collaborators. As our AI models continue to evolve, we’re excited for artists to continue to push the boundaries of where technology and creativity go next.

In the meantime, come down the rabbit-hole to explore Infinite Wonderland for yourself.

Labs.google is Google’s home for the latest AI experimentation. We’re a new community seeking to gather artists, technologists, and fellow creatives. Visit us at labs.google to explore our products and tools, follow us at @labsdotgoogle, and apply to join our private Discord at goo.gle/labsgooglediscord.

Infinite Wonderland is intended for use by an 18+ audience.

Let’s stay in touch. Get the latest news from Google in your inbox.

Subscribe