Grok Imagine has just received its most substantial update since launch, and the changes reach every corner of the product — including the discovery page that greets users at grok.com/imagine. On 7 August 2026, xAI switched on Imagine Image 2.0 as the new Quality Mode across web, iOS and Android, bringing precise photo editing tools and a Smart Resize feature that recomposes a finished image into any of nine aspect ratios.
The company announced the release from the official Grok account on X the following day, and the pitch was unusually specific: “Edit exactly what you mean with new tools: Magic Wand changes one region and leaves the rest, Segmentation selects precise areas, and Smart Resize lets you tailor to any aspect ratio.” Alongside the editing toolkit, the Grok Imagine discovery page now surfaces a gallery of workflow templates — photo edits, product colour changes, e-commerce shots, professional headshots, icons, character sprites, emojis and merchandise mock-ups — that package the new capabilities into one-click starting points.
This article walks through what the Grok Imagine update actually ships, how the photo editing and resizing features work, what the refreshed discovery page offers, and how the new model measures up against OpenAI’s GPT-Image-2 on the public Arena leaderboards, where it now sits second in the world for both generation and editing.
Table of contents
- What the Grok Imagine Update Actually Ships
- Photo Editing in Grok Imagine: Magic Wand, Segmentation and Cut-Outs
- Resizing Features: Smart Resize Across Nine Aspect Ratios
- The Grok Imagine Discovery Page and Workflow Templates
- How Grok Imagine Ranks Against GPT-Image-2
- What the Grok Imagine Update Means for Businesses
- Availability, API Access and What Comes Next
- Grok Imagine FAQ: Editing, Resizing and the Discovery Page
- References
What the Grok Imagine Update Actually Ships
The 7 August release replaces the previous quality model behind Grok Imagine with Imagine Image 2.0, and the change is live for everyone rather than gated behind a waitlist. TestingCatalog, which covered the rollout on 8 August, confirmed the new model is available on the web-based Imagine service and in both the iOS and Android apps, with API access planned but not yet available.
One model, five new capabilities
The headline additions cluster around control. Where the previous generation of Grok Imagine could generate striking images but offered little means to refine them, Image 2.0 ships five tools that target exactly that gap:
| Tool | What it does | What it replaces |
|---|---|---|
| Magic Wand | Edits one pointed-at region, leaves the rest untouched | Full-image regeneration for small fixes |
| Segmentation | Selects precise areas for targeted changes | Manual masking in external editors |
| Smart Resize | Recomposes one image into nine aspect ratios | Cropping, or regenerating per format |
| Background removal | Exports subjects with clean transparency | Third-party cut-out tools |
| Multi-reference editing | Combines up to five input images in one generation | Manual compositing by hand |
Typography that survives small sizes
Beyond the tools, xAI says the new model plans typography and layout deliberately, keeping small text sharp inside dense, multi-part visuals — posters, menus, infographics and interface mock-ups. That has been a weak point for image models generally, and reviewers who tested the release singled out the text rendering as a genuine step forward. The model also follows detailed multi-part instructions more faithfully and preserves supplied elements across successive generations and edits, which matters once an image goes through several editing rounds.
Photo Editing in Grok Imagine: Magic Wand, Segmentation and Cut-Outs
The photo editing side of the update is best understood as a shift from prompting to pointing. Rather than describing a change and hoping the model applies it in the right place, Grok Imagine users now indicate the place itself.
Magic Wand for one-region changes
Magic Wand is the everyday tool of the set. Point at a region — a shirt, a sky, a product label — describe the change, and the model redraws only that region while leaving every other pixel alone. That single constraint fixes the most common frustration with AI image editing: asking for a small change and receiving a subtly different image in which faces, lighting and composition have all drifted.
Segmentation for precise selections
Segmentation goes a level deeper, letting users select precise areas before applying a change. Where Magic Wand infers the boundary of the region you point at, segmentation gives you the boundary itself to work with — closer to a professional selection tool than a chat instruction. For product photography, where the difference between “the bottle” and “the bottle’s cap” matters commercially, that precision is the feature.
Background removal with transparency
The third editing addition exports any subject with clean transparency, ready to drop into decks, listings or design software. Built-in background removal means a generated or uploaded product shot can leave Grok Imagine as an asset other tools can use directly, without a round-trip through a dedicated cut-out service.
Five references in one generation
Multi-reference editing accepts up to five input images in a single generation. In practice that means a brand’s logo, a product photo, a colour reference and a style frame can all steer one output — the manual compositing step that previously stitched sources together by hand simply disappears. Combined with the model’s element preservation, the feature turns Grok Imagine into something closer to a layout engine than a slot machine.
Resizing Features: Smart Resize Across Nine Aspect Ratios
The resizing features are the update’s most quietly practical addition. Smart Resize takes one finished image and recomposes it into any of nine aspect ratios, from a 1:2 tall banner to a 2:1 wide one — extending or restaging the scene to fill the new frame rather than cropping into it.
