Detailed visual briefs
Describe composition, typography, lighting, materials, mood and the exact edits you need.
GPT Image 2 text and image workflows
Create a new image from a prompt or edit and combine up to 16 source images with GPT Image 2. Choose the canvas ratio and resolution, then preview the stored result online.
Output
Your image will appear here
Add a prompt and choose the output settings, then submit a GPT Image 2 task.
The model, explained
GPT Image 2 supports both text-to-image creation and image-to-image editing. Start from a written brief or edit and combine up to 16 source images for product concepts, editorial scenes, illustrations and other visual work.
Both Kie workflows expose canvas ratios from square and portrait formats to wide cinematic compositions. You can also choose 1K, 2K or 4K output when the selected ratio supports the requested resolution.
Requests run as asynchronous tasks instead of blocking the page. Once Kie completes the image, the server stores it privately so the result remains available through the authenticated output panel.
Built for precise image direction
Move from a written brief or source images to a stored, downloadable result with one generator.
Describe composition, typography, lighting, materials, mood and the exact edits you need.
Upload JPG, PNG or WebP references for image editing, combination and visual consistency.
Choose a supported ratio, then select 1K, 2K or 4K output for the workflow.
The asynchronous result is downloaded and stored before the authenticated preview link is shown.
Three simple steps
Choose text-to-image or image-to-image, write the brief, and let Kie complete the task.
Start from a prompt, or select image-to-image and upload between one and 16 source images.
Describe the subject or edits, then choose a supported canvas ratio and output resolution.
Wait for the task to finish, then inspect the stored image or download it.
Specific visual constraints help the model produce a more useful first draft.
Questions, answered
It supports both text-to-image generation and image-to-image editing, with selectable aspect ratios and 1K, 2K or 4K output.
Yes. Image-to-image mode requires at least one and accepts up to 16 JPG, PNG or WebP source images, each up to 30 MB. Text-to-image mode does not use references.
The provider queue and selected resolution affect timing. The output panel refreshes the asynchronous task while it runs.
Yes. After the server stores the completed Kie result, the output panel provides an authenticated preview and download link.
Start from a prompt or source images and generate a GPT Image 2 result online through Kie.
Create with GPT Image 2