
AI Image Generator
Explore the site’s broader image-generation workflow.

Build an original character concept with a readable silhouette and tactile material details, then refine a character or object in your own reference photo. This GPT Image 2.6 AI Image Generator page explains the text-to-image and image-to-image workflows, how to write focused prompts, and what to review before iterating. The page’s model selector offers GPT Image 2.6 Flare and GPT Image 2.6 Sunburst for generating and editing still images. Check your account access and the live credit quote for your selected model and size or aspect choice before you generate.
Start with the subject and its defining visual traits: overall shape, colors, face, clothing, and material. For a handmade figure, name details such as felt fibers, bead eyes, seams, or soft stuffing. Prioritize the features that make the character recognizable; a short, specific description is more useful than a pile of competing style directions.
State the camera view, how much of the subject should appear, and where it stands. A full-length view against a simple craft-studio backdrop makes the silhouette easy to judge. Add the setting, lighting, and floor or wall colors as separate instructions. For a variation, try a small bookshop-inspired set with blank spines and a sage-green wall.
Use Text to Image to create a scene from an editable prompt; this mode has zero uploads. Use Image to Image to upload one reference image and describe the edit in an editable prompt. For reference-based revision, choose an actual photo of the character or object you want to preserve. This guide covers still-image generation and editing.
Before Generate, check account access and the live Required Credits quote for your current model and size or aspect choice. Review the result for a readable outline, clean edges, believable material detail, unwanted lettering, and preserved identity or design. If something needs work, ask for one specific change and iterate rather than assuming an exact match.
A character concept becomes easier to understand when its outline is distinct from the setting. Name the full figure’s pose, proportions, and most recognizable shapes, such as triangular ears or a rounded muzzle. Ask for the entire standing subject when the complete design matters, and leave space around its outline. Keep the background simple enough that it does not compete with the character. Then check whether the face, clothing, and body shape remain legible at a glance.

Material words help explain what the surface should look like. Felt can be described through fibers, seams, and a soft edge; bead eyes through their small dark shape; wood through a pale, visible grain. Distinguish the material from the lighting: specify a soft light that reveals texture without changing the intended colors. Review edges and surface detail after generation, since a prompt guides the image rather than guaranteeing exact construction. If the material feels unclear, revise that instruction directly.

The two native workflows have different inputs. Text to Image accepts an editable prompt and no uploads, making it suitable for a new character scene. Image to Image accepts one reference image plus an editable prompt, so you can request a narrow adjustment while asking to preserve the visible subject’s own identity, design, and colors. Do not use the fox example as a replacement instruction when editing a different subject. Describe the change you want to the supplied image, then inspect what remained intact.

Follow this guide with GPT Image 2.6 Flare or GPT Image 2.6 Sunburst; the native selector also lets you explore other models separately. Both guide models are enabled for generate and edit. The names identify the available choices here; this guide does not promise that one is better or faster. Current model and size or aspect choices determine the live Required Credits quote. Check account access and that quote before Generate. After starting, find status and ready images in History. These are still-image workflows, not GIF or video output.
It guides you through creating a still image from an editable prompt or revising one supplied reference image with a focused prompt. The examples center on readable character silhouettes and visible material detail.
Text to Image uses an editable prompt with zero uploads. Image to Image uses one reference image and an editable prompt. The workflow described here does not use multiple reference images.
Free eligibility is not supported for these two models, so do not assume free access or a free trial. Check your account’s access and the live Required Credits quote for the current model and size or aspect choice before Generate; no fixed price is stated here.
No exact wait time is promised. Check the generation status and look for ready images in History.
Both selectable models are enabled for generate and edit in this page’s selector. No quality or speed comparison is claimed here; choose between the available names and check the live quote for your current settings.
Request one change at a time, such as softer lighting or a quieter background. Review the silhouette, edges, material detail, unwanted lettering, and—when editing—a reference’s preserved visible identity and design. Iterate if needed; exact fidelity is not guaranteed.
southparkcharacter.com · Updated