Elon Musk approves: Grok has learned to generate 18+ photos and videos

Elon Musk’s company xAI has officially launched Grok Imagine, an AI-based image and video generator. The main feature is the so-called 18+ mode.

The tool is available to all SuperGrok and Premium+ X subscribers in the Grok app. It allows text or graphic prompts to be converted into images that can be further animated into videos of up to 15 seconds. Audio is played in the background of the video, which is also generated automatically.

Grok Imagine allows you to create NSFW content — including sexual content, including partial nudity. This tool is not explicitly pornographic, so it is not without technical limitations: some requests are moderated, and the result may be blurry or completely inaccessible. However, some users have managed to obtain semi-nude images.

The existence of the so-called “acute regime” became known last week after a comment by xAI staffer Mati Roy. In a deleted post, she wrote: “Grok Imagine videos have an acute mode that can do nudity.” As you can see, there are still some restrictions.

Separately, Grok Imagine allows you to create content with famous people like Donald Trump and Taylor Swift. However, there are limitations here as well. For example, it was not possible to create a pregnant Trump. The tool only produced images of the US president standing next to a pregnant woman or holding a child.

The company has shown similar openness before. Last month xAI releases hypersexualized anime waifu It talks flirtatiously and strips down to its underwear, and it can swear and insult you. Like the chatbot itself, the new tool also selectively filters content, which can lead to unpredictable consequences — as it did with sexism and anti-Semitism in Grok.

Grok Imagine is still in early development, but so far it’s lagging far behind competitors like Google Veo 3, OpenAI Sora, or Midjourney. The main reason — Grok Imagine does not create full-fledged videos from scratch as, for example, Veo 3 does. Instead, the user needs to create or upload an image, after which it can be animated for up to 15 seconds.

Currently, the quality of the results varies. Some of the images look unnatural: the skin of the characters is waxy, their expressions look like 3D models from previous years. In many cases, the result looks cartoonish. Nevertheless, Grok Imagine is already generating sufficiently detailed scenes and continues to be actively developed.

Source: TechCrunch