X AI that generates almost unrestricted images creates unusual encounters; see some images that went viral

O X announced this week the release of Auroraimage generation model integrated with Grok 2for all users of the platform. The tool was made available on Saturday (7) and was deactivated hours later, but is now available again.

One of its main features, which has generated a lot of repercussion on the web, is the possibility of generating images without all the restrictions related to the numerous filters imposed by different services that try to avoid running into ethical and commercial issues, such as copyright infringement. .

This characteristic of image generation services means that many people turn to options that run locally, such as Stable Diffusion, guaranteeing more creative freedom. Surprisingly, at least for now, Aurora works in much the same way. The tool has fewer barriers when compared to other options.

According to X, Aurora was trained with billions of images available on the internet, enabling her to generate highly detailed representations faithful to the prompt. Examples published include fictitious photos of famous people, allowing some unusual encounters to be generated, such as the one in the cover image of this post.

How does Aurora AI work?

Aurora AI uses an autoregressive network capable of processing multimodal data — combining text and images. In addition to creating photos from descriptions, the model also allows you to edit existing images, such as transforming a daytime scene into a nighttime one or adapting a landscape to different weather conditions.

Another interesting feature is the possibility of uploading personal photos, allowing the user to see themselves alongside famous figures, or in stylized versions, such as caricatures or cartoons.

Format limitations and daily number of images

The maximum image size is 1024 pixels, with no support for panoramic formats. Furthermore, there is censorship of specific content, such as sexualized representations or scenes of extreme violence. The generated images also receive a watermark in the lower right corner.

Users can send just 10 messages every two hours, for a total of up to 120 prompts per day. This number applies to both text and image generation requests. Access is also only available through the web version of X, by clicking on the button corresponding to Grok in the left corner of the screen and then entering the prompt. The tool can interpret commands written in Portuguese.

Ethical debate

Some people joked that many of these images will go viral in Zap groups, especially those involving politics. A dish full of laughs, but also for attempts to spread misinformation.

Check out some unusual images generated by Aurora that went viral below: