GPT Image 2.5 Flare vs Nano Banana 2.1: Ad composition or precise UI structure?
Our tests split by task, with many ties: choose GPT Image 2.5 Flare for closer advertising composition and narrow advantages in preserving portraits, relit scenes and product details during edits. Choose Nano Banana 2.1 for structured UI graphics without extra text and a slight edge in preserving the base scene and replacement positions when combining reference objects.
Side-by-side facts
| Feature | GPT Image 2.5 Flare | Nano Banana 2.1 |
|---|---|---|
| Developer | OpenAI | |
| Availability on Banana Pie | Available now | Available now |
| Credits from | 25 credits | 25 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 taskTie. Both satisfy the main product and lighting requirements. Right is slightly closer to the requested fill level, while left gives a more subdued, fragmented reflection; neither has a decisive overall advantage.
- Single transparent rectangular bottle, with no text, logo, or extra objects: Both are even. Each image shows one upright rectangular glass perfume bottle with an attached cap, and neither contains visible text, logos, plants, or separate props.
- Halfway amber fill: Right is slightly closer. Its amber liquid occupies roughly the lower half of the internal chamber, though it looks a little above halfway. Left's liquid rises more noticeably above the chamber's midpoint.
- Coherent glass refraction and wet-stone reflection: Both are convincing. Left shows displaced edges and a visible pump tube through the glass and liquid, with a broken amber reflection following the rough stone. Right shows plausible distortion through the thick glass base and one aligned reflection in the smoother puddle, although that reflection is stronger than the requested subtle reflection.
- Upper-left lighting, consistent highlights, and contact shadow: Both are convincing. Left has pronounced upper-left illumination, bright left-facing glass edges, and a shadow extending right from the base. Right has softer highlights on the cap's top and the bottle's upper-left edges, with dark contact beneath the base and an amber light patch to the right.
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
Depends on the taskTie. Both render the correct wording and colors clearly, but both break the four-line requirement by splitting the title into two lines.
- Exactly four centered text lines and no additional text: Both are even. Each poster has five centered lines because the title wraps onto two lines. Neither adds extra wording.
- Exact wording and order: Both are even. LEFT renders "MOONLIGHT", "MARKET", "月光市集", "18 OCT", and "RIVER HALL"; RIGHT renders "MOONLIGHT", "MARKET", "月光市集", "18 OCT", and "RIVER HALL". All wording is correctly spelled and ordered, but neither keeps "MOONLIGHT MARKET" on one line as requested.
- Text colors: Both satisfy this criterion. "月光市集" is bright red on LEFT and darker red on RIGHT; the English title, date, and venue are black in both.
- Legibility, spacing, and alignment: Both are comparably legible and centered, with clear character spacing and generous separation between text groups. LEFT uses heavier lettering, while RIGHT uses lighter lettering; neither has a meaningful alignment or readability advantage.
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.
Human anatomy and contact
Depends on the taskTie. Both depict convincing pottery shaping, but neither clearly delivers two fully visible hands with five distinct fingers each. LEFT’s slight advantage in clay contact does not create a clear overall winner.
- Exactly two fully visible hands are present, with no extra hands or people: The two are even. Each image shows two hands belonging to one potter, with no extra hands or people. Neither fully exposes both hands: LEFT hides fingertips behind the bowl’s front rim, while RIGHT overlaps fingers along the bowl’s outer wall.
- Each hand has exactly five distinct, naturally formed fingers: Neither clearly satisfies this requirement. LEFT shows several separated fingers entering the bowl, but the rim hides their ends and the outer hand’s fingers overlap. RIGHT clearly separates four fingers on the hand above the bowl, while its thumb is hidden; the other hand’s fingers overlap beside the bowl.
- The fingers, joints, and contact with the clay are anatomically plausible: Both show plausible bent joints and complementary support inside and outside the bowl. LEFT is slightly closer to the requested contact because RIGHT’s conspicuous outward-pointing thumb is held away from the clay. Overlapping fingers in both images make the finer anatomy less clear.
- The clay bowl, spinning wheel, and shaping action are clearly recognizable: The two are even. Both show a recognizable clay bowl centered on a wheel, circular motion streaks, and clay-covered hands supporting and shaping the walls.
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.
Counting and attribute binding
Depends on the taskTie. Both satisfy the requested counts, arrangement, colors, and handle direction. Right makes the wooden material slightly clearer, while Left provides a cleaner background; neither has a substantial overall advantage.
- Object counts and placement: Both are even. Each shows exactly three red cubes in a row on the left, two blue spheres on the right, and one yellow mug centered behind the foreground objects.
- Colors, shapes, and materials: Right is slightly stronger because the red cubes show more distinct wood grain. Both correctly pair red with cubes, blue with transparent glass spheres, and yellow with a glazed ceramic mug, all on a gray table.
- Mug handle direction: Both are even. Each mug has a clearly visible handle extending to the right.
- No additional objects or text: Both show the six requested objects with no text or additional tabletop items. Left has a cleaner gray backdrop, while Right includes small portions of the room beyond the table at the upper corners.
