OpenAI GPT Image 2
gpt-image-2General image generation, text/img-to-image, 27 fixed sizes.
Run the model and the result shows up here.
Overview
General image generation, text/img-to-image, 27 fixed sizes.
Key capabilities
Strong visual understanding
Follows complex written instructions and applies edits that match intent and context.
Intelligent image editing
Add, remove or alter elements precisely — from subtle tweaks to full restyling.
Multi-image input
Up to 5 reference images to guide the edit or supply a style.
Context-aware refinement
Preserves lighting, tone and composition, changing only what needs to change.
Efficient and affordable
Professional-grade editing billed per image × resolution tier — built for fast iteration.
Parameters
| Parameter | Description |
|---|---|
| Modes | text→img · img→img |
| Size | 10 ratios × 1K/2K/4K (30 total) or auto |
| Duration | —（image） |
| Reference images | up to 5, optional |
| Billing | per image × resolution (1K/2K/4K) |
- Aspect ratios (10): 1:1 · 3:2 · 2:3 · 4:3 · 3:4 · 5:4 · 4:5 · 16:9 · 9:16 · 21:9.
- size must be one of the 30 fixed sizes below or auto; other values are rejected upstream.
- Image models use the async /v1/videos endpoint; results are images, no quality / n params.
- Pass reference_images for image-to-image (up to 5).
Supported sizes
Pricing
Prices are per generation and charged from your prepaid balance. Failed jobs are refunded automatically.
Use cases
E-commerce imageryTurn a product shot into hero images, posters and feature layouts.
Background replacementKeep the subject, swap the background for any scene or flat colour.
Local retouchingAdjust lighting and colour, add or remove elements in the frame.
Creative iterationTry many directions on one asset and iterate cheaply.