In the continued battle for AI market dominance, X (formerly Twitter) has announced the release of its latest image generation model that it says “can render precise visual details of real-world entities, text, logos, and can create realistic portraits of humans.”
Code-named Aurora, the xAi branch of the social media giant has confirmed that it is already available to start using in select countries and that it will be accessible by all X users “within a week.”
Elon Musk has suggested on X that the Aurora name will be deleted altogether, with the new image creating technology to simply be integrated within Grok – X’s AI chatbot.
‘Autoregressive Image Generation Model’
Aurora’s roll out was announced in a blog post from xAi, which confirmed that it would be an enhancement to Grok’s existing image creating abilities.
The post says that the tool is ” an autoregressive mixture-of-experts network trained to predict the next token from interleaved text and image data,” which allows it to create photorealistic images based on the user’s text instructions.
This just in! View
the top business tech deals for 2025 👨💻
“We trained the model on billions of examples from the internet, giving it a deep understanding of the world.” – xAI
Alternatively, it can also be used to edit existing images. The blog post shows examples where Grok has been used to make a photo of a cat appear in anime style, adding a cowboy hat to a cartoon, changing a character’s hair color, and adding snow to a scene.
Grok vs The Competition
In terms of unique image generation, xAi says that Aurora is able to generate high-quality images more effectively than Grok’s competitors.
“Grok can now generate high-quality images across several domains where other image generation models often struggle. It can render precise visual details of real-world entities, text, logos, and can create realistic portraits of humans.” – xAI
To prove its point, it shows examples in the blog post where Aurora’s creations are compared to those of OpenAI’s Dall-E 3, Google’s Imagen 3, Ideogram 2.0 and Flux.1 Pro.
There are direct comparisons to each model’s output for prompts under categories such as ‘Entity generation’ (Cybertruck under an aurora), ‘Artistic text’ (Stars in a galaxy spelling “Grok”), ‘Realistic portraits’ (An Asian woman wearing a long, floral dress surrounded by glowing light stones) and ‘Celebrities’ (Nikola Tesla wearing a VR headset).
It appears that some of the prompts have been deliberately chosen to show impressive Aurora-made images where others struggle, with several where the competitor has refused to generate the image altogether.
Guess What Elon is Using Aurora For…
In a pinned post on X from his own handle, Elon Musk lauded the technology as a way to “create awesome memes super fast”.
Create awesome memes super fast using Grok!
Just tap on the square with the slash. https://t.co/kT0Bk2odcX pic.twitter.com/o41mUudbzr
— Elon Musk (@elonmusk) December 10, 2024
The blog post itself highlights meme generation as another potential use case for the Aurora model, showing an image of a result for the prompt ‘Anime Pepe’ as an example.
Musk has reposted numerous examples of meme-style images created by X users since the announcement, with many including the Pepe the Frog character, Teslas’s Cybertruck vehicle or Musk himself in various scenarios.
The blog confirms that some countries already have access to Aurora’s image-making capabilities within Grok, but doesn’t specify which ones. Asking the X chatbot whether it could shed any light on the matter, it responded that ” there is no explicit mention of specific countries where it has been launched”.
