Create an Image or Edit a Reference
Start from a text prompt, or upload an image when you want to change an existing visual.
GPT Image 2 AI Image Generator
Create images from a prompt or upload a reference to edit objects, text, backgrounds, layouts, and visual style with clearer control.
GPT Image 2
What GPT Image 2 Does Best
See what GPT Image 2 takes in, creates, and works best for.
- Text-to-image
- Create new images from natural-language descriptions of the subject, composition, lighting, style, and text.
- Reference editing
- Change an object, outfit, background, or label while telling the model which parts to preserve.
- Detailed instructions
- Specify composition, lighting, style, text, and small constraints in natural language for clearer control.
- Typography and layouts
- Make posters, ads, covers, comics, and layouts with clearer text placement and structure.
How to Use GPT Image 2
Create a new image from text or upload a reference to make a focused edit. Describe the result, generate, and refine it in a few steps.
Describe the Result Clearly
Name the subject, style, composition, text, and important details. For edits, separate what should change from what should stay untouched.
Generate and Refine
Review the result, adjust one detail at a time, and create another version until the image fits your use case.
What Can You Create with GPT Image 2?
Start with a prompt or reference image, then create new visuals, focused edits, character variations, and text-led designs.
Is GPT Image 2 the Right Model for You?
Use GPT Image 2 when you want to create an image from a detailed brief or make an intentional edit to a reference. Use Seedream 5 for broad visual exploration, then use Seedance 2 when an approved image needs motion.
Explore AI Creation Guides
Explore practical guides, creative techniques, and step-by-step tutorials to improve your AI images and videos.
Read More
Frequently asked questions about GPT Image 2
1. What is GPT Image 2 best for?
It is suited to high-quality image generation and editing from text prompts and image references, including visual concepts, key art, graphic assets, and targeted changes to an existing image.
2. Can I create an image without a reference?
Yes. Describe the subject, setting, style, composition, lighting, and any important visual details in a text prompt.
3. Can I edit an existing image?
Yes, when image reference input is available in the current Arcloop integration. Add the reference and say exactly what should change and what must stay the same.
4. How should I prompt for an image edit?
Start with the intended change, then list the elements to preserve. Include details such as placement, material, lighting, composition, or text requirements when they matter.
5. Can it create posters or images with text?
Use a clear layout description and reserve space for text. Always review generated copy, spelling, layout, and brand requirements before publishing.
6. Can I use generated results in production?
Review the result for visual accuracy, continuity, rights, and your project requirements before treating it as a final production asset.
Ready to create your next image?
Start with a detailed prompt or a reference image, then turn the strongest result into a reusable project asset.