Head to head: Seedream 5.5 Edit vs Google: Nano Banana Pro (Gemini 3 Pro Image)
Seedream 5.5 Edit vs Google: Nano Banana Pro (Gemini 3 Pro Image)
By Ryan Merket · Published
This image-editing matchup tests text rendering, spatial relationships, constraint-following, and several tightly art-directed poster briefs. The comparison turns on different strengths across those tasks rather than a single consistent edge.
Seedream 5.5 Edit takes four task wins, including the Okapi Trail Relay palette, named art style, Harbor Sprint layout, and legible multi-line text. Its strongest pattern is focused art direction: the poster results more closely follow the requested composition, style, and palette. The text task is a narrow call, not a reason to overlook that the judges also praised Nano Banana Pro’s crisp, balanced layout there. Nano Banana Pro takes three wins: attribute binding, negation, and the Bauhaus goalkeeper poster. Its object arrangement and geometric poster treatment are notable strengths. The negation result deserves some caution in the editorial read: one assessment flagged visible foliage outside the window as a conflict with the no-plants instruction, so this is not evidence of uniformly clean constraint-following. Greaves of the Velodrome is saved as a tie. Seedream’s low, intimate framing puts the shoe and track details closer to the brief, while Nano Banana Pro makes the Vanta Circuit branding visible; the wider framing and imperfect lettering keep neither result decisively ahead. Across the sample, the task record is 4 wins for Seedream, 3 for Google, and 1 tie. The aggregate scores are 61.2 for Seedream and 62.8 for Nano Banana Pro, but the saved overall result remains a tie. **Final call: Too close to call (not statistical). This assessment applies only to the saved task sample; it does not establish a general capability advantage for either model.**
Legible multi-line text
A minimalist event poster with three lines of crisp, correctly-spelled text stacked and centered: 'NIGHT MARKET' large on top, 'Fridays · 6–11pm' in the middle, 'Riverside Pier 4' at the bottom, on a deep navy background, clean sans-serif, subtle grain.
OpenAI: GPT-6 Luna: Model B presents all three lines clearly and correctly, with a clean, restrained hierarchy and generous spacing that suits the minimalist poster brief. Model A also renders the text accurately, but its oversized headline feels crowded across the width and less balanced. (Second judge pass, order swapped — scores are the average of both: Model B presents all three lines crisply and correctly in a balanced, minimalist centered layout with subtle grain. Model A also renders the text accurately, but | Anthropic: Claude Fable 5.1: Both render all three lines correctly spelled a…
Greaves of the Velodrome
An ultra-detailed close view of a track cyclist from the fictional club Vanta Circuit tightening one shoe beside lane markings in an indoor velodrome, 16:9, with the camera low and slightly front-left: every surface must read sharply—carbon-fiber weave on the bike frame, brushed titanium pedal hardware, fine ribbing in the maroon socks, perforations and creases in the white shoe, individual arm hairs, tiny scuffs in the lacquered wood track, stitched bar tape, and dust caught in a narrow beam of overhead arena light—rendered in photorealistic sports editorial style with no smudged textures.
OpenAI: GPT-6 Luna: Model A more closely matches the requested low, intimate close-up, with the shoe being tightened on the pedal, maroon sock, track markings, and dramatic arena light all clearly emphasized. Model B includes visible Vanta Circuit branding, but its wider, higher framing makes the shoe and fine material details less central, and the lettering is imperfect. (Second judge pass, order swapped — scores are the average of both: Model A has the stronger low, close editorial composition, with crisp shoe an | Anthropic: Claude Fable 5.1: Model A delivers the requested low, ultra-close…
Attribute binding
A photorealistic tabletop scene: a matte green cube to the LEFT of a glossy red sphere, a small blue cylinder BEHIND both, and a single yellow rubber duck sitting ON TOP of the cube. Colors must stay on the correct objects. Soft studio light, 16:9.
OpenAI: GPT-6 Luna: Model A clearly shows the requested object colors and placements, with the duck on the cube and the blue cylinder behind both; its polished, softly lit tabletop composition is especially convincing. Model B also follows the arrangement well, but Model A has a more cohesive and refined photographic look. (Second judge pass, order swapped — scores are the average of both: Model B clearly satisfies the object colors and spatial relationships, with soft studio lighting and a clean tabletop presentat | Anthropic: Claude Fable 5.1: Both models bind every color to the correct obj…
Negation
A cozy reading nook with an armchair, a stack of books, and a mug of tea by a window — with absolutely NO plants, NO lamps, and NO artwork or picture frames anywhere in the frame. Warm afternoon light, 16:9.
