Skip to main content

Grok Image Pro

xAI's Grok Imagine image model — prompt-faithful generation plus edits and multi-image merge. Its signature trick is fusing up to 3 source images into one coherent output.

Capabilities​

FeatureSupport
Text-to-ImageYes
Image-to-Image (editing / merge)Yes
Max Resolution2K (2048 x 2048)
Reference ImagesUp to 3
Aspect RatiosAdaptive, 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 2:1, 1:2, 9:19.5, 19.5:9, 9:20, 20:9
Quality Tiers1K, 2K
Outputs per RunUp to 4
SeedNo
Negative PromptNo
Inpainting / MaskNo

Multi-image Merge​

Connect up to 3 upstream images and describe how they combine:

Put the character from @[Image 1] into the environment of @[Image 2],
using the color grade of @[Image 3].

Grok treats the sources as equal-weight inputs rather than "one base + hints", which makes it unusually good at composites: character transplants, product-in-scene shots, style fusions.

Adaptive & Device Aspect Ratios​

Two things stand out in the ratio list:

  • Adaptive — the model picks the ratio that best fits the prompt (or the input image in edit mode).
  • Phone-screen ratios (9:19.5, 19.5:9, 9:20, 20:9) — native support for full-bleed mobile wallpapers and app-store screenshots without cropping.

Batch Outputs​

A single run can return up to 4 variants. Use this for cheap A/B exploration — pick the strongest variant and iterate on it in edit mode.

Prompting Tips​

  • Say what should survive the merge. With multiple sources, name the element you want from each: "the face from @[Image 1], the outfit from @[Image 2]".
  • Adaptive ratio + descriptive scene = best framing. If you don't need an exact canvas size, let the model choose.
  • No seed means no exact reruns. Save variants you like — the same prompt will not reproduce them.

Limitations​

  • Max 3 reference images (use Nano Banana 2 for larger reference sets)
  • No seed / determinism
  • No negative prompt, no inpainting
  • 2K ceiling — no 4K tier

See Also​