Stable Difussion’s AI and the rest of the AI models capable of generating images are giving us a lot of joy. Some are beginning to combine it with Photoshop. There are already AI-generated image banks, such as Lexica, and there are even beginning to be problems and debate over copyright when using this type of image.
The tool that we did not see coming was this model capable of creating pokémon from words, even daring to turn famous characters into these creatures. The AI is open to any user with a GitHub account (if you don’t have one, you can sign in with Google).
If it can be written, it can be a Pokémon
Beyond the meme, the technology behind this model is more than interesting. It is a model created by Lambda Labs and trained with images of Pokémon. The creators claim that the total cost of creating this model it was only 10 dollarsone of the possible reasons why it is open to all users, unlike DALL-E or Midjourney.
The operation is quite simple. We just have to enter the word (prompt), always trying to make it correspond to a well-known character. It can be both a real character (politicians, athletes, etc.) as well as a fictional or cartoon character. We can choose a total of four outputs, although if we want it to work quickly, it may be convenient to go one by one.
By clicking on any image we can copy it, save it as an image and any other action that can be carried out on any photo on the web. Currently the model has been run 185,600 times. It is not the most useful AI in a practical sense, but we can have a good time with it.
More information | replicate