Head to head: Google: Nano Banana 2 (Gemini 3.1 Flash Image) vs Seedream 5.0 Pro Image Editing
Google: Nano Banana 2 (Gemini 3.1 Flash Image) vs Seedream 5.0 Pro Image Editing
By Ryan Merket · Published
Eight image-generation challenges put two models through cinematic scenes, precise object placement, material detail, and style imitation. The results show where each is dependable—and where following the brief matters more than surface polish.
Google’s Nano Banana 2 takes the matchup, but this is a **lean, not a blowout**: it leads 61.5 to 59.4, wins four tasks to Seedream’s two, and ties two. The limited confidence points to a real edge, while the narrow score gap argues against calling it a runaway. Google’s clearest advantage is **compositional discipline**. It kept all four color-locked cargo objects in frame, followed the isometric room layout, and preserved more of the aircraft-and-mechanic scene in “Rivets and weave.” It also delivered the more convincing cinematic convoy: a diagonal line of vehicles cresting a ridge through warm, dramatic dust. Seedream’s wins came when a specific visual cue needed to read unmistakably. Its chrome-and-water reflections made the window and red apple easier to see, and its ukiyo-e image stayed focused on a single fishing boat, bold foam, and a restrained indigo-and-cream palette. Those are meaningful wins, not consolation points: Seedream can be the better choice when a brief hinges on a tightly specified visual motif. The ties underline how close the models are. Google’s rain-slick depot scene had stronger atmosphere and framing; Seedream’s close-up made dread more immediate. For the bracelet task, Google better conveyed the candid moment, while Seedream gave the hands and threads a cleaner close-up. **Call: Nano Banana 2 wins on balance and reliable brief-following, with a modest edge—not enough separation to call it a clear-cut victory.**
Golden hour convoy
A wide cinematic landscape of three experimental desert rally vehicles cresting the chalk ridges above Brindle Salt Flats during the last six minutes of golden hour, the low sun skimming across body panels and throwing long apricot shadows, with airborne dust glowing amber and the mood feeling triumphant yet fleeting. Painterly realism with crisp silhouettes, warm sun from frame-left, distant heat haze, expansive sky, layered composition emphasizing time-of-day light, 16:9.
OpenAI: GPT-6 Luna: Model A more fully captures the painterly realism, crisp vehicle silhouettes, warm frame-left light, and dramatic amber dust, with the convoy clearly cresting a ridge. Model B has a strong salt-flat setting and polished detail, but its brighter, more washed-out look feels less cinematic and less aligned with the requested mood. (Second judge pass, order swapped — scores are the average of both: Model A better captures the painterly, cinematic urgency with a strong diagonal convoy, glowing dust, | Anthropic: Claude Fable 5.1: Model A nails the painterly-realism brief with c…
Reflections & glass
A photorealistic close-up of a chrome teapot on a polished dark marble counter next to a glass of water; the window and a red apple on the counter must be correctly reflected in BOTH the chrome and the water's surface, dramatic side light, 16:9.
OpenAI: GPT-6 Luna: Model B more clearly shows the window and apple reflected in the chrome and water, with dramatic side lighting and a focused close-up composition. Model A is attractive and photorealistic, but the required reflections in the water are less evident. (Second judge pass, order swapped — scores are the average of both: Model B more clearly shows the window and red apple reflected in both the chrome and the water, with dramatic side lighting and a strong close-up composition. Model A is polished and | Anthropic: Claude Fable 5.1: Model B delivers the dramatic side light and sho…
Cargo color lock
A clean, stylized product-scene illustration of four freight items arranged on a loading dock with clear separation: a matte cobalt-blue ceramic fuel canister, a glossy saffron-yellow rubber traffic baton, a brushed copper-red metal toolbox, and a pale mint-green frosted glass navigation buoy, each object keeping its exact own color and material without bleeding or swapping attributes. Precision-focused commercial render style, front three-quarter view, soft overcast daylight, tidy shadows, restrained background of stacked pallets, 16:9.
