GPT Image 2.5 for Scenes You Can Describe
Give a story moment a place, a cast and a point of view. GPT Image 2.5 on AINanoBanana2 turns a written scene or a set of image references into one illustration. Use the prompt to connect the subjects instead of leaving their relationship to chance.
Write the scene like a short visual brief
The useful details explain who is present, what is happening and how the viewer sees it.
Name the relationships
Describe where each subject is and what they are doing. Add camera distance, time of day and the emotional tone as separate details.
Assign a role to each reference
Use ordered references for characters, setting or props. A request accepts up to 16 images; explain what each one should contribute.
Pick the shape of the story
A wide scene can show a setting; a portrait composition can center a character. Choose Auto or a fixed ratio before generating.
From a story idea to a single illustrated moment
On AINanoBanana2, GPT Image 2.5 offers Flare and Sunburst for text-led creation and reference-guided editing. A prompt is required in either case. It is useful to define one scene per request, then judge whether the subjects and setting support that moment.
Each generation produces one image. Flare is selected initially, with 1K and Auto as the starting settings. Both versions cost 3 credits at 1K, 5 at 2K and 7 at 4K; references do not add another charge.
Make the scene readable
Describe which action is central and which details belong in the background so the image has a clear focus.
Plan a sequence yourself
Write individual briefs for later moments and reuse suitable references. Inspect consistency across outputs rather than treating it as guaranteed.
Keep the story brief organized
Use references and settings to support a specific moment.
Character, place, action
Separate those three parts in the prompt to make crowded scenes easier to describe and evaluate.
A reference is visual context
Tell the model whether an image represents a character, costume, object or atmosphere instead of asking it to copy everything.
Explore versions deliberately
Flare and Sunburst share the same input limits and price. Compare them through separate requests without assuming a promised style or speed difference.
Return to your scene studies
My Creations stores the submitted tasks, helping you find earlier results when planning the next illustration.
Prompt studies for an illustrated story
Begin with a single event, then check the resulting staging.
The rooftop observatory
Prompt: "A young astronomer and an older mechanic on a small rooftop observatory at dusk; the astronomer looks through a brass telescope on the left, the mechanic holds a folded star map on the right; blue city roofs below, warm lamp between them, watercolor illustration, no text." Choose 3:2 and check that each action is assigned to the intended person.
The lost parcel
Upload a character image as reference 1 and a street image as reference 2. Prompt: "Use the coat and appearance of the character in image 1. Show them standing beside the bakery in image 2, discovering a red parcel on the doorstep, light rain, eye-level view, storybook ink and muted colors." Choose 4:3 and inspect the coat, parcel and location.
A character introduction
Describe a portrait that includes one meaningful prop and a restrained background. Use a portrait ratio and review whether the expression fits the scene.
A second scene with shared references
Reuse the relevant input images in a new request, but describe the new location or action explicitly. Compare clothing, proportions and props with the earlier result.
A repeatable way to describe a scene
Three steps keep the prompt, references and result connected.
Outline one moment
Write who, where and what happens. Add style and framing after the main action; upload references only where they help explain the brief.
Choose the view and output
Set Flare or Sunburst, a ratio and 1K, 2K or 4K. Confirm the credit cost and submit one scene.
Review the visual relationships
Check character positions, objects, gestures and the background. A revised scene needs a new request; find previous tasks in My Creations.
Questions about scenes and references
Useful boundaries for planning illustrations.
Does the model remember the previous scene?
Do not rely on an earlier request being carried forward automatically. Include the references and descriptions needed for each new scene.
Can I use several character references?
Yes, within the 16-image input limit. Identify each reference by order and explain the character or visual element it represents.
Can one request create a complete comic?
One request produces one image. This generator does not provide a multi-page comic layout or automatic sequence workflow; plan individual illustrations separately.
What is the cost of another attempt?
Each new request costs 3 credits at 1K, 5 at 2K or 7 at 4K. Flare and Sunburst cost the same.
Which output shapes can I choose?
Auto and ten ratios are available: 1:1, 16:9, 9:16, 4:3, 3:4, 3:2, 2:3, 5:4, 4:5 and 21:9. All work with 1K, 2K and 4K.
What happens to a failed scene request?
A confirmed generation failure receives a full refund. A temporary inability to check status can be retried; My Creations lets you revisit the submitted task. Submitted tasks cannot be cancelled.
Choose the moment your story needs next
Describe the scene, assign the references and generate one illustration. Let the result show what needs to become clearer in the next brief.