Create surreal Pokémon lookalikes of Jeff Bezos, The Rock and more with AI

“It was really fun to see all the stuff people do with it,” Pinkney wrote in an email to The Washington Post. “I’m always impressed with how it seems to capture the essence of famous characters, but in a very weird Pokémon form.”

Kids Who Grew Up With Pokémon Are Raising Their Own Trainers

Very weird is an apt way of saying it. People have used it to generate all kinds of Pokemon with the uncanny resemblance of famous people. Here’s what he spat using Hollywood superstar Dwayne Johnson as a guest:

Here are some Pokemon variants from Amazon founder Jeff Bezos. (Bezos also owns the Washington Post).

Here, by the way, is Wordle as a Pokémon.

But the generator is also capable of producing some surprisingly poignant creations, as can be seen in this mosaic of sample images shared by Pinkney. The first Pokémon in the line was generated from the prompt “Girl with a Pearl Earring”, the famous painting by Dutch master Johannes Vermeer. See if you can guess the others.

This is a Pokémon created from “The Return of the Prodigal Son”, one of the best-known paintings by another Dutch master, Rembrandt Harmenszoon van Rijn.

Pinkney was inspired to create a Pokemon generator art by her 6-year-old daughter, who recently became a huge fan of the franchise.

“I thought it would be a fun project to train a model so she could describe her ideal Pokemon,” he said. “Looks like it worked!”

Text-to-image art generators work through a process called deep learning, in which algorithms make predictions and complete tasks in a process that mimics neurons in the human brain. In the case of AI-generated art, generators draw from a database of existing images and artwork to put together a discrete piece based on the user’s prompt. Pinkney explained that his own creation is adapted from an open-source deep learning model called Stable Diffusion, which already has large datasets of information. Text-to-Pokémon works by matching datasets from Stable Diffusion to a dataset of 850 Pokémon images from a previous university research project, which Pinkney deposited using a captioning system. automated to categorize each image with a text description.

He used AI to win a fine art contest. Was it cheating?

This is why Text-to-Pokémon can create more or less convincing Pokémon facsimiles of well-known public figures such as Dwayne Johnson or Taylor Swift, but will generate more abstract results for most people using their own name (apart from unless they are also high profile celebrities). Pinkney posted a detailed article on how his generator works on Lambda’s blog.

If you want to try your hand at creating some of your own Pokemon, you can access them here. You’ll need a GitHub account to run the template through the website, but once that’s set up, all you need to do is enter a phrase (Keanu Reeves, “The Last Supper”, “Doom Eternal”, etc.) and press the submit button to generate one or more Pokémon.

Pinkney intends to continue working on Text-to-Pokémon. He mentioned the possibility of getting more Pokemon images to augment his dataset with more detailed captions, perhaps in conjunction with someone with more Pokemon knowledge. He is also interested in developing art generators capable of compiling images beyond text prompts.

“I think people have only seen the tip of the iceberg so far with these techniques!” Pinkney wrote. “Personally, while I think the text prompts are a nice and easy interface to tell the AI ​​what type of image you want, I’m really interested in how to add other ways to influence and to control the output, to make it much more useful for artists and creatives to guide it to deliver the kind of images they are looking for.