OpenAI image models

GPT Image 2.6 Image Generator

Create images from text or edit your photos with GPT Image 2.6. Choose Flare for everyday work or Sunburst for detail-focused edits, then review the credit quote before confirming.

Image direction

0 / 5000

PNG

Prompt, files, and settings stay in this browser for 24 hours.

The exact credit amount is shown before generation.

Finished images will be delivered privately to your account.

Flare or Sunburst

Choose a starting point for your image

GPT Image 2.6 brings two OpenAI image models into one workspace. Flare is the everyday starting point for exploring a composition, a product scene or an illustration. Sunburst is the option to consider when fine detail and consistency through edits matter more. OpenAI describes those differences in its November 4, 2026 announcement; they are a useful selection guide, not a guarantee that one setting will win on every brief. Both variants share this page, so you can compare their settings without moving between separate tools.

Start with the job you need the image to do. A rough campaign concept needs a clear idea and an appropriate frame; a product visual needs readable details and careful review. Changing models does not replace those decisions. Keep your subject, instructions, output size and quality setting the same when comparing results. Movelume has not published a controlled GPT Image 2.6 comparison here, and this page does not present another model’s images as 2.6 examples. Choose a variant, describe your image or add references, and review the credit quote before generating.

Cost and availability

Check the credits before committing to an image

GPT Image 2.6 text-to-image starts at 17 credits and reference editing starts at 22 credits for one reference. Account plans below describe credit packages, not a fixed number of images. Review the exact credit quote for your chosen variant, quality, size and references before confirming. Changing those inputs can change the quote. Do not use GPT Image 2.5 prices as an estimate for 2.6.

OpenAI API token rates describe developer usage and are not Movelume’s retail prices. A short text request and an edit with several references are different jobs, even if the downloaded pictures have the same dimensions. Plan your work around a small number of useful revisions rather than assuming unlimited attempts. For an available model, check the existing task when a result is delayed instead of repeatedly submitting. A confirmed failed generation follows the account’s credit-release or refund status; an unresolved task should not be treated as a confirmed refund.

Text to image

Give the image a clear visual purpose

For a new image, describe the subject, its surroundings, the composition and the intended use. A product photograph might need a simple background and space for a headline. An editorial illustration might need a specific mood and a limited palette. Pick the frame before writing lots of small details: a portrait layout and a wide banner leave very different amounts of room around the subject. Text-to-image produces one PNG per generation, giving you a single result to inspect and refine.

Separate essential requirements from preferences. Name the object or person that must be present, explain where the focus belongs, and then add lighting or material details. If lettering matters, keep the requested text short and review every character in the result. For a presentation or advertisement, allow room for final typography in your design editor. An image can be visually convincing while containing misspellings, invented labels or awkward geometry. Review these details at the actual size your audience will see, not only in a small preview.

Reference editing

Explain what should change and what should stay

Reference editing accepts one to sixteen images. Use a clear primary reference and add others only when each contributes something specific, such as an alternate angle or an essential material detail. Explain the change in ordinary language, then state the features to preserve. Changing a background while retaining the product shape is easier to judge than a broad request to improve everything. Avoid contradictory references, and use only images you have permission to upload and transform.

Image editing is generative: details outside the requested change may also shift. Inspect faces, product marks, edges, reflections and the relationship between objects after each edit. For a sequence of revisions, keep an original copy and compare against it periodically. OpenAI positions Sunburst for stronger consistency across edits, but that does not promise exact identity or pixel-perfect preservation. If an exact logo, dimension or regulated label must remain untouched, finish and verify that part with a suitable design tool before publishing.

Output settings

One PNG, five explicit frames, two quality levels

The first Movelume interface is scoped to PNG images with an opaque background. The five frames are square 1024×1024, landscape 1024×768, portrait 768×1024, wide 1280×720 and tall 720×1280. Those correspond to 1:1, 4:3, 3:4, 16:9 and 9:16. The same explicit dimensions apply to reference edits; uploading a differently shaped picture does not automatically preserve its original canvas. Choose a frame that suits your final placement and allow for any later crop.

Medium and high are the quality choices for both Flare and Sunburst. Quality and pixel dimensions answer different questions: increasing quality does not change the selected frame into a larger resolution. Use medium when assessing composition and high when detail is worth the additional cost shown in the quote. The interface does not provide transparent-background output. If transparency is essential, choose a tool that explicitly offers it, and verify edges and fine structures before using the result on a new background.

Compare fairly

Judge the result against your own brief

Choose a few concrete acceptance checks before generating: the subject is present, the composition fits the placement, the text is correct, and the requested changes are visible. For an edit, add a preservation check against the reference. Comparing those properties gives you a more useful answer than asking which model is universally best. Keep the same reference set, instructions, dimensions and quality level across any comparison, and include the actual credit cost and waiting time in your decision.

OpenAI’s announcement describes faster generation and positions Flare for fast iteration. Those statements are not Movelume speed measurements or delivery guarantees. Your waiting time can vary with the request and current service conditions. This page makes no six-second promise and does not offer the low-quality setting associated with that official example. If you have a deadline, leave time to inspect, revise and export the image. A successful download is a delivery result; it is not a guarantee that the artwork is ready for your particular use.

Before uploading or publishing

Check rights, privacy and the final context

Use source material you own or have the necessary permission to use. That includes photographs, artwork, branding and identifiable people. Generated images can resemble existing works or contain recognizable marks, so review the output before using it commercially. This page does not promise exclusive copyright, trademark clearance or universal commercial safety. Consider where the image will appear and whether permissions, disclosures or further professional review are needed for that use.

