Grok Imagine Image 2.0 brings precision editing, crisp text rendering, smart resizing, and multi-reference image generation to xAI’s image tool.

Grok Imagine Image 2.0 brings precision editing to xAI’s image generator

Grok Imagine Image 2.0 has arrived with a clear focus on practical image creation rather than simple text-to-image generation. xAI says the new model improves instruction following, text rendering, factual accuracy, and consistency while adding a set of editing tools aimed at designers, marketers, e-commerce teams, and everyday users.

The model is available as Quality Mode on Grok’s Imagine website and through the Grok apps for iOS and Android. API access has been announced as “coming soon,” although xAI has not provided a specific release date.

More controlled image editing

The main improvement is greater control over individual parts of an image. The new Magic Wand tool lets users select an area and change only that section while preserving the rest of the composition. A segmentation tool offers more precise object selection, which should be useful when editing products, clothing, backgrounds, or individual elements in a busy scene.

Grok Imagine Image 2.0 also supports background removal, allowing users to export a subject with a transparent background. That could save time for people creating product listings, social media graphics, marketing assets, thumbnails, and presentation materials.

Up to five reference images

The model can combine up to five reference images in one generation. For example, a user could provide a product photo, a preferred model, a background, a colour reference, and a layout example, then ask Grok to combine them into one visual.

This multi-reference workflow reduces the need for manual compositing and gives creators more control over the final result. It could be especially valuable for e-commerce campaigns, where the same product needs to appear in multiple styles and environments.

Smart Resize and sharper text

The new Smart Resize feature expands an existing image into different aspect ratios instead of simply cropping it. xAI says Image 2.0 supports formats ranging from 1:2 to 2:1, including common portrait, square, landscape, and banner layouts. The model fills the additional space while attempting to preserve the original subject and composition.

xAI has also placed greater emphasis on typography and layout planning. The company says the model is better at rendering small, readable text and maintaining complex designs with multiple visual elements. That improvement could make AI-generated posters, advertisements, infographics, product cards, and social media graphics more usable.

Templates for real-world work

Image 2.0 includes preconfigured templates for recurring tasks such as product photography, professional portraits, e-commerce images, game assets, icons, emojis, merchandise, and promotional posters. Instead of starting with a blank prompt, users can select a workflow and provide the required images or instructions.

xAI is also demonstrating a system for creating characters, locations, and props separately while maintaining a consistent visual style. That could help creators develop storyboards and reusable assets for future AI video projects.

Early benchmark results

Early third-party reports place Imagine Image 2.0 near the top of image-generation and editing benchmarks. The model reportedly scored an Elo rating of 1,439 in the Image Edit Arena, behind OpenAI’s GPT-Image-2 at 1,463. These are ongoing community benchmark results, not a complete measure of real-world performance, but they indicate that xAI is targeting the leading image models directly.

Summary: Grok Imagine Image 2.0 turns xAI’s image tool into a more capable editing platform, adding region-level changes, background removal, multi-reference generation, Smart Resize, and improved text rendering. Its success will depend on how reliably these features perform in everyday creative and commercial workflows.