GPT Image 2.5 Sunburst vs Nano Banana Pro for Image Tasks
The choice depends on the task. Pick Nano Banana Pro for multilingual typography, counting and attribute binding, advertising composition, and outpainting; pick GPT Image 2.5 Sunburst for faithful object edits, text replacement, garment transfer, and scene relighting. Product rendering and structured UI graphics ended in ties.
Side-by-side facts
| Feature | GPT Image 2.5 Sunburst | Nano Banana Pro |
|---|---|---|
| Developer | OpenAI | |
| Availability on Banana Pie | Available now | Available now |
| Credits from | 30 credits | 80 credits |
| Max resolution | 4K | 4K |
| Reference images | Up to 8 | Up to 8 |
Scenario by scenario
Same prompt. Two models. Judge the difference in the published images below.
Product material and lighting
Depends on the taskThe images split the strongest rubric details: the right has the subtler, cleaner reflection, while the left has more consistently modeled upper-left lighting and grounding. Both meet the bottle-count and fill requirements.
- Exactly one transparent rectangular perfume bottle: tie. Each image shows one upright rectangular glass bottle with no visible rendered text, logo, plants, or separate objects; the closures and spray mechanisms are parts of the bottles.
- Half filled with amber liquid: tie. Both bottles have a clearly visible amber liquid level near the midpoint of the glass body, with transparent empty space above.
- Glass refraction and wet-stone reflection: right does better. The right image has one restrained, continuous reflection beneath the bottle, while the left image's reflection spreads into several intense amber and white streaks and is less subtle. Both show visible refraction through the thick glass.
- Upper-left lighting and contact shadow: left does better. Its bright upper-left-facing edges, darker right side, and grounded shadow extending to the right are especially consistent; the right also has upper-left illumination and a contact shadow, but the broad diagonal background shadow competes with the bottle lighting.
Prompt & settings used
Prompt
A premium product photograph of exactly one transparent rectangular perfume bottle, half filled with amber liquid, standing upright on wet black stone. Light comes from the upper left, creating coherent refraction, a contact shadow, and one subtle reflection. No text, logo, plants, or extra objects.
Multilingual typography
Better here: Nano Banana ProRIGHT satisfies the poster specification more accurately, chiefly because it preserves "MOONLIGHT MARKET" as one line and therefore has exactly four lines.
- Exactly four centered text lines and no additional text: RIGHT does better. It shows four centered lines, while LEFT wraps "MOONLIGHT MARKET" into separate "MOONLIGHT" and "MARKET" lines, producing five lines total.
- Exact text and order: RIGHT does better. Its lines read exactly "MOONLIGHT MARKET", "月光市集", "18 OCT", and "RIVER HALL" in the requested order. LEFT renders "MOONLIGHT" and "MARKET" as two separate lines before "月光市集", "18 OCT", and "RIVER HALL".
- Text colors: Tie. Both render "月光市集" in red and all English and numeric text in black.
- Legibility, spacing, and alignment: RIGHT does better. All four lines are clearly legible, centered, and arranged as a compact, consistently aligned group. LEFT is also legible and centered, but the unintended title wrap disrupts the requested four-line spacing and structure.
Prompt & settings used
Prompt
Design a clean cream-colored vertical event poster. Show exactly four centered text lines and no other text: "MOONLIGHT MARKET", "月光市集", "18 OCT", and "RIVER HALL". Render "月光市集" in red and all other lines in black.
Counting and attribute binding
Better here: Nano Banana ProBoth outputs follow the requested counts and arrangement closely. The right image has the clearer material distinctions, especially the visible wooden grain and ceramic glaze, giving it the overall edge.
- Exactly three red cubes, two blue spheres, and one yellow mug: Tie. Both images visibly show three cubes in a row on the left, two spheres on the right, and a single mug centered behind them.
- Colors, shapes, and materials: Right does better. Both bind red to cubes, blue to spheres, and yellow to the mug, but the right image shows clearer wood grain on the cubes and a more visibly ceramic, glazed surface on the mug; its spheres also read as transparent glass.
