01
What is Grok Image 2?
Grok Image 2 is the common name for xAI’s Imagine Image 2.0, released on August 7, 2026 as the new Quality Mode in Grok on the web, iOS, and Android. It covers image generation plus targeted editing.
Grok Image 2 · Released August 7, 2026
Grok Image 2: precise image generation and editing
Grok Image 2 is xAI’s new Quality Mode for Grok on the web, iOS, and Android. This independent workspace documents the release and uses the current Grok Imagine API while its developer API is still listed as coming soon.
Independent workflow preview
Generation · editing · references · smart resize
Use English for better results
What Grok Image 2 adds
Layout and small-text planning
xAI says Grok Image 2 plans typography and layout so dense, multi-part visuals remain coherent and small text stays sharp. Important copy still needs human review before publication.
Targeted editing controls
The release adds a magic-wand region editor, segmentation-based selection, background removal with transparent output, and multi-reference editing in the consumer product.
Smart resize and reuse
The model can fill a new frame for nine documented ratios, from 1:2 and 9:16 to 16:9 and 2:1, while aiming to preserve the subject and visual direction.
What · Why · How
Grok Image 2 in three practical questions
Start with the released product, the evidence behind its capabilities, and the correct way to test it. Image 2.0 consumer features and the current Grok Imagine API are related but not interchangeable.
02
Why does the release matter?
The release groups layout and small-text planning, selection-based edits, five consumer reference images, transparent cutouts, and nine reframing ratios in one workflow. xAI also reported second place in two Arena categories on release day; that dated claim is not an independent benchmark.
03
How should you use it?
Use Grok Quality Mode for the actual Grok Image 2 consumer model. Write testable constraints, keep references purposeful, edit the smallest necessary region, and inspect text and geometry. This site uses the current public Grok Imagine API until xAI publishes Image 2 API access.
What Grok Image 2 adds
The verified Image 2.0 release focuses on instruction following, planned typography and layout, targeted edits, reusable references, and reframing. The sections below separate the consumer release from the current developer API.
01
Instruction-following for dense compositions
Grok Image 2 is designed to keep multi-part instructions, type hierarchy, and layout relationships together. Specify the subject, required text, placement, style, and exclusions instead of relying on vague adjectives.
02
Region and segmentation edits
The release adds pointed region edits and segmentation selection. That is useful when one object, color, or area needs to change while the rest of the image should remain stable.
03
Multi-reference editing
The Grok Image 2 consumer workflow accepts up to five input images. The current API used by this site documents up to three references, so the upload control follows that lower limit.
04
Transparent background output
Background removal can export an isolated subject with transparency. Inspect hair, glass, fabric, and soft shadows at full size before using the result in production.
05
Nine smart-resize ratios
xAI demonstrates 1:2, 9:16, 2:3, 3:4, 1:1, 4:3, 3:2, 16:9, and 2:1 reframing. The model fills the new frame instead of applying a simple crop.
06
Clear API boundary
Grok Image 2 API access is not public at launch. This workspace uses documented Grok Imagine models and will not label those outputs as Image 2 results.
A practical Grok Image 2 workflow
Use Grok Image 2 in Grok Quality Mode for the released consumer experience. In this independent workspace, generation and editing run on the currently documented Grok Imagine API until xAI opens the Image 2.0 API.
01
Write the constraints first
State the subject, composition, exact text, visual style, aspect ratio, and elements that must not change. Concrete constraints make review faster than general quality terms.
02
Add only useful references
Use references for identity, product shape, palette, or layout. This workspace accepts three current-API references; the Grok Image 2 consumer editor accepts up to five.
03
Edit the smallest necessary area
Prefer a targeted region or segmented object when the rest of the result is already correct. Check faces, labels, logos, fingers, product geometry, and transparent edges.
04
Resize, compare, and export
Create the ratios the channel actually needs, compare subject preservation across versions, and keep the approved prompt and references with the exported asset.
Public posts on X
What users report after trying Grok Image 2
These public X posts were observed on August 9, 2026 and selected because each describes a concrete generation, comparison, or editing experience. The set includes positive, mixed, and critical reactions.
Positive
“Grok Image 2 is impressive. Generate an image, inspect its selectable segments, then edit, regenerate or animate from one surface.”
Highlights the single-surface flow from generated artifact to selectable parts and follow-up actions.