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: GPT Image 2.5 FlareLeft is closer overall because its headline placement and orange trail better match the requested composition. Both fail the text-only constraint by adding "AERO" to the shoe.
- Shoe and orange trail placement: Left is closer. Both show exactly one silver shoe in the lower-right. Left’s orange trail sweeps through the bottom-left foreground, while Right’s trail enters from the bottom-right before curving toward the left.
- Visible text: The two are even. Both correctly render "RUN LIGHT" and "42 KM", but both also print "AERO" on the shoe, violating the requirement for no other text.
- Empty top area and headline placement: Left does better. Both leave the top 15 percent uncluttered, but Left places the headline closer to directly below that reserved space. Right leaves a much larger empty area and pushes the headline farther down.
- Visibility and visual hierarchy: The two are even. Both keep the entire shoe, headline, and round badge visible, with a clear separation between the large headline, the product, and the smaller badge. Left emphasizes the shoe through its larger scale; Right gives it more surrounding space.
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
Better here: Nano Banana 2.1RIGHT wins because it preserves the requested structure and includes no extra text. LEFT matches the main layout but adds unrequested plan values.
- Title and three columns: Both sides satisfy this criterion equally. Each displays the exact title "CHOOSE YOUR PLAN" above exactly three columns labeled "STARTER", "PRO", and "TEAM".
- Three aligned rows per column: Both sides satisfy this criterion equally. Every column contains "PROJECTS", "STORAGE", and "SUPPORT" in that order, with corresponding rows aligned horizontally across the graphic.
- One blue SELECT button per column: Both sides satisfy this criterion equally. Each column has exactly one blue button labeled "SELECT", positioned below the three rows.
- Even alignment and no additional columns or text: RIGHT wins. Both graphics have evenly spaced, equal-width columns, but RIGHT includes only the requested text. LEFT adds "5", "10 GB", and "Email" in STARTER; "25", "100 GB", and "Priority" in PRO; and "Unlimited", "1 TB", and "24/7" in TEAM, violating the instruction against additional text.
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: Nano Banana 2.1RIGHT has a slight overall edge for preserving the base scene and replacement positions. Both correctly bind all requested objects and include exactly two lemons.
- Mug and napkin replacement: The two are close. Both show a dark navy mug with the gold rim from Reference 2, a right-facing handle, and a folded green-and-white striped napkin in the correct areas. LEFT better retains the mug’s broad body; RIGHT keeps the napkin’s footprint closer to the original towel’s position in Reference 1.
- Orange and notebook replacement: RIGHT is slightly closer. Both contain exactly one orange with a green leaf and a mustard-yellow notebook with the reference’s black elastic band, and neither retains any original source object. RIGHT places the orange and notebook closer to their base positions; LEFT shifts both upward.
- Two lemons to the right of the mug: The two are even. Both show exactly two adjacent lemons to the mug’s right, with a leaf on the left lemon and a bare right lemon, matching Reference 6’s recognizable arrangement. Neither adds extra fruit, duplicate assets, or other objects.
- Base scene preservation: RIGHT does better. Its tabletop knots, upper-right table edge, napkin footprint, and notebook placement align more closely with Reference 1. LEFT shifts several objects and recognizable grain landmarks upward and gives the table a warmer, more saturated appearance. RIGHT still changes the mug’s proportions, but preserves the overall base layout more closely.
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
Depends on the taskTie. Both replace the text correctly and preserve the sign’s style and surrounding storefront closely, with no decisive visible advantage.
- Exact sign text: Both sides are even. LEFT and RIGHT each render "NIGHT OWL", and the reference’s "MORNING BEAN" is completely gone.
- No additional or malformed characters: Both sides are even. Each sign contains clean, recognizable uppercase letters, with no extra characters or remnants of the old lettering.
- Original font character, spacing, and perspective: Both are close to the reference. Each retains cream sans-serif capitals, generous letter spacing, and lettering aligned with the receding sign. LEFT’s lettering appears slightly taller within the board, while RIGHT’s placement is slightly farther right; neither has a decisive advantage.
- Unchanged material, lighting, and surrounding scene: Both closely preserve the green wood grain, dark frame, lamp shadows, storefront, plants, and bench. Small texture differences are visible, but the scene composition and warm lighting remain very similar to the reference on both sides.
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 FlareLEFT wins narrowly because it preserves the original portrait more faithfully. Both transfer the jacket's main features successfully; RIGHT has a slight advantage in natural garment integration.
- Identity, expression, hair, pose, proportions, and framing: LEFT does better. Its face, hair silhouette, head position, and portrait crop closely follow Reference 1. RIGHT retains the recognizable person and relaxed expression but changes the face shape slightly and uses a narrower composition with different head and shoulder proportions. Both shift the hands somewhat.
- Jacket material, color, collar, buttons, pockets, and sleeve patch: The two are close. Both reproduce the light blue denim, cream fleece collar, metal buttons, pointed chest pockets, vertical lower pockets, and red sleeve patch. LEFT shows the patch and pocket construction more clearly, but its denim has a more mottled surface than Reference 2. RIGHT's denim texture is closer, while its patch is reduced to a narrow visible strip.