- Mug handle direction: Tie. In both images, the mug handle is clearly attached on and extends toward the right.
- No additional objects or text: Tie. Neither image contains any visible extra objects or rendered text.
Prompt & settings used
Prompt
On a matte gray table, exactly three red wooden cubes form a row on the left, exactly two blue glass spheres sit on the right, and one yellow ceramic mug stands centered behind them. The mug handle points right. No other objects or text.
Advertising composition
Better here: Nano Banana ProRight follows the text restriction and top-spacing direction more closely while keeping every required element clear and fully visible. Left is visually strong but adds the forbidden text "AERO" and leaves more empty space than requested above the headline.
- Exactly one silver shoe and orange trail: Both show exactly one silver shoe in the lower-right and an orange trail entering from the bottom-left. The right places both elements clearly, while the left shoe is larger and more decisively anchored in the lower-right; slight edge to left.
- Only visible text: Right does better. It shows only "RUN LIGHT" and "42 KM", both spelled correctly. Left also renders "AERO" on the shoe, which violates the requirement for no other visible text.
- Top spacing and headline position: Right does better. Its upper area is visibly empty, with "RUN LIGHT" placed immediately below it in the upper-left. Left also leaves the top uncluttered, but the empty region is substantially deeper before the headline begins.
- Visibility and visual hierarchy: Both keep the required headline, badge, trail, and shoe fully visible. Right has the clearer prompt-compliant hierarchy, with "RUN LIGHT" leading into the prominent "42 KM" badge, trail, and shoe without extra text competing for attention.
Prompt & settings used
Prompt
Create a vertical social ad for a fictional running shoe named AERO. Keep the top 15 percent empty. Directly below it, place the headline "RUN LIGHT" in the upper-left. Show exactly one silver shoe in the lower-right, an orange trail curving from the bottom-left, and a round badge reading "42 KM". No other shoes or text.
Structured UI graphic
Depends on the taskBoth images satisfy the title, column, row-label, and button requirements, while both add plan-detail text forbidden by the prompt. Neither has a clear rubric-level advantage.
- Criterion 1: Tie. Both display the exact title "CHOOSE YOUR PLAN" and exactly three columns labeled "STARTER", "PRO", and "TEAM".
- Criterion 2: Tie. Each column in both images contains exactly three aligned labeled rows: "PROJECTS", "STORAGE", and "SUPPORT".
- Criterion 3: Tie. Each column in both images contains exactly one blue button labeled "SELECT".
- Criterion 4: Tie. Both use three evenly aligned columns, but both violate the no-additional-text requirement. LEFT adds "3", "5 GB", "Email", "10", "50 GB", "Priority", "Unlimited", "1 TB", and "24/7". RIGHT adds "5 active projects", "10 GB", "Email support", "20 active projects", "50 GB", "Priority email & chat support", "Unlimited projects", "200 GB", and "24/7 phone, email & chat support".
Prompt & settings used
Prompt
Create a clean horizontal pricing comparison graphic titled "CHOOSE YOUR PLAN". Use exactly three equal columns labeled "STARTER", "PRO", and "TEAM". Under each column, show exactly three aligned rows labeled "PROJECTS", "STORAGE", and "SUPPORT", followed by one blue button labeled "SELECT". White background, dark navy text, no additional columns or text.
Six-reference object binding and edit
Better here: GPT Image 2.5 SunburstLeft wins overall by preserving the base scene and the replacement objects more faithfully. Right has the better lemon pair, but its tighter framing and flatter, smaller object renderings depart more from Reference 1.
- Criterion 1: Left does better. Its dark navy-blue mug more closely preserves the reference mug's broad cylindrical shape, glossy finish, gold rim, and original lower-left placement, while its striped napkin retains the thicker folded form and stronger green-and-white appearance of Reference 3. Right includes both replacements correctly but renders the napkin flatter and the mug smaller.
- Criterion 2: Left does better. Both replace the pear and blue notebook without leaving any original source object, but Left's orange more closely matches Reference 4 in size, textured peel, stem, and leaf, while its mustard-yellow notebook more clearly retains the hardcover texture, black elastic, and visible page block of Reference 5.
