Grok Imagine adds preset palette recolouring and in-app crops as Image 2.0 lands on third-party playgrounds

In-flow colour and crop controls land inside Grok Imagine

Editors using Grok Imagine can now recolour an image using either preset palettes or colours sampled directly from the image itself, and crop output without leaving the Imagine interface. Both functions are positioned as practical editing utilities rather than a new model release, and they target the common friction of round-tripping assets between a generator and a separate editor.

Grok Image 2.0 ships with designer typography, magic-wand region edits, and Smart Resize

The Imagery 2.0 release packages several editing capabilities into one surface: designer-grade typography, magic-wand region edits, multi-reference generation, and Smart Resize, alongside prompt examples that show the intended use cases. Coverage from more than one source frames Image 2.0 as a precision-editing product, with explicit prompt-and-edit patterns laid out for practitioners.

Third-party playground exposes 1K and 2K renders, Low and Medium tiers, 13 aspect ratios, and reference edits

A third-party AI playground has opened up Grok Imagine 2 with both 1K and 2K resolution outputs, Low and Medium quality tiers, 13 aspect ratios, and reference-image edits, putting the model in front of a creator base that already uses that platform's editing surface. For practitioners targeting specific export presets, the 13-aspect-ratio set is the most concrete planning input in today's evidence.

Provider and API surfaces add Image 2.0 with reference-guided edits and visible credit accounting

An image-generator listing presents Grok Image 2.0 with reference-guided editing, credits visible to the caller, and explicit limits stated up front. Separately, an API listing for the same family describes a Grok Imagine API covering image, video, and speech models, with 1K and 2K stills across a stated number of aspect ratios; cross-check the exact aspect-ratio count at the API surface, because the third-party playground article lists 13 aspect ratios and the API summary lists a different count, so the precise supported set is not agreed across the two items and should be confirmed at the endpoint before targeting exports.

Workflow note: building a matched image set from a single input

A workflow guide for a different model family — Flux — argues for designing a single prompt-plus-reference setup that yields a finished set of images, rather than regenerating one image multiple times until a set matches. The article is presented as a working example of multi-reference generation, which overlaps directly with the multi-reference capabilities now exposed in Grok Imagine 2.0 and gives practitioners a tested pattern to port across models.

Editor-side follow-ups for practitioners on August 25, 2026

Three concrete checks are worth making today. First, confirm whether your target preset — 1K or 2K, Low or Medium — matches the aspect ratio you need; the third-party playground lists 13 aspect ratios while a separate API summary lists a different count, so verify the supported set at the exact endpoint you will call. Second, when prepping source images for recolouring with Grok Imagine's new palette tools, drop the original into the Image Resizer so the sampled colours are pulled from a pixel-exact crop rather than a scaled-down preview — the Resize an Image to Exact Pixel Dimensions in Your Browser guide covers the workflow.

Third, when shipping final stills to the web, run outputs through the Compress Images for Web: A Practical Guide so that 2K renders do not blow up payload budgets.