Elon Musk’s AI company has officially rolled out Grok Imagine, xAI’s image and video generator, to all SuperGrok and Premium+ X subscribers on its iOS app. And true to form for Musk, who positions Grok as an unfiltered, boundary-pushing AI, the generator allows users to make NSFW content.
Grok Imagine, which promises to turn text or image prompts into a 15-second video featuring native audio, has a “spicy mode” that allows users to generate sexually explicit content including partial female nudity . There are limits to how explicit one can get. Many of our spicier prompts – made in the name of Journalism! – generate blurred out images that are “moderated” and therefore inaccessible. We were, however, able to generate semi-nude imagery.
The NSFW content is unsurprising for xAI, given the release last month of a raunchy, hyper-sexualized anime AI companion. But just as Grok’s unrestrained nature was entertaining until it started spewing hateful, antisemitic, misogynistic content, Grok Imagine could be poised to bring its own set of unintended consequences.
CNBC first reported the existence of a “spicy mode” last week after an xAI employee Mati Roy said in a now-deleted post on X: “Grok Imagine videos have a spicy mode that can do nudity.”
TechCrunch has reached out to xAI for more information.
The existing limitations with Grok Imagine are somewhat heartening given the model also lets you create content of celebrities – anyone from Donald Trump to Taylor Swift – and there appears to be additional restrictions on those. For example, TechCrunch tried, and failed, to generate an image of Trump pregnant. Grok Imagine only generated images of Trump holding a baby or next to a pregnant woman.
While it’s still early days for Grok Imagine, which aims to compete with incumbents like Google DeepMind, OpenAI, Runway, and Chinese rivals, the images and videos generated of humans are still a bit lost in the uncanny valley, with waxy-looking skin that verges on cartoonish at times.
Still, the generator is impressive. It produces images in seconds from a text prompt, and continues to auto-generate new images as you scroll. Those images can then be animated into stylized videos. The user-interface is also seamless and intuitive.
Musk said on X that the model would “get better every day.”