- Criterion 3: Right does better. Both show exactly two lemons to the right of the mug with no extra fruit, but Right preserves the distinctive leaf on the left lemon from Reference 6; Left omits that leaf.
- Criterion 4: Left does better. It remains much closer to Reference 1's wide camera framing, object scale, tabletop coverage, right-edge notebook placement, visible wood grain, and overall lighting. Right changes the framing to a much tighter, nearly square composition and shifts the apparent scale and spacing of the scene.
Reference images
Prompt & settings used
Prompt
Use Reference 1 as the base tabletop scene. Replace the red mug with the exact dark navy-blue mug from Reference 2 in the same position and orientation. Replace the folded white towel with the exact green-and-white striped napkin from Reference 3 in the same folded area. Replace the green pear with exactly one orange from Reference 4 in the same position. Replace the blue notebook with the exact mustard-yellow hardcover notebook from Reference 5 in the same position. Add exactly the two lemons from Reference 6 to the right of the mug. Preserve the wooden table, camera, framing, wood grain, lighting, shadows, and all other spatial relationships from Reference 1. Do not copy the white product backgrounds from References 2–6, duplicate any asset, or add other objects.
In-image text replacement
Better here: GPT Image 2.5 SunburstLeft fulfills the replacement while preserving the original image much more faithfully. Right gets the text correct but alters much of the scene.
- Exact text replacement: Both sides read exactly "NIGHT OWL", with no visible remnants of "MORNING BEAN".
- No additional or malformed characters: Both sides render only "NIGHT OWL" with clean, correctly formed letters.
- Font, spacing, and perspective: Left is closer to the reference, retaining the original tall sans-serif character, wide letter spacing, and the sign's original perspective. Right uses similar lettering but changes the framing and perspective of the storefront.
- Material, lighting, and scene preservation: Left is substantially better; the sign texture, shadows, lamps, storefront, plants, bench, and framing remain closely aligned with the reference. Right visibly changes the crop, viewpoint, storefront proportions, lamp positions, windows, planter, bench, foliage, and surrounding details.
Reference images
Prompt & settings used
Prompt
Replace only the sign text with exactly "NIGHT OWL". Preserve the original font style, spacing, perspective, sign material, lighting, and everything else.
Multi-reference identity and garment transfer
Better here: GPT Image 2.5 SunburstLEFT more successfully performs the identity-preserving garment transfer. It retains the original portrait more closely and reproduces substantially more of the reference jacket's construction and details.
- Identity, expression, hair, pose, body proportions, and framing: LEFT does better. Its face, head size, hair silhouette, shoulder width, hand position, and crop closely match Reference 1, while RIGHT renders a narrower face, shifts the hands upward, and uses a somewhat wider body framing.
- Jacket material, color, collar, buttons, pockets, and sleeve patch: LEFT does better. It more closely reproduces the light-blue denim, thick cream shearling collar, centered bronze buttons, pointed flap pockets, lower welt pockets, vertical panel stitching, and red upper-sleeve patch from Reference 2. RIGHT omits the visible lower welt pockets and changes the front into an open, asymmetrical arrangement.
- Full replacement without mannequin or product background: Both are even. Each fully replaces the black suit jacket with denim while retaining the portrait's gray wall and left-side curtain; neither carries over the ghost mannequin shape or white product background.
- Hands, garment fit, occlusion, shadows, and lighting: LEFT does slightly better. The hands remain closer to Reference 1, the sleeves meet the wrists cleanly, and the jacket's folds and shading follow the original scene consistently. RIGHT is also physically plausible, but its altered hand placement and uneven open-front button and buttonhole arrangement are less faithful and slightly less coherent.
Reference images
Prompt & settings used
Prompt
Use the portrait in Reference 1 for the person and the isolated jacket in Reference 2 for the garment. Dress the person from Reference 1 in the exact jacket shown in Reference 2. Preserve the person's identity, face, expression, skin, hair, hands, pose, body proportions, background, framing, and lighting from Reference 1. Preserve the jacket's material, color, collar, buttons, pockets, and sleeve patch from Reference 2. Do not copy the ghost mannequin or white product background.
