Grok Imagine Image

Cheap prompt-to-image at 2 credits, from the house that makes Grok Imagine Video 1.5, which is the reason it is here. If a clip is going to be animated on the Grok video model, its first frame made on the Grok image model tends to need less persuasion than one imported from another family's look.

Its reference channel is unusual and worth understanding before you plan around it. Where most of this catalog takes a tray of up to seven references, this takes exactly one image, and it does not treat it as a reference at all: it edits it. Hand it a picture and a prompt and you get that picture changed, not a new picture influenced by it. That makes it a good editing tool and a poor consistency tool, because one photo cannot show a subject from three angles the way a seven-slot tray can.

Editing is metered separately from generating: the picture you send in costs about half a credit per megapixel on top of the 2, so an edit runs 2.5 to 3.5 credits depending on how large the input is. Seven aspect ratios, JPEG out, and no resolution or quality dropdown at all. There is exactly one way to run it. The Quality variant is where the resolution ladder lives.

Describe what you want in your own words and Kuro writes the prompt for you. If you would rather write it yourself, this is what Grok Imagine Image responds to.

Two modes: a prompt alone makes a picture, and a prompt plus one picture edits that picture. When editing, write the change rather than a new brief — the picture carries everything else.

Prompt in full sentences and name the medium and the light. It responds to cinematic direction: "shot on 35mm, overcast daylight, shallow depth of field".

A still made here can open a Grok Imagine Video clip as its first frame.

The same advice covers Grok Imagine Image Quality. Their pages note what differs.

What it can't do

• One image seat: it edits a single picture and does not take a set of references. Tag a saved Subject that has a picture and that picture becomes the one being edited; to edit your own picture, leave the Subject out and describe it in words.
• The aspect ratio you choose applies to prompt-only pictures; an edit keeps its source's shape. Ratios: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2 and 2:3.
• Output is JPEG: no transparency and no lossless master.
• The base model renders at 1K only.

Use something else when…

• You need several references at once — a face, a product and a set: Nano Banana 2.1.
• An edit must keep identity, layout and light exactly: GPT Image 2.5 Sunburst.
• The lettering must be dense or exact: GPT Image 2.5 Flare.