OpenAI: GPT-6 Luna: Model A includes the requested reading nook elements and avoids lamps and artwork, with only blurred foliage visible outside the window; Model B is a polished, well-composed nook but shows a distinctly leafy garden through the window, more clearly violating the no-plants constraint. Model B has the stronger still-life composition, while Model A’s warm light and cozy reading moment are appealing. (Second judge pass, order swapped — scores are the average of both: Model B more directly centers the | Anthropic: Claude Fable 5.1: Both outputs respect the negations (no indoor p…
Okapi Trail Relay Palette
A mountain trail relay race threading through the Okapi Ridge switchbacks at sunset, shown as a stylized panoramic poster, 16:9, restricted entirely to the named palette ‘Japanese Indigo’: indigo, aizome blue, pale rice-paper cream, charcoal sumi black, and muted slate only, with no other hues anywhere in the runners, sky, rocks, banners, or shadows; feature three runners passing a baton on a ridge path under raking low light, with strong silhouette design and impeccable palette discipline.
OpenAI: GPT-6 Luna: Model A more fully matches the panoramic relay-poster brief, with a clear baton handoff, three runners, dramatic ridge scenery, and polished, legible typography. Model B has a strong silhouette-driven design, but its brighter cyan sky strays from the restricted palette and its Japanese banner text is less aligned with the requested poster. (Second judge pass, order swapped — scores are the average of both: Model A depicts a clear baton handoff among three runners in a polished panoramic mountain | Anthropic: Claude Fable 5.1: Model A delivers rich, disciplined indigo/cream…
Bauhaus Goalkeeper Dive
A goalkeeper in mid-air making a dramatic save during a fictional handball final in the Kestrel Dome, rendered faithfully in strict Bauhaus style, 16:9: geometric simplification, asymmetrical balance, bold circles and rectangles, flat planes, clean black outlines, minimal facial detail, dynamic diagonal composition, and typography integrated as abstract match signage; use hard directional lighting and a poster-like composition that unmistakably reads as Bauhaus rather than generic modernist art.
OpenAI: GPT-6 Luna: Model B more clearly delivers a Bauhaus poster, with a simplified diving goalkeeper, bold geometric goal and prominent Kestrel Dome match signage; some smaller lettering is garbled. Model A has a strong dramatic dive, but its detailed arena and more illustrative rendering feel less strictly Bauhaus. (Second judge pass, order swapped — scores are the average of both: Model B more clearly reads as a strict Bauhaus poster, with a geometric goalkeeper, unmistakable goal, and integrated match signage | Anthropic: Claude Fable 5.1: Model B reads unmistakably as Bauhaus with flat…
Named art style
A ukiyo-e woodblock print of a fishing boat riding a large cresting wave at dawn, faithful to the flat color planes, bold outlines, and stylized foam of the tradition, muted indigo and cream palette.
OpenAI: GPT-6 Luna: Model A more directly centers a fishing boat riding a towering cresting wave, with a clear dawn setting and a cohesive muted indigo-and-cream palette. Model B also evokes ukiyo-e convincingly, but its multiple boats and less focused wave-and-boat arrangement stray from the prompt’s singular scene. (Second judge pass, order swapped — scores are the average of both: Model A more clearly captures the ukiyo-e tradition with crisp flat color planes, bold contours, stylized foam, and a strong, readabl | Anthropic: Claude Fable 5.1: Model A delivers a single fishing boat clearly…
Harbor Sprint Layout
A tense sideline photo-illustration of a women’s kayak sprint training drill at the foggy Brinehook regatta basin at dawn, 16:9: a neon yellow kayak with black paddle is exactly in the center foreground, a coach in a crimson rain jacket stands on the dock to the left of the kayak, a stack of three orange buoys is to the right of the kayak, a silver bicycle lies on its side behind the coach, a white rescue dog wearing a blue harness sits between the coach and the buoy stack, and a green timing banner hangs directly above the kayak; render the scene in a crisp vintage travel poster style with long cool shadows and precise object placement.
OpenAI: GPT-6 Luna: Model A captures the active women’s sprint and nearly all requested placements in a crisp vintage travel-poster style, while Model B has a strong foggy dawn setting but shows an empty kayak and leans photorealistic rather than poster-like. Model A’s kayak is slightly off the exact center, but its overall arrangement is more faithful and visually cohesive. (Second judge pass, order swapped — scores are the average of both: Model A better captures the active women’s sprint drill and the requested | Anthropic: Claude Fable 5.1: Model A nails nearly every placement constraint…
Matchup powered by OpenRouter.