Coherent scene relighting
Better here: GPT Image 2.5 SunburstLEFT delivers the more faithful relighting: it creates convincing golden-hour illumination and coherent directional shadows while preserving the original room much more closely. RIGHT achieves warm directional light but noticeably reconstructs the scene.
- Warm golden-hour light clearly enters from the left window: LEFT does better. Visible low sun appears through the left window and casts strongly warm light deep into the room; RIGHT also shows warm left-side illumination, but the source is less directly apparent.
- Highlights and shadow directions respond coherently to the new light source: LEFT does better. Window-frame and foliage shadows extend across the wall, sofa, rug, and floor away from the left window, while the coffee table casts a consistent elongated shadow toward the right. RIGHT has directional shadows too, but some large shadow shapes across the sofa and wall look less naturally connected.
- The result is more than a uniform yellow color filter: LEFT does better. It adds distinct sunlit patches, window and foliage projections, directional cast shadows, and varied highlight intensity across surfaces. RIGHT also goes beyond a flat tint, though its overall amber cast is more uniform.
- No furniture or decor is moved, added, removed, or redesigned: LEFT does better. Its window, plant, sofa, lamp, table, and rug closely retain the reference arrangement and designs. RIGHT substantially changes the framing and relative placement and proportions of the sofa, plant, coffee table, rug, lamp, and window.
Reference images
Prompt & settings used
Prompt
Change the lighting to warm golden-hour sunlight entering from the left window. Do not move, add, remove, or redesign any object. Update highlights and shadows coherently.
Cross-ratio outpainting
Better here: Nano Banana ProThe right output is the stronger outpainting because it preserves the reference composition much more faithfully while producing clean, natural side extensions.
- 16:9 format and natural side content: Both outputs are landscape and extend the beach, ocean, and sky across both sides, but the right extension is more consistent with the reference's restrained cloud patterns and shoreline.
- Complete original centered without cropping or stretching: Right does better; the reference's square composition remains centered with matching vertical placement of the roof peak, horizon, shoreline, and foreground, while the left shifts the horizon upward and makes the cabin noticeably narrower relative to the preserved central area.
- Original cabin, shoreline, and internal composition unchanged: Right does better; its cabin size, roof geometry, horizon height, wave bands, and sand proportions closely match the reference. Left alters the cabin proportions and changes the relative amounts of sky, ocean, and beach.
- No visible seams, mirrored filler, or repeated objects: Right does better; its sky, wave bands, and beach texture continue smoothly without obvious repetition. Left has more conspicuous changes in cloud density and texture across the expanded scene, though no hard seam is visible.
Reference images
Prompt & settings used
Prompt
Expand the canvas to a 16:9 landscape by naturally continuing the beach, ocean, and sky on both sides. Keep the complete original image centered without cropping, stretching, letterboxing, or modifying it.
Human anatomy and contact
Prompt & settings used
Prompt
A photorealistic close-up of an adult potter shaping a clay bowl on a spinning wheel. Both hands are fully visible, each with five natural fingers touching the clay. No other people or hands. Soft window light.
Product preservation across scene change
Reference images
Prompt & settings used
Prompt
Place the sneaker on a wet outdoor basketball court at dusk. Preserve the exact sneaker shape, black geometric side mark, materials, stitching, sole geometry, and camera angle. Add physically coherent contact, reflections, and dusk lighting.
Cost & latency
GPT Image 2.5 Sunburst starts at 30 credits, while Nano Banana Pro starts at 80, making Sunburst the lower-credit option. In this small run, observed latency ranged from about 69 s to 155 s per scenario for Sunburst and from about 39 s to 68 s for Nano Banana Pro; these are run-specific samples, not benchmark figures.
How we compared & disclosure
We used one fixed prompt per scenario and made one run with each model per scenario, publishing the outputs as generated. Banana Pie sells paid access to both models.
Banana Pie sells paid access to this model alongside other models in one studio. Our verdicts come from tests run through the same pipeline our users get.