A new machine-learning model by Lambda Labs allows users to input any prompt they desire and generate a unique Pokemon. While the results are complex, the process is relatively simple. The model pulls from a database of captioned images of Pokemon and generates a new image based on the user's prompt. While examples accompanying the model include Barack Obama, Donald Trump, and My Neighbor Totoro, another natural place to begin would be to pair two of Nintendo's largest franchises.
Similar to the DALL-E program, which uses artificial intelligence to generate a user's image, the model by Lambda Labs yields better images with more specific prompts. Using a simple prompt such as «Link» will likely produce a fairly abstract image compared to more precise inputs like «Link from Breath of the Wild as a fox.» The resulting image will always be rendered in a Pokemon style regardless.
Shiny Pokemon Looks Suspiciously Like Sith Kermit Meme
ThePokemon version of Princess Zelda is fairly spot on, especially considering the prompt was intended to render Zelda in her Ocarina of Time version. As the image produced is clearly based on young Zelda in Ocarina of Time, it's appropriate that the result is a Pokemon that looks like a base evolution.
This Pokemon looks vaguely like it could fit as a Psychic-type—befitting of Zelda's increasingly prominent magical abilities that allow her to stand against The Legend of Zelda's strongest villains. The only sort of anomaly in the rendering is the overwhelming green used for Zelda's Pokemon, though this is likely pulled directly from Link's tunic.
In another astonishingly accurate rendering, Ganondorf's wicked expression and reddish-yellow eyes remain perfectly intact in a Pokemon that looks fairly
Read more on gamerant.com