Loading...
Loading...
A video idea often needs a visible starting point: a room at dusk, a costume choice, or the placement of a subject in a wide shot. Use GPT Image 2.5 on Seeddance to turn those decisions into still-image concepts. Start from words or add up to 16 references, then choose Flare or Sunburst and 1K, 2K or 4K output.
Work through setting, light and framing before choosing your next video-generation step.
Describe the room, time of day, camera position and materials. A text-only request gives you a still image to assess against that brief.
Use one image for costume, another for the location and another for a lighting direction. State those roles in upload order and identify what should change.
Try 16:9 for a landscape shot, 9:16 for a vertical composition or 21:9 for a wider concept. Each available ratio can be selected at up to 4K.
GPT Image 2.5 generates and edits still images. On Seeddance, it is useful for exploring the visual questions around a video idea: how a character might dress, where a scene takes place, and what belongs inside the frame. It does not produce a moving shot on this page. Review the image, then decide separately whether it is suitable as reference material for another tool.
On Seeddance, GPT Image 2.5 defaults to Flare, 1K and Auto aspect ratio, with Sunburst also available. Each request produces one image and can include up to 16 references. A 1K, 2K or 4K image costs 3, 5 or 7 credits respectively. Both versions cost the same and references have no extra fee.
If framing needs work, keep the subject description and revise the camera instruction. Compare the actual results before advancing the scene.
Find the submitted task and its image in My Creations when you return to the project.
Keep the brief small enough to evaluate: subject, environment, light and framing.
A location that does not exist yet can begin as a text description. Add references later when you have a direction to communicate.
Run the same brief with Flare or Sunburst as separate requests. Both versions share the same output controls; assess their results for your particular scene.
Set a ratio before submitting so the draft addresses a wide establishing shot or a portrait composition.
Choose 2K or 4K when you want a larger output to examine. Higher resolution does not replace checking faces, props and scene details.
Choose 1K, 2K or 4K for the output you need, at 3, 5 or 7 credits per image.
Follow the request and find its result in My Creations. Confirmed generation failures refund the full generation charge.
These are starting briefs for individual still images. Change the details to fit your own production.
Prompt: "A small rural train station after rain, empty platform, amber light from one waiting-room window, blue evening sky, wet concrete in the foreground, eye-level wide shot, no signs or readable text." Use text-to-image, choose 16:9 and 1K, then inspect the depth and light before revising.
Upload a costume reference first and a workshop photo second. Prompt: "Use image 1 for the olive work jacket and image 2 for the timber workshop. Show an adult mechanic standing beside a workbench, sleeves rolled up, soft daylight from the left, three-quarter full-body view, no text." Choose 3:4 and check how each reference was interpreted.
Use a selected scene image as a reference and ask for morning window light instead of evening light. Name the setting and composition details you want retained, then review the new output.
Describe a medium shot of the same scene idea in a new request. Ask which props should remain visible rather than assuming a model will preserve every detail between drafts.
Start with the visual question, then choose output settings.
Name the subject, setting, lighting and camera distance. In reference mode, upload the relevant images and explain their order; a written prompt is still required.
Select Flare or Sunburst and a ratio that fits the shot. Choose 1K, 2K or 4K and review the displayed credits.
Submit one image request. Inspect the result here or in My Creations, and revise the prompt if the scene needs another pass.
What this page generates, what to check, and how to manage each draft.
No. GPT Image 2.5 produces one still image per request here. Any later video generation is a separate step with its own settings and charges.
This tool returns one image, not a sequence of separate shots. For distinct scene references, submit separate prompts and evaluate the results individually.
Reference images can guide the request, but details still need review. Describe the traits you want retained and compare each output; consistent appearance across requests is not guaranteed.
Choose the output size that fits the review. A 1K draft can help evaluate the broad direction; 2K or 4K provides a larger image when you need it. Each submission is charged at the selected resolution.
Seeddance charges by the selected resolution: 3 credits for 1K, 5 for 2K and 7 for 4K. Flare and Sunburst cost the same, with no additional charge for reference images.
Check the submitted task here or in My Creations. A confirmed generation failure refunds the full generation charge; a delay or unavailable status check does not mean it failed. Submitted tasks cannot be cancelled.
Describe one location, costume or camera setup. Generate a still reference, review its details and decide where the project goes next.