Why recomposition beats cropping
A crop discards pixels; a recomposition rebuilds the scene for the new shape. When a square product shot becomes a 9:16 story, Smart Resize fills the vertical frame with coherent new content while trying to keep the important elements intact. The same visual can therefore ship as a story, a square feed post, a banner and a widescreen slide without regenerating from scratch — which also keeps the subject consistent across every placement.
The formats that matter
TestingCatalog’s rundown of the Grok Imagine update names 1:2, 9:16, square, landscape and 2:1 among the supported shapes, with nine ratios in total spanning the two extremes:
| Ratio | Shape | Typical placement |
|---|---|---|
| 1:2 | Tall banner | Display ads, pinned graphics |
| 9:16 | Vertical | Stories, Reels, TikTok, phone wallpapers |
| 1:1 | Square | Feed posts, avatars, thumbnails |
| 16:9 | Landscape | Slides, YouTube, blog covers |
| 2:1 | Wide banner | Website heroes, email headers |
For anyone who publishes the same creative across several channels, this is the difference between one asset and five. It is also a pointed answer to a real workflow cost: teams have long paid for resizing passes in design tools or accepted awkward crops, and the resizing features fold that work into the generation step itself.
The Grok Imagine Discovery Page and Workflow Templates
The discovery page is where the update becomes visible before a user types anything. The Imagine interface at grok.com/imagine opens onto a Discover view that showcases featured templates alongside trending and highlighted community creations, and the 2.0 release stocks that page with workflow templates built on the new photo editing and resizing features.
Templates as packaged workflows
The template gallery covers photo edits, product colour changes, e-commerce shots, professional headshots, icons, character sprites, emojis and merchandise mock-ups. Each template packages a working prompt and settings behind a preview card, so the route from “I need a headshot” to a usable result is one click and one upload rather than prompt engineering. Reviewers covering the release framed the templates as the professionalisation of Grok Imagine — the moment the product stopped assuming its users enjoy writing prompts.
Custom, shareable templates are coming
The template system has been building for months. Back in April 2026, TestingCatalog spotted xAI developing custom, shareable Imagine templates for web users — creators can customise the underlying prompt, configure separate fields for chained image-edit and video options, add a name and description, and generate shareable links, with custom templates appearing in a dedicated tab. A fourth template type, Image Reference Edit, was seen in development using an @-mention syntax for referencing multiple images — a natural companion to the five-reference editing that Image 2.0 has now shipped.
A discovery feed beyond the web
The same reporting spotted a separate vertical discovery feed in development for Grok on iOS, designed to surface Imagine videos generated through Grok and posted to X, sourced from native X content rather than fresh uploads. Taken together, the direction is clear: the discovery page is becoming the front door of Grok Imagine, stocked with templates on the creation side and a community feed on the consumption side — the same quiet, ship-first cadence we documented when xAI introduced its Voice Agent Builder on the console and when the mysterious Ox Alpha model surfaced on an evaluation platform with no attribution.
How Grok Imagine Ranks Against GPT-Image-2
xAI shipped this release with a scoreboard attached. On the public Arena leaderboards updated for 7 August, Imagine Image 2.0 placed second in the world in both text-to-image generation and image editing, behind OpenAI’s GPT-Image-2 in each case.
The numbers behind the ranking
The reported Arena figures put Grok Imagine at 1,439 Elo in image editing against GPT-Image-2’s 1,463 — a gap of just 24 points — and at 1,320 in text-to-image against 1,380, a 60-point gap. The improvement over xAI’s own previous quality model is larger than either gap: the earlier grok-imagine-image-quality model scored 1,390 in editing and 1,228 in text-to-image, so Image 2.0 gained 49 Elo points in editing and 92 points in generation in a single release.
Reading the leaderboard honestly
Two caveats belong next to those numbers. First, Arena rankings move as new models arrive and votes accumulate, so “second in the world” is a snapshot of early August, not a permanent state. Second, Unite.AI notes the models are listed on Arena under the SpaceXAI name following the companies’ corporate restructuring, so anyone checking the leaderboard should look for that label rather than xAI. Neither caveat changes the substance: in one release, Grok Imagine closed most of the editing gap to the best model on the board while remaining free to use at the point of generation.
What the Grok Imagine Update Means for Businesses
Strip away the leaderboard drama and the update is really about production work. The combination on offer — photo editing that respects the rest of the image, resizing that rebuilds instead of cropping, transparent cut-outs and template-driven workflows — describes the day-to-day of a marketing team more than a hobbyist’s toy.
Where the tools earn their keep
An e-commerce operation can generate a product shot, recolour the product per variant with Magic Wand, strip the background for the listing, and push the same visual through Smart Resize for the story, the feed, the banner and the email header. A recruiter can standardise team headshots from the template gallery. A small studio can keep characters consistent across sprite sheets using multi-reference editing. Each of those was previously a chain of separate tools; inside Grok Imagine it is one surface. The neural network behind the product handles the pixels; the templates on the discovery page handle the process.
