---
title: "GPT Image"
description: "Exact text, careful edits and layouts with OpenAI's image models."
section: "Models"
url: "https://batchbanana.com/docs/models/gpt-image"
markdown_url: "https://batchbanana.com/docs/models/gpt-image.md"
last_verified: "2026-10-03"
---

# GPT Image

Exact text, careful edits and layouts with OpenAI's image models.

![A six-panel brand board for Halvard chocolate: the cream HALVARD wordmark on cobalt, four contour-line colour swatches with their hex codes, an Aa type specimen named Halvard Serif, the contour pattern alone, a cobalt bar on an old map of Norway, and four bars in the four flavour colours in hard sun](https://batchbanana.com/docs/models/gpt-image/g-f.webp)

*Prompt: "The brand identity board for Halvard, a small chocolate maker in Bergen. The identity hangs on one idea from the bar in Image 1: a wrapper printed with fine white contour lines like a sea chart, cut short so the chocolate shows. Six panels with thin cream gutters, laid out like Image 2: the wordmark "HALVARD" in a tall cream serif on cobalt; four flavour swatches labelled "#1F3A93", "#F28C28", "#2F5D50", "#C8893B"; a type specimen "Aa" with "Halvard Serif"; the contour-line pattern alone; the bar from Image 1 on the atlas; four bars in the four flavour colours fanned out in hard sun. Take only the board layout from Image 2. Each line of text appears exactly once." Model: GPT Image 2.5 Sunburst.*

## Best for

GPT Image is OpenAI's image model, and it takes your words literally. Give it the exact headline, the real numbers for a chart or the screen you want mocked up, and it puts them where you said, spelled as you wrote them. It is also careful in edits: change one thing and the rest stays put.

Sunburst is one of the slower models here. For painterly or cinematic images, try [Nano Banana](https://batchbanana.com/docs/models/nano-banana) or [Seedream](https://batchbanana.com/docs/models/seedream). For vectors, use [Arrow](https://batchbanana.com/docs/models/arrow). [Which model for what](https://batchbanana.com/docs/models/which-model) shows all five families on the same prompts.

**App screen**

![A pixel-art phone level-select screen for Lantern Keep: a ruined castle at dusk, eight numbered stone nodes on a torch-lit path with three gold stars on levels 1 to 5, level 6 glowing, coins 1,240 and five hearts at the top, a green Play button at the bottom](https://batchbanana.com/docs/models/gpt-image/g-s1.webp)

*Prompt: "An iPhone level-select screen for Lantern Keep, a pixel-art dungeon game: a top-down map of a ruined castle at dusk with eight numbered stone level nodes on a torch-lit path, levels 1 to 5 with three gold stars, level 6 glowing. Top bar: coins "1,240" and hearts "5". Bottom: a green "Play" button." Model: GPT Image 2.5 Flare.*

**Character turnaround**

![A character turnaround of Fern the brown hare in an olive coat and a leather satchel, in front, three-quarter, side and back views at the same height, in ink and watercolour on off-white](https://batchbanana.com/docs/models/gpt-image/g-s2.webp)

*Prompt: "A character turnaround sheet of Fern, the hare in Image 1: front, three-quarter, side and back views in one row, same height, on plain off-white. Same ink and watercolour style, coat, satchel and colours as Image 1." Model: GPT Image 2.5 Sunburst.*

**Slide with real numbers**

![The same Halvard sales slide at High quality, with heavier black figures and a thin rule beside the chart](https://batchbanana.com/docs/models/gpt-image/g-q-high.webp)

*Prompt: "A pitch-deck slide titled "Halvard, first half of 2026". A bar chart of bars sold per month: Jan 1,200, Feb 1,350, Mar 1,800, Apr 2,100, May 2,600, Jun 3,100, each value above its bar, June in orange and the rest in cobalt. Beside it three numbers: "12,150 bars", "4.8 average rating", "41% buy again". Grotesk type on cream." Model: GPT Image 2.5 Flare.*

**Game sprite strip**

![Six pixel-art frames of a small armored knight in a red scarf running right with a glowing lantern, evenly spaced on near-black](https://batchbanana.com/docs/models/gpt-image/g-s4.webp)

*Prompt: "A sprite strip for a 2D side-scrolling game: a small armored knight carrying a lantern, six frames of a run cycle facing right in one row, evenly spaced, the same size, on flat near-black. Pixel art at 32 by 32 per frame scaled up with hard edges, a dark outline and a warm lantern glow." Model: GPT Image 2.5 Sunburst.*

## Which version

| Model | Best for | Credits per image | Reference images |
| --- | --- | ---: | --- |
| GPT Image 2.5 Sunburst | OpenAI's most capable image model, built for precise edits | 30 | Up to 8 |
| GPT Image 2.5 Flare | OpenAI's fastest model for everyday image generation | 30 | Up to 8 |

