What will you create next?
AI Image Generation
Start with a written scene or an image you want to change. GPT Image 2 offers 1K, 2K and 4K output, with frame shapes for different uses. Review the result, continue editing, or pass a chosen image into the video workspace.
Choose the canvas, the reference and the next step
Choose 1K, 2K or 4K and an available portrait, landscape or square ratio. Match the canvas to the layout before generating.
An uploaded reference establishes what you are working from. Explain the desired change and name any details that should stay the same.
The result's Image to Video action sends the chosen frame into the Seedance workspace, where you can write a motion brief without uploading it again.
A clear subject and a deliberate canvas make a useful starting brief.
Describe the subject, composition and lighting. If you are revising an image, attach it and state what should change.
Select 1K, 2K or 4K and the aspect ratio needed for the final layout.
Inspect the image at full size before downloading it or using Image to Video. A selected frame can become the starting point of a separate video task.
GPT Image 2 is an image-generation and editing model. You can begin with a text brief or supply a reference image and describe the revision you need.
Choose 1K, 2K or 4K output and one of the available landscape, portrait or square frame shapes. The selected ratio determines the composition's canvas.
Use an image reference and explain the edit in words. Specify both the intended change and the parts of the picture that should remain recognizable.
Resolution, generation option and account pricing determine the credit total. The workspace shows the current estimate before submission, including any applicable paid-account rate.
The Image to Video action transfers a result to the Seedance workspace. There you can describe its movement and submit a separate video-generation task.
Prepare an image brief in the workspace and check the available credits or plans when you are ready to generate.