- Full replacement without mannequin or product background: The two are even. Both fully replace the black blazer with the denim jacket and retain the gray wall and bright window edge from Reference 1. Neither introduces the product background or a mannequin.
- Physical coherence of hands, fit, occlusion, shadows, and lighting: RIGHT is slightly stronger. Its open jacket naturally reveals the original white shirt, with convincing folds and shadows around the elbows and cuffs. LEFT also has plausible hands and directional lighting, but removes the visible shirt neckline and brings the jacket fronts together while leaving opposing buttons and buttonholes visibly separated, making the closure less convincing.
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 FlareLEFT wins narrowly: both deliver convincing directional golden-hour lighting, but LEFT preserves the reference scene more faithfully.
- Warm golden-hour light clearly enters from the left window: Both succeed. LEFT shows a stronger amber glow through the window and across the sofa, while RIGHT shows warm rectangular sunlight patches on the wall and floor. Both clearly establish the left window as the source.
- Highlights and shadow directions respond coherently to the new light source: The two are close. Both illuminate the left-facing plant leaves and furniture surfaces, cast the plant’s shadow onto the wall to its right, and extend table shadows toward the lower right across the rug. RIGHT has somewhat cleaner separation between direct sunlight and shaded areas.
- The result is more than a uniform yellow color filter: Both succeed. Each introduces distinct window-shaped pools of sunlight, foliage shadows on the wall, and elongated furniture shadows. LEFT has a stronger overall orange cast, but its localized highlights and shadows clearly go beyond a color filter.
- No furniture or decor is moved, added, removed, or redesigned: LEFT does better. It closely preserves the reference’s framing, sofa contours, plant silhouette, table construction, and rug placement. RIGHT retains the same object types but changes the plant’s foliage, the table’s proportions and leg details, and the rug’s outline and foreground coverage.
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
Depends on the taskTie. RIGHT better preserves the original framing, while LEFT better retains the cabin’s weathered surface. Both produce convincing widescreen extensions, but neither leaves the original image unchanged.
- 16:9 with natural new content on both sides: Both sides satisfy this criterion. Each presents a wide landscape with beach, ocean, and sky extending naturally to both edges, without letterboxing.
- Complete original image centered without cropping or stretching: RIGHT is closer. Its cabin, horizon, and foreground retain the reference’s relative vertical positions and proportions. LEFT keeps the cabin centered but makes it smaller relative to the frame height and moves the shoreline upward, changing the original framing.
- Original cabin, shoreline, and internal composition unchanged: Neither fully satisfies this criterion. LEFT better retains the cabin’s conspicuous chipped paint and weathered boards, but alters its scale and the surrounding composition. RIGHT better preserves the cabin silhouette and scene layout, but visibly smooths and changes the cabin’s paint texture and redraws details in the sand and surf.
- Extensions without visible seams, mirrored filler, or repeated objects: The two are even. Both have continuous horizons, plausible cloud and sand transitions, and no obvious joining seams or duplicated cabins. Neither shows conspicuous mirrored filler.
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.
Product preservation across scene change
Better here: GPT Image 2.5 FlareLEFT has a slight overall edge because it preserves the reference sneaker’s proportions and side mark more closely. Both successfully place it on a wet basketball court with coherent evening lighting.
- Wet outdoor basketball court at dusk: Both sides satisfy this criterion. LEFT shows a hoop, fencing, court lines, and wet pavement beneath an orange evening sky. RIGHT shows more of the painted court and several hoops beneath a cooler twilight sky.
- Silhouette, sole geometry, stitching, materials, and camera angle: LEFT is closer to Reference 1, retaining the elongated profile, collar opening, panel seams, and proportions of the textured toe bumper. RIGHT makes the shoe appear somewhat more compact and changes the collar and heel-panel contours. LEFT adds conspicuous water droplets, but the underlying construction remains close to the reference.
- Black geometric side mark: LEFT is slightly closer to the reference in the angles and proportions of the three branches. RIGHT retains the same recognizable mark, though its descending branch appears shorter and its central junction slightly altered. Both keep the mark crisp and readable.
- Contact shadow, wet-surface reflection, and dusk lighting: Both are convincing. LEFT has a dark contact shadow, a broken reflection on textured pavement, and warm edge highlights matching the sunset. RIGHT has a subdued puddle reflection, shadow beneath the sole, and cooler illumination consistent with its twilight setting.
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
Both models start at 25 credits, so the starting credit budget is the same whichever you choose. In the small latency sample from this comparison, GPT Image 2.5 Flare took 70.5–101.0 seconds and Nano Banana 2.1 took 18.3–126.4 seconds. Nano Banana 2.1 generally returned images sooner, but also had the longest observed wait; these are observations from this run.
How we compared & disclosure
This comparison is a reproducible fixed-prompt run: we used one fixed prompt per scenario, with one run per model per scenario, and published the images as generated. The test results on this page compare how closely each model follows the requested composition, text and editing requirements. 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.