Both versions cost the same at every setting. Use **Flare** for new images; it takes about 16 seconds to Sunburst's 30. Use **Sunburst** for edits, where keeping the untouched parts exactly as they were matters most. On a fresh poster prompt, with the same reference poster attached for the look, both got every word right:

![A stack of vinyl records in magenta and blue flash light, with LATE NIGHT RECORD FAIR set vertically down the right edge and the date and address small in the left corners](https://batchbanana.com/docs/models/gpt-image/g-p1b.webp)

*GPT Image 2.5 Flare*

![The record fair poster made on Sunburst: a close stack of vinyl in pink and blue light, LATE NIGHT RECORD FAIR down the right edge, the date and address small in the left corners](https://batchbanana.com/docs/models/gpt-image/g-v-sunburst.webp)

*GPT Image 2.5 Sunburst*

*Prompt for all 2: "A poster in the look of Image 1 for a record fair: a stack of vinyl records shot close in hard on-camera flash against black, grainy, magenta and electric blue light in the grooves. Down the right edge, vertical: "LATE NIGHT RECORD FAIR". Top left, small: "Friday March 6, 8pm to 2am". Bottom left, small: "Dock 4, Rua do Cais 22". No other text."*

## Write the prompt

**Open with the subject and what the image is for.** OpenAI: "Name the subject and intended use, such as a product photograph, advertisement, or diagram." For anything busy, write the prompt in labelled parts. OpenAI suggests "scene, subject, details, and constraints, using labeled sections". The brand board at the top lists its six panels one by one, in reading order.

**Quote the text, say where it goes, and say it appears once.** "Put required wording in quotes and describe its position and typography." Spell unusual names letter by letter. Then say that nothing else may appear: "Each line appears exactly once. No other text." Both posters below had the same reference poster attached for the look. The one whose prompt only mentions "the dates and address" got a date in 2025, an address in California and a tagline, all made up, and a lily where the records should be:

**Mentions the text**

![A neon-lit lily on a poster for the Late Night Record Fair, with invented dates in May 2025, an invented address in Riverdale, CA, and a tagline nobody asked for](https://batchbanana.com/docs/models/gpt-image/g-p1a.webp)

*Prompt: "A poster in the look of Image 1 for the Late Night Record Fair, with the dates and address." Model: GPT Image 2.5 Flare.*

**Quotes the text**

![A stack of vinyl records in magenta and blue flash light, with LATE NIGHT RECORD FAIR set vertically down the right edge and the date and address small in the left corners](https://batchbanana.com/docs/models/gpt-image/g-p1b.webp)

*Prompt: "A poster in the look of Image 1 for a record fair: a stack of vinyl records shot close in hard on-camera flash against black, grainy, magenta and electric blue light in the grooves. Down the right edge, vertical: "LATE NIGHT RECORD FAIR". Top left, small: "Friday March 6, 8pm to 2am". Bottom left, small: "Dock 4, Rua do Cais 22". No other text." Model: GPT Image 2.5 Flare.*

**To change one thing, say "change only" and list what stays.** OpenAI: "say 'change only X' and list the details to preserve, such as identity, geometry, layout, lighting, or labels." Attach the image you're changing. On the Halvard board, a bare "Make it a night scene." took it literally: every panel went dark, and a moon, a lantern and a compass appeared. Naming the one panel to change, and saying the rest stays, redid the bottom-right photo and left the other five panels and every line of text alone:

**Original**

![A six-panel brand board for Halvard chocolate: the HALVARD wordmark on cobalt, four swatches with hex codes, a black Aa specimen named Halvard Serif, the contour pattern, a cobalt bar on an old map, and four coloured bars fanned in sunlight](https://batchbanana.com/docs/models/gpt-image/g-board.webp)

*Prompt: "The brand identity board for Halvard, a small chocolate maker in Bergen. The identity hangs on one idea from the bar in Image 1: a wrapper printed with fine white contour lines like a sea chart, cut short so the chocolate shows. Six panels with thin cream gutters, laid out like Image 2: the wordmark "HALVARD" in a tall cream serif on cobalt; four flavour swatches labelled "#1F3A93", "#F28C28", "#2F5D50", "#C8893B"; a type specimen "Aa" with "Halvard Serif"; the contour-line pattern alone; the bar from Image 1 on the atlas; four bars in the four flavour colours fanned out in hard sun. Take only the board layout from Image 2. Each line of text appears exactly once." Model: GPT Image 2.5 Sunburst.*

