ingenious AI lets you create your own Pokémon

Text-to-Pokémon is an AI that lets you create Pokémon from text. You can enter the name of a character, celebrity or even a physical description, and let the artificial intelligence generate pocket monsters in the pure style of Nintendo’s famous franchise. Find out all you need to know about this brilliant tool, and how to use it!

For several months now, AI-based Text-to-Art image generators are enjoying viral success. From MidJourney to DALL-E, these tools can automatically create amazing drawings from just a few words. To find out more, read our complete guide to Text-to-Art AIs.

These artificial intelligences are being exploited by graphic artists and designers, and one architect has even used it to create the skyscrapers of the future. However, these programs also generate controversy, as when an AI-created drawing won an art competition in Colorado.

What’s more, as these AIs are trained on existing images, they can cause copyright problems. This is why Getty has banned this type of content from its image bank…

However, far from polemics and other debates, machine learning researcher Justin Pinkney of Lambda Labs has created an image synthesizer that will make everyone agree: Text-to-Pokémon.

As its name suggests, this tool uses AI to create a new Pokémon from a text description. This can be as simple as a word, an object or even the name of a celebrity or famous character.

Simply enter a text, and the artificial intelligence model automatically generates a character in pure Pokémon style by drawing on its database of images and legends.

A version of the AI Stable Diffusion trained on Pokémon

This program is executed with an API on Replicate, the platform for training your own models machine learning models and run them in the cloud. It is a version of the AI image generator Stable Diffusion, trained specifically on Pokémon.

To create the model, Justin Pinkney “refined Stable Diffusion by training it on Pokémon as a dataset. The AI was trained on captioned BLIP images of Pokémon with two NVIDIA RTX A6000 GPUs on the Lambda GPU Cloud for around 15,000 steps.

According to the researcher, ” Stable Diffusion is an excellent generalist model, but getting a certain style of result can be difficult and usually requires a lot of prompt engineering. “. The ” prompt engineering “consists in modifying the texts repeatedly until the AI produces a satisfactory result in terms of style or appearance.

As an alternative, the expert explains that ” refining the model itself is a simple approach to focusing only on what you want, provided you have data “. Thanks to this process, Text-to-Pokémon generates only Pokémon.

If you’d like to know more about Stable Diffusion’s refinement process for Text-to-Pokémon, Justin Pinkney has posted a full explanation on GitHub at this address.