Generative AI models have improved a hell of a lot in the last year alone, with text prompts now able to produce often eerily realistic images. Regardless of concerns surrounding ethics and copyright, AI image generators continue to add new tools – including this impressive (and somewhat worrying) image style mimicking tool in Dall-E 3.
As revealed on Twitter (sorry, X) below, the third generation of Dall-E, which now includes ChatGPT integration, is capable of creating images that echo the style of other images – which could be helpful for maintaining a consistent style of illustration (something that could arguably also be achieved by hiring a single illustrator).
Digital writer @iamfakhrealam shared the tool yesterday, demonstrating how it can be used directly within ChatGPT 4. Simply by including the gen_id of an image created within Dall-E 3 in the text prompt, users can mimic that style with impressive accuracy.
Right now, the tool can only be trained on an initial image created by Dall-E 3 itself – which might go some way to assuage those concerned about AI's ability to breach copyright by recreating the style of other artists. But even so, can we be sure that the initial DALL-E produced image doesn't include aspects resulting from AI's training on artists' work? If not, perhaps this tool has the potential to exponentially grow an ecosystem of egregiously copied art styles.
Indeed, with even Adobe recently coming under fire for Adobe Stock's alleged involvement in so-called copyright infringement, it seems even the most apparently ethical AI models aren't immune from the possibility of treading on artists' toes. From realistic images to actual walking robots, AI has proven capable of creating a dizzying amount of content and objects in recent months – but we arguably mustn't lose sight of the impact on artists.