Simone
@mustbesimo ·
Mixed
“生图：细节以及对于世界理解能力还是gpt强 编辑：审美gpt稍微占一些优势，但是图像编辑能力Grok强 中文小字：两者都没有任何问题”
After seven comparisons, the author preferred GPT for generation detail and aesthetics, but Grok for image editing and faster output.
码农暖爸
@Delroy715 ·
Mixed
“Grok 直出效果已经很能打，质感、光影、氛围都在线。短板是复杂细节容易糊，精细度还不如 GPT Image 2。”
Praises direct output, lighting, atmosphere, and local editing while flagging weaker precision in complex details.
羊羊羊
@encryptjustice ·
Critical
“Have a coffee-making scene and delete the coffee art from ONE spot? It deletes every single coffee art instance across the entire image.”
Reports a targeted deletion affecting repeated elements elsewhere in the image—a specific failure mode worth retesting.
Silverion Stormbringer
@Lord_Silverion ·
These are individual user reports, not reproducible benchmarks or proof of general model quality. Quotes remain in their original language and link to the source post; authors may later edit or delete them. Grok Image 2 results should still be tested against your own prompts and review criteria.
FAQ
Frequently asked questions
Verified answers about Grok Image 2 availability, editing controls, references, resize options, benchmarks, and the current API boundary.
What is Grok Image 2?
It is the common search name for xAI’s Imagine Image 2.0 release. xAI describes the model as precise image generation and editing, and made it generally available as the new Quality Mode on Grok web, iOS, and Android on August 7, 2026.
Where can I use Grok Image 2 today?
The official release lists grok.com/imagine and the Grok iOS and Android apps. Availability refers to the consumer Quality Mode. This website is independent and is not an xAI product.
Does Grok Image 2 have a public API?
Not at launch. The xAI release page says API access is coming soon. The generator on this site uses the existing grok-imagine-image-quality and grok-imagine-image API models, not an undisclosed Image 2.0 endpoint.
What changed in typography and layout?
xAI says the model plans typography and layout so dense, multi-part visuals hold together and small text is sharper. That is a capability statement, not a guarantee: proofread every name, number, label, and legal line.
How does precise editing work?
The consumer release includes a magic-wand tool for a pointed region and segmentation for selecting an exact area. The stated goal is to change the selected content while leaving the rest intact.
How many reference images are supported?
The Image 2.0 consumer release accepts up to five input images in one generation. Current Grok Imagine API documentation describes up to three reference images, so this site limits uploads to three.
Can it remove a background?
Yes. xAI lists background removal with transparent output as an Image 2.0 editing tool. Review fine edges and semi-transparent materials before placing the cutout in another design.
Which resize ratios are documented?
The release demonstrates 1:2, 9:16, 2:3, 3:4, 1:1, 4:3, 3:2, 16:9, and 2:1. Smart resize fills the added frame rather than only cropping the original.
How did xAI report its Arena rank?
xAI reports that its model ranked second in both Text-to-Image Arena and Image Edit Arena using overall Elo on August 7, 2026. Rankings change, so the release date and leaderboard snapshot matter.
What templates shipped with the release?
xAI lists workflows for photo editing, product color changes, product posters, collage, mascots, background changes, ecommerce photos, UGC photos, headshots, icons, character sprites, UI kits, emoji, and merchandise.
Can it keep a visual world consistent?
The release demonstrates a character, locations, and props generated separately while holding one look. Treat consistency as something to test across a set, using the same references and explicit style constraints.
What model does this site call?
This site exposes Grok Imagine Quality and Grok Imagine through the documented xAI API. It will not route requests to Image 2.0 until xAI publishes an API model identifier and access terms.
What does the current Grok Imagine API support?
Current documentation covers text-to-image generation, natural-language editing, batch outputs, aspect-ratio control, 1K and 2K output, and reference-image input. Capability and pricing depend on the selected API model.
How much does the current API cost?
xAI lists Grok Imagine Quality at $0.05 per 1K output and $0.07 per 2K output, plus image-input cost. Grok Imagine is listed at $0.02 per output. Site credits and subscription fees are separate.
Is grokimage2.app affiliated with xAI?
No. grokimage2.app is an independent workflow site. Grok, Grok Imagine, and Imagine Image 2.0 are xAI names. Official availability and model details come from xAI sources.
Review a Grok Image 2 brief in the workspace
Generate or edit with the current Grok Imagine API, keep the prompt and references, and verify every result before use. Image 2.0 API access remains pending.