Do not treat OpenAI’s general statements about its products as Movelume’s privacy policy. Review Movelume’s own privacy policy and terms before uploading sensitive or confidential material. They explain the service relationship that applies here. Keep unnecessary personal information out of references and text instructions. Movelume is a separate service, and offering an OpenAI model does not imply OpenAI endorsement of Movelume. If you are uncertain about an input’s rights or confidentiality, resolve that question before uploading it.

Plans for image and video

Pricing

Annual plans are billed upfront. The same plan credits refresh every 30 days and expire at the end of that window; unused subscription credits do not roll over.

Starter

$336/ year

~~$34.99 / month~~Equivalent to $28 / month · billed upfront

For occasional image and video experiments with a predictable monthly ceiling.

2,400

credits every 30 days

About 6 short videos per credit cycleSeedance 2.0 · 5s · 720p · 380 credits each

2,400 credits refresh every 30 days
2 generations at once
New results saved for 30 days
Subscription credits expire at each refresh
Private results and generation history

CoreMost popular

$672/ year

~~$69.99 / month~~Equivalent to $56 / month · billed upfront

For creators who iterate every week and need room for several directions.

6,000

credits every 30 days

About 15 short videos per credit cycleSeedance 2.0 · 5s · 720p · 380 credits each

6,000 credits refresh every 30 days
4 generations at once
New results saved for 90 days
Subscription credits expire at each refresh
Private results and generation history

Max

$1,680/ year

~~$174.99 / month~~Equivalent to $140 / month · billed upfront

For high-volume creation, longer result storage, and Seedance 2.0 4K.

18,000

credits every 30 days

About 47 short videos per credit cycleSeedance 2.0 · 5s · 720p · 380 credits each

18,000 credits refresh every 30 days
6 generations at once
New results saved for 365 days
Seedance 2.0 4K access · credits charged per generation
Private results and generation history

Subscriptions renew automatically at the selected monthly or yearly rate until canceled. Cancel future renewals in your account.

Video counts are estimates for the settings above, not a separate allowance. Other models and settings use different credits. Review the exact quote before generating.

Pay safely and securely with

Compare what is included

Compare what is included
Plan benefit	Free	Starter	Core	Max
Monthly billing	$0	$34.99	$69.99	$174.99
Annual billing · paid upfront	$0	$336	$672	$1,680
Annual equivalent per month	$0	$28	$56	$140
Credits every 30 days	10 once · eligible new users	2,400	6,000	18,000
Simultaneous generations	1	2	4	6
New result storage	30 days	30 days	90 days	365 days
Seedance 2.0 4K access	Not included	Not included	Not included	Included · credits apply
Private results and history	Included	Included	Included	Included

Subscription credits refresh every 30 days, expire at the end of that window, and do not roll over. Annual billing does not issue all yearly credits upfront.

Storage is set when a result is generated. Upgrading does not extend the storage period of older results. Download files you want to keep.

The Max requirement applies to Seedance 2.0 4K. Nano Banana 2 4K does not require Max; your selected model and settings determine the credit quote.

Active paid subscribers can buy 1,800, 4,000, or 7,500 extra credits. Packs last 365 days and do not upgrade plan benefits.

Fixed-result templates have their own cash or credit options on each template page. A subscription does not include every $9.90 finished video.

The signup grant is one-time and subject to eligibility. Local drafts are free; daily check-in rewards are separate. Generation requires enough valid credits for the exact quote.

Checking payment availability…

Prices are in USD. Subscriptions renew automatically; credit packs are one-time purchases.

Build a free draft Compare plans and credit packs

Questions before you start

GPT Image 2.6 questions, answered

01Can I generate with GPT Image 2.6 on Movelume now?

Yes. Sign in, choose Flare or Sunburst, describe a new image or add references, then review the credit quote before confirming. Exact credits depend on your settings; check the amount shown for this request rather than assuming another model’s price.

02How do I choose between Flare and Sunburst?

OpenAI positions Flare for everyday speed and iteration, and Sunburst for detail and greater consistency across edits at medium or high quality. Choose according to your brief, then review actual results. This is official product positioning, not a Movelume benchmark or a promise of an exact completion time.

03Can I create a new image and edit an existing photo?

Yes. Text-to-image and reference editing are available for both Flare and Sunburst. Editing accepts one to sixteen reference images; use clear inputs and describe both the change and the details to preserve. Exact preservation is not guaranteed—inspect the result before publishing.

04Which formats, sizes and quality settings are included?

The initial choices are one opaque PNG, medium or high quality, and five frames: 1024×1024, 1024×768, 768×1024, 1280×720 or 720×1280. Reference edits use the frame you select. These are the choices prepared for this Movelume page, rather than a claim to offer every capability of the underlying models.

05How many credits will an image cost, and what if it fails?

GPT Image 2.6 text-to-image starts at 17 credits and reference editing starts at 22 credits for one reference. Review the exact quote before confirming; quality, size and reference count can change the amount. For a confirmed failed generation, check the task and account for the release or refund outcome. A pending task is not evidence that a refund has completed.

06Can I use the output commercially, and are my uploads private?

Review Movelume’s terms and privacy policy and make sure you have rights to your inputs. Commercial suitability depends on the content and intended use; generated output is not automatically cleared of copyright, trademark or likeness concerns. OpenAI’s policies do not replace Movelume’s own policy, and no absolute training or confidentiality promise is made here.

Your next step

Generate with Flare or Sunburst

Choose Flare or Sunburst, describe your image, and review the credit quote. Keep your intended format, source rights and budget in mind.

Generate with GPT Image 2.6 Available image models GPT Image 2.5 Movelume home Privacy policy Terms of service