How the workflow arithmetic changes
The cost case is easiest to see as a count of handoffs. A typical multi-channel creative today passes through an image generator, a background-removal service, a design tool for the resize pass and a compositor for reference-matching — four tools, each with its own subscription, learning curve and export step. The Grok Imagine update collapses those into a single surface, and the discovery page templates remove the prompt-writing step for the most common jobs on top.
For a small team producing dozens of assets a week, the saving is measured less in licence fees than in the version-control chaos that disappears when one asset flows through one tool. None of that arithmetic requires the model to be the best in the world; it requires it to be good enough everywhere at once, which is precisely what the leaderboard numbers now suggest.
The questions to ask before adopting
The sensible caution is the same as for any fast-moving AI creative tool. Moderation behaviour, brand-safety controls and commercial licensing terms deserve a read before a business commits production assets to the platform, and the absence of a public API today means anything automated must wait. Teams that already have provenance policies for AI imagery will also want to check how exports are labelled. For organisations working through where tools like this fit, our AI strategy consulting practice helps map exactly these adoption decisions, and our earlier look at Grok Voice Think Fast 1.0 shows how quickly xAI’s platform pieces iterate once shipped.
Availability, API Access and What Comes Next
Everything described above is live now for Grok users: Imagine Image 2.0 runs as the Quality Mode on grok.com/imagine and inside the iOS and Android apps, and the discovery page templates are available on the web today. Access is free at the point of use, with the usual platform limits.
The API gap
The one missing piece is programmatic access. xAI lists the API as coming soon with no date given, which means developers cannot yet build the new photo editing and resizing features into their own products through official channels. Developers Digest reports a preview build is already visible on Vercel’s AI Gateway as xai/grok-imagine-image-2.0-preview, which suggests the official endpoint is close — but until it lands, Grok Imagine remains an app-first proposition.
The trajectory
The wider pattern is worth tracking. In the space of four weeks xAI shipped character references and 1080p output for Imagine Video 1.5, then this full image-model replacement with editing, resizing and a template-stocked discovery page. The company builds in public and announces afterwards, if at all — which is exactly why watching product surfaces, as we do in this series, keeps turning up changes before the press releases arrive. Grok Imagine is no longer the novelty corner of Grok; it is being assembled, release by release, into a production creative suite.
Grok Imagine FAQ: Editing, Resizing and the Discovery Page
The update raises the same practical questions from most readers, so here are the short answers, drawn strictly from what the release and its coverage actually state.
Is Grok Imagine free to use?
Yes at the point of use. Imagine Image 2.0 runs as the Quality Mode on grok.com/imagine and in the Grok apps for iOS and Android without a separate purchase, subject to the platform’s usage limits. Heavier usage sits behind Grok’s subscription tiers, and pricing for the forthcoming API has not been announced.
What exactly is the discovery page?
The discovery page is the Discover view that opens when you visit Grok Imagine on the web. It showcases featured workflow templates — the photo edits, product colour changes, e-commerce shots, headshots, icons, character sprites, emojis and merchandise categories described above — alongside trending and highlighted community creations. It is the fastest way to see what the new model can do, because every template card is a working example you can run against your own image.
How is Smart Resize different from cropping an image?
Cropping cuts pixels away to fit a new frame; Smart Resize regenerates the scene to fill it. Move a square visual to 9:16 and the model extends the composition vertically — more sky, more floor, more background — while trying to keep the subject intact. The result looks designed for the format rather than trimmed into it, which is why the feature spans nine aspect ratios rather than offering a crop handle.
Can businesses use the output commercially?
Coverage of the release has flagged the editing tools’ suitability for commercial work, and the template gallery clearly targets business use cases. As with any generative platform, though, the licensing terms, moderation rules and brand-safety behaviour are the user’s responsibility to review — they evolve quickly, and nothing in this article substitutes for reading the current terms before shipping a campaign built on Grok Imagine assets.
When will the API be available?
xAI says the API is coming soon, with no date given. The preview build spotted on Vercel’s AI Gateway suggests the endpoint is in active preparation, but until an official release, the editing and resizing features are only available inside the Grok apps and the web interface.
Does the update change Grok Imagine video generation?
No — Image 2.0 is the image model. Video generation continues on Imagine Video 1.5, which received its own update on 1 August 2026 adding character references and 1080p output. The two models share the same Imagine surface and the same discovery page, which is why the template system spans both image edits and chained image-to-video workflows.
References
xAI launches Imagine Image 2.0 in Grok Quality Mode — TestingCatalog
xAI rolling out custom, shareable Imagine templates for Grok — TestingCatalog
xAI adds character references and 1080p to Imagine Video 1.5 — TestingCatalog
Grok’s image tools get a professional upgrade — The Next Web
xAI Ships Grok Imagine Image 2.0 With Precise Editing and a Top Arena Ranking — Unite.AI
xAI Launches ‘Imagine Image 2.0’ With Advanced Editing Tools for Grok — iClarified
Grok launches Imagine Image 2.0 with advanced editing and workflow templates — AlternativeTo
Grok Imagine Image 2.0 Ships: xAI’s Typography-Aware Image Model — Developers Digest
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.