Digital artist Greg Rutkowski, known for his vivid and surreal style, finds himself at odds with the world of artificial intelligence (AI) art. Rutkowski has expressed his opposition to AI-generated art, but that hasn’t stopped AI creators from persisting in mimicking his unique artistic style.
Rutkowski’s name has become the most popular keyword used by AI artists seeking to replicate a specific artistic style. However, he wants nothing to do with this technology. In response to his vocal criticism, Stability AI, the creators of the popular AI image generator Stable Diffusion, removed Rutkowski’s work from their dataset. But it seems that the AI art community has found a workaround.
They have created a tool that can mimic Rutkowski’s style against his wishes. Since Stable Diffusion is open source, both Rutkowski and Stability AI are powerless to prevent its use. This has plunged Rutkowski into a digital quagmire, where it becomes difficult to distinguish between AI-generated pieces and his genuine works.
Despite his popularity within the AI art scene, Rutkowski remains one of the technology’s fiercest critics. In response to feedback from him and other digital artists, Stability AI made a major change in the release of Stable Diffusion 2.0. They removed the ability to emulate the styles of specific artists, which caused some discontent among users. This update, deemed nerfed by many, no longer allowed the generation of images in Rutkowski’s unique style and encountered difficulties reproducing human anatomy. It also required a more complex technique for prompting.
The introduction of Stable Diffusion 2.0 made the previous version, Stable Diffusion 1.5, the preferred choice among Open Source Text to Image Generators. However, the art community saw a void with Rutkowski’s style missing in the updated model and sought to fill it. They trained a model called LoRa specifically to recreate Rutkowski’s style, and it is now freely available on Civitai.
The introduction of LoRa sparked controversy, with debates on Civitai and Reddit. Some argued that it was ethically wrong to create a model based on Rutkowski’s art when he explicitly asked people not to train on it. However, others supported the move, noting that Rutkowski’s art had already been circulated in Stable Diffusion 1.5 for years.
Lykon, the creator behind the customized SDXL model, defended the decision, stating that since Rutkowski’s name was frequently used in prompts for Stable Diffusion 1.5, training a LoRa model seemed like a suitable solution. He claimed to be acting in the greater interest of the art community.
While Lykon stated he would remove the LoRa model if Rutkowski requested it, he currently believes having an accurate immortal depiction of Rutkowski’s style benefits everyone.
The evolving world of AI and art blurs the line between innovation and infringement. The concepts of styles, decentralization, and open source play a significant role. The art community remains unafraid to challenge the boundaries of the digital canvas.
In the dynamic interplay between artists and AI, life often imitates art, and vice versa. If Rutkowski can’t beat them, perhaps he could find a way to paint them?
