GPT Image 2.5: Design the Frame Before the Video
A video brief is easier to discuss when there is a picture to point to. Use GPT Image 2.5 on Veo Omni to explore a location, settle a camera view, or draft a starting frame. Write a scene description or supply visual references, generate one image, and decide what should move in a later video.
Give your video idea a visible starting point
Work out the setting, light, and framing in a still before preparing a separate video request.
Explore the location
Describe architecture, weather, time of day, and the empty space a subject will need. Use a location photo as a reference when words alone leave too much open.
Choose the opening composition
Select 16:9 for a landscape scene or 9:16 for a vertical shot. Ask for a clear foreground, a readable background, and room for the intended action.
Review a single frame
Each request returns one image. Inspect the layout, save the result, and revise the prompt for another attempt before using the image elsewhere.
A still-image step in your Veo Omni workflow
GPT Image 2.5 creates images; the video tools in Veo Omni handle a separate generation step. This page is useful for preparing the visual direction of a scene without asking a video model to resolve every design decision at once.
The default is Flare, 1K, and Auto composition. Sunburst is also available. Both versions accept text or up to 16 reference images and produce one image per request. On Veo Omni, 1K costs 3 credits, 2K costs 5, and 4K costs 7; adding references does not increase that price.
Separate design from motion
Describe where objects belong in the image first. Put camera movement and timing into your later video prompt.
Carry the image forward yourself
Download a chosen result, open an image-to-video tool, and upload it there if that model accepts the image. Generation settings and charges are separate.
Make scene choices easier to review
A concrete frame helps you identify what to keep before the next production step.
References with specific jobs
Name the references by order: the first sets the location, the second supplies a prop, and the third suggests lighting. The prompt explains how they should contribute.
A draft resolution for exploration
Start a composition study at 1K for 3 credits. When requesting another image, select 2K or 4K if your next review needs more detail.
Landscape and vertical studies
All three resolution settings support Auto and ten aspect ratios. Choose the frame shape before generating rather than assuming a later crop will work.
A record of the request
Open My Creations to return to the generated frame or check a submitted task. A confirmed failed generation receives a full refund.
Two scene briefs to try
These examples describe still images. Use the resulting frame as a reference, then write a separate motion brief in a video tool.
Prompt: Rainy station opening frame
Select 16:9, 1K, and Flare. Prompt: "A quiet rural railway platform at blue hour after rain. A yellow bench in the lower left, wet paving reflecting two warm lamps, mountains in the distance. Eye-level wide view. Keep the center walkway empty. No train, people, or lettering." Review the empty walkway before considering how a subject could enter the shot.
Prompt: Product reveal setting
Upload a product photo as reference 1. Select 9:16 and 2K. Prompt: "Use the ceramic bottle in reference 1 as the main subject. Place it on a pale stone plinth in a sunlit room, with a soft leaf shadow on the back wall. Bottle in the lower middle; leave the upper third clear. Show the whole bottle and a small amount of floor. No additional objects." Check the bottle shape and framing against the reference.
Build a usable reference frame
Keep the brief focused on what one image can communicate.
Define the visible scene
Describe subject, setting, camera height, lighting, and composition. If uploading references, say which details come from each image.
Set the image output
Choose Flare or Sunburst, a frame ratio, and 1K, 2K, or 4K. Check the credit amount before generating; one request produces one image.
Review, download, and reuse
Inspect the image and download it when ready. For animation, open the appropriate Veo Omni video tool and submit the downloaded image there as a new request.
Before you turn a frame into a video
Does this page generate video?
No. GPT Image 2.5 returns a still image. Veo Omni has separate video tools; you can download an image and upload it to a compatible image-to-video workflow. That video task has its own settings and cost.
Can I ask for a complete storyboard in one request?
Each request produces one image. For individually downloadable storyboard frames, submit a separate request for each scene and state the recurring visual details in each prompt. Consistency should be checked in the results.
Which version and resolution should I start with?
Flare at 1K is the default. Sunburst is an alternative, not a promised faster or higher-quality tier. Both versions cost 3, 5, or 7 credits for 1K, 2K, or 4K respectively.
How do multiple reference images work?
You can supply up to 16 images with a required text prompt. Explain the role of every useful reference. The 16-image limit refers to inputs, while the output is one image.
Is a slow task automatically refunded?
No. A delay or temporary status error is not confirmation of failure. Recheck the task in My Creations. Confirmed failures refund the full generation charge, and submitted tasks cannot be cancelled.
Start with the frame you want to animate
Describe a location, upload a visual reference if needed, and make one concrete scene choice with GPT Image 2.5.