on Toshareproject.it - curated by Bruce Sterling
*When it bleeds, it leads.
(…)
Despite being the hottest name in the game, Rutkowski became one of AI art’s fiercest critics.
In response to feedback from him and other digital artists, a major change was introduced with the release of Stable Diffusion 2.0. Stability AI chose to remove the ability of emulating the style of specific artists, causing some discontent amongst users. The update was declared “nerfed,” as it no longer allowed generation of images in Rutkowski’s unique style. It also had problems reproducing human anatomy, and it required a whole new and more difficult technique for prompting.
The move made the updated model highly unpopular, and cemented Stable Diffusion v1.5 as the king of Open Source Text to Image Generators.
Enter the LoRa, a small, very specific model trained to emulate specific things like art styles, faces, stances or color gradings. With Stability AI omitting Rutkowski’s style in SD 2.0, the art community saw a void and sought to fill it. The result? A Lora trained to recreate Rutkowski’s style, now freely available on Civitai….