imageby OpenAIAvailable in Cinemagiq
GPT Image 2.5
OpenAI's GPT Image 2.5: Flare for speed, Sunburst for precision
An image costs from about $0.03
What is GPT Image 2.5?
GPT Image 2.5 is OpenAI's image generation model and the step after GPT Image 2. Cinemagiq runs it in two lanes. GPT Image 2.5 Flare is the fast lane for everyday generation. GPT Image 2.5 Sunburst is the precision lane, built for careful edits where the details have to hold. Both sit on the Generation Studio's Image tab and return up to four images per run.
Unlike GPT Image 2, both lanes render an exact size for each of ten aspect ratios: 1:1, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 3:2, 2:3 and 21:9, for example 1536 x 864 at 16:9 and 1792 x 768 at 21:9. Quality runs from low and medium through high to xhigh and max, and the top two levels exist only on GPT Image 2.5. You can attach up to 4 reference images. Camera Angle Control is not available on OpenAI models.
In production, use GPT Image 2.5 Flare to iterate on concept frames, product shots and title cards, then switch to GPT Image 2.5 Sunburst when a frame needs tighter control, such as a multi-panel storyboard or comic page with captions. The 16:9 and 21:9 sizes suit widescreen keyframes that you can set as a First Frame on the Video tab. Results land in the Asset Hub under All Generations.
GPT Image 2.5 examples
Made in Cinemagiq with GPT Image 2.5. The exact prompt is under each one.
16:9
PromptProduct hero shot for a commercial: a faceted glass perfume bottle standing on black wet stone, a single beam of light cutting through mist, droplets on the glass, luxury advertising photography, no logo.
16:9
PromptTitle card for a documentary series: the words 'OCEANS AFTER DARK' in elegant serif letters over a dark underwater scene with glowing jellyfish, and the smaller subtitle 'Episode 1: The Glow' underneath.
16:9
PromptA comic page with four panels telling a tiny story: a girl finds a lost little robot dog built from copper pipes and a tin can, with mismatched button eyes, in the rain; she shares her umbrella with it; they walk home together; the robot dog sleeps by the fireplace. Clean ink lines, flat colors, short readable speech bubbles.
16:9
PromptPhotorealistic still: a teenage skateboarder in mid-air over a concrete ledge at golden hour, wearing plain unbranded clothes and plain dark shoes with no logos or stripes, city skyline behind, motion frozen, low angle, 35mm photograph.
What GPT Image 2.5 can do in Cinemagiq
Straight from the model list Cinemagiq runs today.
| Variant | Resolution | Reference images | Camera Angle Control |
|---|---|---|---|
| GPT Image 2.5 Flare | Model default | Up to 4 | No |
| GPT Image 2.5 Sunburst | Model default | Up to 4 | No |
Strengths
- Two lanes: Flare for speed, Sunburst for precision edits
- Ten exact aspect ratios, including 21:9 and 9:16
- Five quality levels, up to xhigh and max
- Up to 4 reference images
Best for
- Title cards and graphic frames
- Product and commercial hero shots
- Multi-panel storyboard and comic pages
- Widescreen keyframes for image-to-video
Watch-outs
- ·Output sizes are fixed per ratio (for example 1536 x 864 at 16:9), with no 2K or 4K setting.
- ·No panoramic ratios such as 1:4 or 8:1, and no Camera Angle Control.
- ·Reference tags like @image1 reach the model as plain text, so describe what each reference is for in words.
Prompt tips for GPT Image 2.5
- Start on GPT Image 2.5 Flare at medium quality to find the composition, then rerun the chosen prompt at high, xhigh or max.
- Write on-image text in quotes, exactly as it should appear, and say where it sits, for example a smaller subtitle underneath the title.
- For multi-panel pages, give the panel count and one short line of action per panel, in order.
- Describe each reference in words ('use the bottle shape from the reference photo'), because @image citations are not translated on OpenAI models.
How to use GPT Image 2.5 in Cinemagiq
- Open the Generation Studio in your project and switch to the Image tab.
- Pick GPT Image 2.5 in the AI Model picker. Each card says what drives the result and its limits.
- Write the prompt, set the aspect ratio and resolution, and attach reference images if you have them.
- Generate. The button shows the token cost before you start.
- Find every take in the Asset Hub under All Generations, and add the one you keep to its shot on the Production Board.
GPT Image 2.5 FAQ
What is the difference between GPT Image 2.5 Flare and Sunburst?
Both are OpenAI's GPT Image 2.5, with the same ten aspect ratios, five quality levels and up to 4 reference images. GPT Image 2.5 Flare is the fast lane for everyday generation. GPT Image 2.5 Sunburst is the precision lane for careful edits where details must hold.
Which aspect ratios does GPT Image 2.5 support?
Ten, each at an exact size: 1:1, 4:3, 3:4, 5:4, 4:5, 16:9, 9:16, 3:2, 2:3 and 21:9. Panoramic ratios such as 1:4 or 4:1 are outside OpenAI's limits, so for those use Nano Banana 2.1, which also offers 1:8 and 8:1.
How is GPT Image 2.5 different from GPT Image 2?
GPT Image 2.5 renders ten aspect ratios at exact sizes, where GPT Image 2 offers only 1:1, 3:2 and 2:3. It also adds the xhigh and max quality levels above high, and it comes in two lanes, GPT Image 2.5 Flare and GPT Image 2.5 Sunburst.
Does GPT Image 2.5 support Camera Angle Control?
No. Camera Angle Control works only on the Google Nano Banana models. With GPT Image 2.5, describe the camera in the prompt instead: shot size, camera height and lens, for example 'low angle, 35mm photograph, subject in the lower third of the frame'.
Related reading
Use GPT Image 2.5 inside Cinemagiq
Generate with GPT Image 2.5 and the other models Cinemagiq runs in one AI-native workspace, and pick the right one per shot.