Explore GPT Image 2.6 Flare and Sunburst on Felo, from fast image generation to precise editing, SVGs, GIFs, and limited-time trials.

GPT Image 2.6 on Felo: Try Flare and Sunburst

A good image often takes more than one try. You might need to explore a dozen campaign concepts, adjust a product shot without changing its label, or turn a character into a looping GIF. GPT Image 2.6 gives you two models for different parts of that process: Flare for fast iteration and Sunburst for detailed visual work.

Explore GPT Image 2.6 Flare on Felo and GPT Image 2.6 Sunburst on Felo. Both models are listed on the Felo API Platform with a limited-time free trial. You can explore their model pages and prepare to test them in the Playground. At the time of writing, both pages also show the provider as temporarily unavailable, so check the current status and trial terms before starting a session.

Meet the two GPT Image 2.6 models

OpenAI released GPT Image 2.6 Flare and Sunburst as image generation and editing models. Both accept text and image inputs, so you can start with a written brief or use existing images as references. They share capabilities such as transparent backgrounds, markup-guided editing, SVG output, and animated GIF generation in OpenAI's Image API. Availability of individual controls through Felo depends on its current integration.

	GPT Image 2.6 Flare	GPT Image 2.6 Sunburst
Best starting point	Fast drafts and frequent revisions	Detailed images and precise edits
Model emphasis	Lower latency and interactive workflows	Detail, layout, and editing accuracy
Useful for	Exploring concepts, social variations, rapid prototyping	Product imagery, polished designs, multi-round refinement
Model ID	gpt-image-2.6-flare	gpt-image-2.6-sunburst

The best choice depends on the work in front of you. Test both with the same prompt, reference images, size, and quality setting; then compare the details and response time that matter to your project.

GPT Image 2.6 Flare: move from idea to draft quickly

Flare is the smaller, speed-oriented model. It is a useful starting point when a person is waiting for an image on screen or when a team wants to compare several directions before choosing one.

OpenAI reports that Images 2.6 generates images 50% faster than before, and that Flare can generate an image in as little as six seconds at low quality. That is a best-case figure, not a promise for every request: image settings, prompt complexity, demand, and integration can affect the wait.

Felo's Flare page lists support for up to 16 reference images for editing, prompts up to 20,000 characters, and 1K, 2K, and 4K output with a long edge up to 3,840 pixels. It also lists six quality choices: auto, low, medium, high, xhigh, and max. Check the Playground for the options currently exposed.

Consider Flare when you want to:

• Explore several compositions for a campaign before choosing a direction.
• Generate quick visual concepts for a presentation or social post.
• Combine reference images into a first product mockup.
• Make successive edits while keeping a character or scene recognizable.

For example, a small coffee brand could ask for a set of poster concepts featuring its cold brew bottle in a sunlit café. Once it finds a layout it likes, it can reuse the result as a reference and ask for a different background or headline. Review the bottle shape and lettering after each revision; consistency is a goal, not a guarantee.

GPT Image 2.6 Sunburst: spend more attention on the details

Sunburst is the precision-oriented model. Start here when the image has a demanding layout, a product must remain recognizable, or a local edit needs to fit naturally into an existing scene.

Imagine a skincare team preparing a launch image. The product bottle, label, and lighting are already approved, but the background needs to change from a bathroom shelf to a pale stone surface. A useful prompt would identify the desired change and name what should remain fixed. Sunburst is designed for this kind of detail-sensitive work, though the final image still needs a careful review for label text, geometry, and unintended changes.

Sunburst is also a good candidate for:

• Product and campaign visuals that need detailed composition.
• Posters or layouts with several related elements.
• Character-led illustrations that develop across multiple images.
• Repeated rounds of targeted editing.
• Brand assets and simple vector concepts that can be refined in design software.

What you can create with GPT Image 2.6

Edit with visual markup

Sometimes it is easier to point at an object than describe its position. In OpenAI's Image API, markup-guided editing uses a separate transparent annotation layer to indicate where a change should happen. A prompt explains the requested edit and what to preserve. For example, circle a vase and ask for blue ceramic while keeping the rest of the room unchanged.

This is guidance, not a pixel-perfect lock. Inspect the entire result for changes outside the marked area. Felo lists image markup as a Sunburst capability, but its model pages do not yet publish model-specific protocol details; confirm the available controls in Felo before building a production workflow around them.

Create SVGs and animated GIFs

OpenAI's Image API supports SVG and GIF output for both 2.6 models. SVG can be useful for scalable icons, logos, and illustrations. Open the file in your design tool and check its paths and typography before using it in production.

GIF output opens up another kind of brief: a mascot waving, a candle flickering, or a product turning in a short loop. Describe the motion, camera, and starting and ending pose. For a seamless loop, tell the model to return the subject to its starting position. ### A GIF generated with GPT Image 2.6 Sunburst

A lantern-holding monkey lifts its light, blinks, and returns toward its starting pose. This example was generated with gpt-image-2.6-sunburst at 512 × 512 pixels, with nine frames at 200 milliseconds each. The result is a 1.8-second animation.

alt="GPT Image 2.6 Sunburst generated animation of a monkey holding a lantern" src="https://blog-static.felo.ai/sunburst_lantern_monkey_661cf9e755.gif" />

This example was generated outside Felo. Check Felo's current output options before assuming GIF generation is available through a particular interface.

