OpenAI

GPT Image

GPT Image 2 creates new images from prompts and edits existing images using natural-language instructions. In Omnipix it also supports a masked inpainting workflow, where the source, reference, and mask must be supplied together. Quality, aspect ratio, output format, and compression are explicit controls, so choose them for the intended use rather than treating them as post-processing details.

How to use this model

  1. 1

    Start with the target use and describe subject, composition, lighting, materials, style, and exact text in one coherent prompt.

  2. 2

    For an edit, add reference images and state which identities, objects, colors, or layout elements must remain unchanged.

  3. 3

    For inpainting, provide the source image, a matching mask, and the required reference input before submitting.

  4. 4

    Select aspect ratio, quality, and output format deliberately; compression applies only to compatible formats.

  5. 5

    Check small text, edges around edited areas, hands, and reference fidelity at full resolution before iterating.

2Available

GPT Image 2

GPT Image 2 follows detailed prompts for new image creation and can edit existing content from references. Omnipix exposes a dedicated masked inpainting workflow in addition to ordinary generation and editing. Output quality, aspect ratio, format, and compression can be selected explicitly, while transparent-background output is not available in this Omnipix configuration.

Try now

Properties

  • Supports text-to-image and reference-image editing.
  • Includes a masked inpainting workflow in Omnipix.
  • Exposes output quality, aspect ratio, format, and compression controls.

How to use this model

  1. 1

    Write a complete generation or edit brief with exact visible wording and preserve/change constraints.

  2. 2

    Add reference images for identity, product, layout, or style and label their roles.

  3. 3

    For inpainting, supply the source, matching mask, and required reference input together.

  4. 4

    Choose aspect ratio, quality, format, and compatible compression before submitting.

  5. 5

    Inspect typography, edited boundaries, identity, and small geometry at full resolution.

Examples

Iced coffee beside a handwritten note
A product-style scene with legible text generated by GPT Image.

Parameters

Prompt

prompt
Required

Aspect Ratio

aspect_ratio
Optional
Default: 1:1

  • 21:9

  • 16:9

  • 4:3

  • 3:2

  • 1:1

  • 2:3

  • 3:4

  • 9:16

Resolution

resolution
Optional
Default: 1K

  • 1K

  • 2K

  • 3K

Quality

quality
Optional
Default: medium

  • low

  • medium

  • high

Output Format

output_format
Optional
Default: png

  • png

  • jpeg

  • webp

Output Compression

output_compression
Optional
Range: 0 – 100

Background

background
Optional
Default: auto

  • auto

  • opaque

Moderation

moderation
Optional
Default: auto

  • auto

  • low

N

n
Optional
Default: 1Range: 1 – 10

Reference Images

reference_images
Optional

Image

image
Optional

Mask

mask
Optional

Size

size
Optional