Create a new image from a written brief, or use existing visuals to guide a new version. Choose the output size and number of results in the workspace, review the calculated credit total, and generate only when the setup is ready.
Two starting points
A written brief or visual references
Make the fit decision first
GPT Image 2 is an OpenAI image generation and editing model. On aiimg.me, it is available in text-to-image and image-to-image workflows. This page describes the controls currently exposed by the aiimg.me workspace rather than every option in the official API.
The model is most relevant when you are still making visual decisions: what the scene should contain, how it should be framed, or how an existing source should be reinterpreted. If the job is mainly about correcting a finished file without changing it, a conventional editor is the clearer choice.
Choose by starting material
The useful distinction is simple: decide whether the new image should begin from an open brief or from visual material that already matters.
Text to image
Start with the subject and intended use. Then describe the setting, framing, light, material, color, and mood. Concrete visual decisions are more useful than general quality words: “centered translucent blue lamp, dark studio, soft rim light, generous space above the subject for a headline” gives the model a clearer scene to interpret.
Image to image
Upload at least one image when an object, character, composition, pose, palette, or camera angle should inform the result. Write the anchor before the change: “Keep the centered bottle silhouette and front camera angle. Change the background to pale travertine and add warm morning side light.” The source provides context; it is not a locked template.
Settings and credits
The workspace calculates the total from the selected resolution and number of requested images, then displays it before you press Generate. The estimate updates with those controls, so you can compare a smaller test with a larger batch before committing. Opening either workspace from this page does not submit a task or consume credits.
| Images | 1K | 2K | 4K |
|---|---|---|---|
| 1 | 10 credits | 20 credits | 40 credits |
| 2 | 20 credits | 40 credits | 80 credits |
| 4 | 40 credits | 80 credits | 160 credits |
At 10 credits per image, 1K is the lower-cost way to check the subject, composition, palette, and prompt before committing more credits.
At 20 credits per image, 2K is the middle option when the direction is established but the highest available output setting is not yet necessary.
At 40 credits per image, 4K costs four times as much as 1K. Use it when output size matters, not as a way to repair an unclear prompt, unwanted detail, or weak composition.
Request one image while testing a prompt or setting. Choose two or four when comparing directions is worth the additional cost. Credits are deducted when you press Generate and the server creates each requested task.
Available choices are Auto, 1:1, 5:4, 9:16, 21:9, 16:9, 4:3, 3:2, 4:5, 3:4, 2:3. Use Auto when the layout is open, 1:1 for square placements, 4:5 or 3:4 for portrait layouts, 9:16 for vertical screens, and 16:9 or 21:9 for wide compositions. Choosing the destination first is usually more useful than forcing a finished image into a different crop.
Reference-image guidance
The image-to-image workspace accepts up to 16 references. Each file can be up to 30 MB, and supported formats are JPG, JPEG, PNG, WEBP. The limit is useful when different images provide different parts of the direction—for example, one for the subject, one for the composition, and another for material or palette.
Start with the smallest set that communicates the task clearly. If references disagree about shape, camera angle, or style, explain which one should guide each part of the result. More inputs add context; they do not remove the need for a clear instruction.
Upload only material you are allowed to use, and review our privacy policy before adding confidential or sensitive images.
Know the boundary
GPT Image 2 creates a new result from the prompt and any supplied visual context. Use the output as a generated interpretation that still needs review. A convincing image can still contain a detail that is wrong for the brief, so inspect the full-size result rather than approving it from a small preview.
For exact typography, identity-sensitive edits, measured product geometry, or compliance-critical assets, finish the selected result in an appropriate professional tool.
Before you open the workspace
No. The links on this page only open the selected workflow with GPT Image 2 preselected. A generation task is submitted after you review the settings and press Generate.
No. This page documents the current aiimg.me configuration: text to image, image to image, the listed ratios, 1K, 2K, and 4K output, one, two, or four results per run, and the current reference-image limits. The official API may expose a different set of controls.
Use text to image when the composition is still open. Use image to image when an existing subject, layout, or visual direction should guide the new result. Both paths open with GPT Image 2 selected.
