From visual brief to first frame

Create and edit with GPT Image 2.6

Plan the image before you plan the motion. Use GPT Image 2.6 in WAN3 to explore a scene, revise a reference or prepare the opening frame of your next video.

Create with GPT Image 2.6

Open Workspace with this model or workflow selected. Adjust settings, add your media, and create in one place.

Start Creating

What is GPT Image 2.6 on WAN3?

GPT Image 2.6 is an OpenAI image model. WAN3 offers Flare and Sunburst for text-to-image creation and reference-guided editing. Choose the version in the model menu, then describe a new scene or upload your images and explain what should change.

Three briefs to make your own

These are starting prompts, not examples of generated results. Replace the subject and details with your own before creating.

Product scene: A matte red travel kettle on slate beside a rain-streaked window. Cool daylight, close framing, empty space on the right. No text.
Campaign concept: A lime-green bicycle beneath a concrete stairway at noon. Graphic shadows, wide composition, room above for a headline added later.
Opening frame: A tiny observatory on a snowy ridge at blue hour. One amber window, distant mountains, a clear foreground path. Static wide shot.

Give each reference a job

For an edit, decide which image defines the subject and which supplies the setting or visual style. State what should stay and what should change; reference guidance is not a guarantee of exact reproduction.

Upload 1–16 JPEG, PNG or WebP references, up to 30 MiB each.
Name the details to retain, such as the silhouette, materials or palette.
Review labels, faces and small details before using the result in a campaign.
You can add one PNG mask or markup image to guide edits to the first reference. A mask marks areas to change or keep; markup adds visual instructions.

Choose the output for its destination

Choose low, medium, high, extra-high or maximum quality, and low, medium or high complexity. Both versions offer 1K, 2K and 4K output with square, portrait, landscape and panoramic ratios. Pixel dimensions depend on the ratio.

Use PNG, JPEG or WebP for still images. Text-to-image also offers animated GIF and vector SVG; each task returns one file.
Check the credit quote before submitting. Quality, resolution, edit references, mask or markup images and SVG output affect the total.
Test your composition at 1K before spending more on a larger version.

Build the opening frame, then direct the motion

Get the composition right in Image Studio, download the result and upload it in WAN3’s image-to-video workflow. Select a video model there and describe the movement and camera action. Video generation is a separate task with its own credit quote.

Before your first generation

Open the image studio with GPT Image 2.6 selected, choose text-to-image or image editing, add your brief and review the quote before submitting.

Which versions can I use? Both GPT Image 2.6 Flare and Sunburst are available in the model menu.
Can I edit a photo? Yes. Add your reference images and describe the changes in words.
Is generation free? Generations use WAN3 credits. The studio shows the cost for your settings.
How do GIF and SVG work? Choose text-to-image, then GIF or SVG. GIF accepts 2–16 frames and 20–2000 ms per frame. Image editing outputs PNG, JPEG or WebP.