More GIF ideas from the community

Creators have also shared GIF experiments they made with GPT Image 2.6, from playful characters to visual explanations. These examples were created outside Felo. The local GIF files are converted from the versions shared online; the model variant and generation settings were not disclosed for every example.

alt="A cat in sunglasses dances in a GPT Image 2.6 GIF" src="https://blog-static.felo.ai/dancing_cat_44c17891df.gif" />

A prompt turns a sunglasses-wearing cat into a dancing meme.

alt="A small animated robot mascot generated with GPT Image 2.6" src="https://blog-static.felo.ai/robot_mascot_dcd37b9d9b.gif" />

A robot mascot shows how a character can become a short animation.

alt="An animated owl character in a GPT Image 2.6 experiment" src="https://blog-static.felo.ai/animated_owl_ef87edc704.gif" />

An illustrated owl becomes a moving character.

alt="A pelican rides a bicycle in a GPT Image 2.6 GIF" src="https://blog-static.felo.ai/cycling_pelican_f97b322832.gif" />

A whimsical pelican-on-a-bicycle prompt, animated.

alt="An illustrated pelican rides a bicycle in a GPT Image 2.6 animation" src="https://blog-static.felo.ai/illustrated_pelican_test_932523df04.gif" />

An illustrated take on the pelican-on-a-bicycle animation test.

alt="A GPT Image 2.6 animation illustrates an AI model comparison dashboard" src="https://blog-static.felo.ai/dashboard_animation_653c2eec5c.gif" />

An animated concept explores a comparison dashboard for AI models.

alt="A fictional character from a novel brought to life as a GPT Image 2.6 GIF" src="https://blog-static.felo.ai/novel_character_8eff685601.gif" />

A character described in prose becomes a photorealistic animated portrait.

Adjust quality and complexity separately

OpenAI describes two distinct controls. Quality affects rendered detail; complexity controls how much planning the model does before rendering. Lower complexity can be useful for a focused change, while medium or high complexity may help with a crowded layout or a prompt containing many relationships.

The side-by-side example below uses Flare to animate a standing backflip at low, medium, and high complexity. Watch the movement and surroundings across all three versions; a single example shows possible differences rather than a reliable ranking of the settings.

alt="A standing backflip generated with GPT Image 2.6 Flare at low, medium and high complexity" src="https://blog-static.felo.ai/flare_complexity_backflip_e936e57b9f.gif" />

The same action tested at three complexity settings.

alt="An illustrated researcher animated at low, medium and high complexity with GPT Image 2.6 Flare" src="https://blog-static.felo.ai/flare_complexity_lab_2517245f3b.gif" />

A second Flare experiment compares an illustrated character at three complexity settings. The displayed times are individual measurements, not guaranteed response times.

More planning can also take longer or introduce unwanted changes. In OpenAI's API, complexity supports low, medium, and high, with high as the documented default. Felo's listed parameters include quality, background, aspect ratio, output format, and reference images; confirm whether complexity is exposed in your chosen Felo interface.

Create transparent assets

A transparent background is useful for product cutouts, stickers, and assets that need to sit on different page colors. Ask for an isolated subject and select a transparent background when the interface supports it. Check the actual alpha channel: a checkerboard painted into an image is not transparency.

How to try the models on Felo

1. Choose a model page. Start with Flare for quick exploration or Sunburst for a more demanding visual brief.
2. Check availability and the trial. Both pages currently advertise a limited-time free trial, but show the provider as temporarily unavailable. Trial eligibility, limits, and end date are not specified on the pages; verify the current terms before generating.
3. Open the Playground. Use the model page's Open Playground button and sign in if prompted. The pages also link to API key creation.
4. Write a specific prompt. Describe the subject, composition, style, and any text. For an edit, identify what to change and what to preserve.
5. Choose the available settings and review the result. Compare versions, inspect small details, and refine one change at a time.

For developers, the model IDs above identify the two listings. Felo's pages link to its model API documentation, but say model-specific protocol details are not yet published. Confirm the endpoint, request schema, supported parameters, and current availability there before integrating. The pages also describe adding balance as part of general console setup; a free-trial banner does not establish that unlimited API usage is free or that a top-up is unnecessary. Paid usage depends on the applicable provider and published terms.

Three prompts to get started

1. Explore a campaign concept with Flare

Create a square social campaign image for a fictional sparkling tea brand called Luma Leaf. Show a slim green can on a sunlit outdoor café table, with soft shadows and a fresh editorial photography style. Leave clear space at the top for a headline. Do not add any text or logos.

2. Refine an existing product image with Sunburst

Using the supplied image, replace the background with a pale limestone surface and soft morning light. Keep the bottle's shape, label, colors, camera angle, and position unchanged. Preserve the existing label text. Do not add props.

3. Design a looping GIF concept

Create a short, seamless looping GIF of a small blue paper bird perched on a branch. It gently opens its wings, blinks, and returns to its original pose. Keep the camera still, the background simple, and the bird's appearance consistent throughout the loop.

These prompts are starting points. Add reference images when you have approved colors, packaging, or character details, and check text, edges, and consistency before publishing a result.

Start creating with GPT Image 2.6

Flare helps you get to a first visual quickly; Sunburst gives you a starting point for work where small details matter. Compare them on your own prompts, then choose the one that fits the task.

Explore the limited-time trial: GPT Image 2.6 Flare on Felo · GPT Image 2.6 Sunburst on Felo