xAI has just added image generation to Grok, a step forward for the platform. The feature is currently available to X users in some countries, and a global rollout is expected in about a week.
The image generator, initially called Aurora but now part of the Grok family uses a fancy autoregressive mixture-of-experts system trained on billions of examples from across the internet. Basically, it can predict the next bits of information by combining text and images, which lets it create much more realistic images than before.
But it's not just about making images from scratch. This system can also take existing images and let users tweak them or use them as inspiration for new creations. xAI says the model does a particularly good job in areas where other generators struggle, like getting text, logos, and human faces just right.
This update comes right after the release of Grok 1.5V in April, which gave the platform its first taste of visual processing. And xAI's not stopping there—they're busy upgrading their Colossus supercomputer in Memphis, which already has 100,000 Nvidia H100 and H200 GPUs and plans to double that number soon.
The timing of this launch is interesting since OpenAI just rolled out its own video generation model, Sora, making it clear that the competition in generative AI is heating up between the big players.
Are you a techie who knows how to write? Then join our Team! Wanted:
- News Writer (Romania based)
Details here