Login
iscover
Share this article on:
Powered by
MOMENTUM
MEDIA
Powered by MOMENTUMMEDIA
xAI has officially launched Grok’s new image generator, Grok Imagine, which is set to rival the likes of Google Veo 3 and OpenAI’s Sora, but stands out with the ability to generate Not Safe For Work (NSFW) content.
Grok Imagine has been rolled out for Premium+ X and SuperGrok subscribers on iOS, and sets itself apart from the competition with its “spicy mode”, which allows users to generate 15-second long videos containing content that is sexually explicit.
X and xAI CEO Elon Musk himself demonstrated the capabilities in a post on X, which features a 6-second clip of an angel wearing minimal clothing.
Imagine with @Grok pic.twitter.com/UIay5yNp97
— Elon Musk (@elonmusk) August 4, 2025
It's unsurprising that xAI has moved in the direction of NSFW content generation, considering X and Musk’s stance on free speech and not removing content that may be deemed inappropriate for at least some audiences. The company also released a sexualised anime AI companion last month.
However, considering that Grok recently went on a racist tirade, calling itself “MechaHitler”, its ability to generate images and videos could lead to some troubling consequences.
However, the bot has limitations regarding the generation of content including celebrities. While images of celebrities like Donald Trump are allowed, TechCrunch tried and failed to generate images of Trump pregnant, only instead able to generate images of him holding a baby or standing next to a pregnant woman.
Still, Musk stands by the product, saying that Grok Imagine is improving on a daily basis, with video and image generation speeds getting faster.
“Grok Imagine is now making videos in ½ to ¼ the time that major competitors to make a single image!” he said on X.
That being said, Grok Imagine still lags behind its rivals, according to a report by Mashable journalist Timothy Beck Werth.
“How does Grok Imagine compare to other AI video tools? In my initial tests — not great! I started with a simple prompt based on a viral AI video trend: ‘Security camera footage of rabbits jumping on a trampoline at night’.” he said on X.
“Grok Imagine completely failed to understand the "security camera" part of the prompt. I tried modifiers like "Ring camera footage" or "black and white surveillance footage," but the results were identical.”
Werth added that when the same prompt was entered into Sora, the results were much better, but it was Google Veo 3 that generated the best video.
Google Veo 3 aced the assignment. The video has the kind of errors common to AI video, but Veo knew what it was doing. pic.twitter.com/wOsKDONm6T
— Timothy Beck Werth (@beck_werth) August 4, 2025
Be the first to hear the latest developments in the cyber industry.
