The artificial intelligence of image generation has been entitled to an update, and it is more precise than ever – but Internet users regret certain novelties.
Stable Diffusion already had its version 1, which made it possible to create impressively realistic images, just from an instruction, whether it was a mountain landscape, a portrait, or even a still life. Since November 24, version 2 of the artificial intelligence is there, and it is even more spectacular than before.
More competent AI and higher quality images
It was StabilityAI, the company behind Stable Diffusion, which announced the new version by press release. Among the main novelties, the improvement of the quality of the images created – which until then were small and regularly blurred –, and better management of creations from a model. Depths of field are also better.
Another important factor: the AI is equipped with a better ability to detect NSFW content, which means that it filters even better ” adult content than the previous version. It thus becomes even more difficult to create pornographic content by artificial intelligence.
One downside though: as The Verge spotted, some users are complaining on Reddit that the new changes would make it more difficult to generate images that copy other artists’ style. The ability to create tables in the style of Van Gogh by Picasso or Frida Khalo was indeed one of the favorite features of users. ” They clamped down on the AI “said a user on Reddit.
The artificial intelligences used to generate images are getting stronger and stronger – to the point where it is sometimes difficult to tell the difference between a work created by a human artist and one produced by AI. Numerama wanted to do the test, and tried to trap two specialists in art and artificial intelligence: will you be able to do better than them?