Scene is a feature that automatically analyzes the content of your uploaded image and uses it to generate a new one, without copying the style. It works like Image to Prompt, but the generated prompt stays hidden from the user.
The system analyzes your image and builds a detailed internal description of the scene — including objects and composition. You can generate an image based solely on that description, or combine it with your own text prompt for more control.
The visual style of the reference image is ignored. Only the scene — what’s happening in the image — is used.
You don’t need to write a prompt. But if you do, the system will blend your prompt with the extracted scene. When there’s a conflict, your prompt takes priority.
How to use it
Click Reference image in the prompt panel
Set Mode = Reference
Choose Method = Scene
(Optional) Add a prompt to guide the output
Choose a style model
Click Create Image
How it works
Scene generates a hidden prompt based on what’s visible in your image: objects, actions, layout, and mood.
The generated prompt is not shown to you.
You can generate an image based only on this extracted scene, or enhance it with your own prompt
If there’s a conflict between your prompt and the scene, your prompt takes priority
The visual style of the original image is not transferred
Model compatibility
Scene works with all models — no limitations
Updated
14 August 2025