OpenAI: GPT-6 Luna: Model A presents all four distinct items fully in frame with the requested color assignments and a clear loading-dock arrangement, though some materials read more like plastic than the specified ceramic or rubber. Model B has strong material rendering, but its objects are cropped at the bottom and its lighting is less consistent with the requested soft overcast scene. (Second judge pass, order swapped — scores are the average of both: Model B better matches the clean, restrained product-render b | Anthropic: Claude Fable 5.1: Model A delivers all four objects with correct,…
Platform dread
A cinematic close-up portrait of a night-shift tram conductor on the rain-slick platform at Kestrel Quay Depot, one hand gripping a punched-paper timetable, convincingly showing barely-contained dread after seeing the destination board suddenly change to an impossible route number; the expression must read instantly in the eyes, brow, mouth, and neck tension, with wet lashes, cold breath, and subtle trembling realism. Gritty live-action film still style, shallow depth of field, sodium-vapor station lights mixed with distant teal signal glow, off-center composition, 16:9.
OpenAI: GPT-6 Luna: Model A better captures the rain-soaked depot atmosphere, off-center framing, teal-and-sodium lighting, and a conductor holding a timetable, with convincing tension in the face. Model B gives a more overtly frightened close-up and visible cold breath, but its setting feels less rain-slick and the composition is more centered. (Second judge pass, order swapped — scores are the average of both: Model B more convincingly delivers the requested close-up and instantly readable dread, with a tense fac | Anthropic: Claude Fable 5.1: Model A nails more of the brief: strong sodium-…
Hands & anatomy
A natural candid photo of a person tying a friendship bracelet, both hands clearly visible with anatomically correct fingers and a plausible grip on the threads, soft daylight, shallow depth of field, 16:9.
OpenAI: GPT-6 Luna: Model A has a warm candid setting and clearly shows the bracelet being tied around a wrist, but the hands and threads are somewhat cluttered and partly obscured. Model B gives a clearer close-up of both hands with a plausible grip and well-defined threads, though its tight crop provides less environmental context. (Second judge pass, order swapped — scores are the average of both: Model B is a crisp close-up with a plausible grip, but Model A more clearly conveys the candid bracelet-making momen | Anthropic: Claude Fable 5.1: Model A delivers a convincing candid scene with…
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 B more directly centers a lone fishing boat beneath a dramatic cresting wave, with classic ukiyo-e foam, bold outlines, and a restrained indigo-and-cream palette. Model A includes a clearer dawn setting but adds several boats and a busier scene that weakens the prompt’s focused composition. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the requested single fishing boat, muted indigo-and-cream palette, and bold, stylized ukiyo-e wave. Model | Anthropic: Claude Fable 5.1: Model B adheres more tightly to the requested s…
Spatial layout
A clean isometric illustration of a bedroom: a bed against the LEFT wall, a round rug centered on the floor, a desk under the WINDOW on the back wall, and a floor lamp in the FRONT-RIGHT corner. Flat-vector style, consistent perspective.
OpenAI: GPT-6 Luna: Model A closely follows the requested isometric flat-vector style and places the bed, centered round rug, desk under the back-wall window, and lamp in the front-right corner. Model B includes similar furnishings but uses a realistic, non-isometric style and its lamp is less clearly placed in a front-right corner. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested isometric flat-vector style and places the bed, centered round rug, desk beneath the ba | Anthropic: Claude Fable 5.1: Model A delivers a true isometric flat-vector r…
Rivets and weave
An ultra-detailed studio photograph of a restored 1931 Veloric mail plane nose section beside a mechanic’s stool draped with a frayed indigo herringbone jacket and thick wool gloves, every surface rendered crisply: brushed aluminum with fine scratches, rows of oxidized rivets, cracked varnish on a wooden prop blade, grease in screw heads, fabric weave on the jacket, individual wool fibers, and dust caught in seams, with no smudging or melted detail. Neutral museum lighting, three-quarter angle, high micro-contrast, dark charcoal backdrop, 16:9.
OpenAI: GPT-6 Luna: Model A more clearly depicts the labeled 1931 Veloric mail plane nose section beside a mechanic’s stool, with the jacket, gloves, exposed engine, and weathered prop all visible. Model B has crisp rivets and fabric texture, but its tighter crop obscures the plane’s identity and leaves less of the stool and jacket in view. (Second judge pass, order swapped — scores are the average of both: Model A gives the scene a stronger three-quarter composition, with the aircraft nose, exposed mechanical deta | Anthropic: Claude Fable 5.1: Model A delivers the richer scene—exposed radia…
Matchup powered by OpenRouter.