**Make it a night scene**

![The Halvard brand board turned to night in every panel: a moon over the sea behind the wordmark, dark swatches, a cream specimen on navy, and the bars lit by a lantern beside a compass](https://batchbanana.com/docs/models/gpt-image/g-p2a.webp)

*Prompt: "Make it a night scene." Model: GPT Image 2.5 Sunburst.*

**Change only one panel**

![The Halvard brand board with five panels unchanged and only the bottom-right panel redone: the four bars on a dark wooden table under one warm lamp](https://batchbanana.com/docs/models/gpt-image/g-p2b.webp)

*Prompt: "Change only the bottom-right panel: the four bars now lie on a dark wooden table under one warm lamp from the right. Keep the other five panels, the layout and every line of text exactly as they are." Model: GPT Image 2.5 Sunburst.*

When you edit the same image again, repeat the list. OpenAI's cookbook says to "repeat the preserve list on each iteration to reduce drift".

## More tips

**For slides and charts, give it the data.** OpenAI's advice is to write "like an artifact spec rather than an illustration request" and to put the "numbers and labels directly in the prompt". The slide above was made from a prompt that listed every month's value.

**Describe a screen as if the app already exists.** "Describe the product as if it already exists. Focus on layout, hierarchy, spacing, and real interface elements." Name the buttons, the counters and what they read.

**Give each reference a job.** "Identify each input by number and purpose: subject, style, clothing, or background."

**Keep a character fixed with a paragraph you repeat.** Describe the character's defining details the same way in every prompt, and add "Do not redesign the character." The turnaround of Fern above attached Fern's picture and named what to keep: the ink and watercolour style, the coat, the satchel and the colours.

**Write "photorealistic" when you want a photo.** OpenAI's cookbook: include the word "to strongly engage the model's photorealistic mode". Lens and camera words help, but OpenAI calls them "cues for appearance, not a guarantee".

## Settings

**Quality** is GPT Image's only setting besides the aspect ratio. It runs from **Low** to **Max**, and the default is **High**.

| Model | Credits per image | Other settings |
| --- | ---: | --- |
| GPT Image 2.5 Sunburst | 30 | Quality: Low 3, Medium 12, Extra high 59, Max 118 |
| GPT Image 2.5 Flare | 30 | Quality: Low 3, Medium 12, Extra high 59, Max 118 |

OpenAI's advice is to "compare quality levels before rewriting the prompt", and to use the top two levels "only when they improve an unmet quality requirement". Draft on **Low**, at 3 credits an image against 30 on **High**. In our test, Low got every number and label on this Halvard slide right; the higher levels added an axis and heavier type, not better facts:

![A pitch-deck slide titled Halvard, first half of 2026, with a bar chart of bars sold per month from 1,200 to 3,100, June in orange, and three figures: 12,150 bars, 4.8 average rating, 41% buy again](https://batchbanana.com/docs/models/gpt-image/g-q-low.webp)

*Quality Low*

![The same Halvard sales slide at Medium quality, now with a labelled vertical axis behind the bars](https://batchbanana.com/docs/models/gpt-image/g-q-medium.webp)

*Quality Medium*

![The same Halvard sales slide at High quality, with heavier black figures and a thin rule beside the chart](https://batchbanana.com/docs/models/gpt-image/g-q-high.webp)

*Quality High*

![The same Halvard sales slide at Extra high quality, with the three figures in blue and the month names in capitals](https://batchbanana.com/docs/models/gpt-image/g-q-xhigh.webp)

*Quality Extra high*

*Prompt for all 4: "A pitch-deck slide titled "Halvard, first half of 2026". A bar chart of bars sold per month: Jan 1,200, Feb 1,350, Mar 1,800, Apr 2,100, May 2,600, Jun 3,100, each value above its bar, June in orange and the rest in cobalt. Beside it three numbers: "12,150 bars", "4.8 average rating", "41% buy again". Grotesk type on cream." Model: GPT Image 2.5 Flare.*

GPT Image offers seven aspect ratios, from 16:9 to 9:16. It has no 21:9, 5:4 or 4:5.

## Limits

- It reads up to 8 reference images.
- It doesn't make transparent backgrounds. Generate on a plain background and remove it afterwards.
- Tiny text can still come out wrong. OpenAI suggests **Medium** or **High** for small text and dense panels.

## Sources

OpenAI's own guides: [Image prompting](https://developers.openai.com/api/docs/guides/image-prompting) and the [image prompting cookbook](https://developers.openai.com/cookbook/examples/multimodal/image-gen-models-prompting-guide).

Images made in Batch Banana on October 2, 2026. Select Prompt on an image to read the prompt and model that made it. Side-by-side pairs and comparison rows show the first result.

---

Last verified October 3, 2026. Web version: https://batchbanana.com/docs/models/gpt-image