Head to head: Wan vs Google: Nano Banana Pro (Gemini 3 Pro Image)
Wan vs Google: Nano Banana Pro (Gemini 3 Pro Image)
By Ryan Merket · Published
Wan and Google’s Nano Banana Pro go head to head across eight image-generation tasks, from room layouts and anatomy to detailed scenes and poster text. The comparison rewards both prompt fidelity and the clarity of the finished image.
Google’s Nano Banana Pro (Gemini 3 Pro Image) wins this saved task sample with an aggregate score of **65.2** to Wan’s **57.6**. It took six task wins, compared with Wan’s one, with one tie. The editorial result is decisive on this sample—not a statistical claim—and the scores should not be read as a general ranking of the models beyond these tasks. Google’s edge came from range and finish: it won on spatial layout, the negation prompt, Alpine Loom Detail, Cliffside Leap, Harbor Market Binding, and Ferry Poster Text. Its carved cedar chair, indigo-and-rust weaving, and dawn-versus-lamplight contrast stood out in Alpine Loom Detail; in the poster task, it made the moonlit bay and ferry more compelling while keeping the requested text prominent. Its market image also separated the five objects cleanly, with more convincing material textures. Wan had a clear bright spot in Hands & Anatomy, where the hands and finger structure read more convincingly and the grip on the threads was clearer. Perspective & Scale was a tie: Wan’s centered aisle and consistent recession were easier to read, while Google’s more photorealistic shelves and foreground columns made the scene feel less open. Those results show why the aggregate lead is not a claim that Google won every kind of image task. The verdict is specific to the saved eight-task sample. Google’s combination of prompt coverage and polished scene-making carried the comparison, while Wan showed an advantage on a focused anatomy prompt and matched Google on perspective.
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 places the bed, centered rug, desk beneath the window, and floor lamp clearly in the requested positions, with a crisp, consistent flat-vector isometric style. Model B also includes the main elements, but its oversized rug overlaps the bed area and its arcing lamp makes the composition less clean. (Second judge pass, order swapped — scores are the average of both: Model B places the desk directly beneath the window and keeps the bed against the left wall, while Model A’s desk is clearly | Anthropic: Claude Fable 5.1: Model B correctly places the desk directly under…
Perspective & scale
A photorealistic one-point-perspective shot down a long empty library aisle, shelves receding to a single vanishing point, books getting consistently smaller with distance and nothing warped or wrongly sized, warm overhead light, 16:9.
OpenAI: GPT-6 Luna: Model A presents a cleaner, more centered one-point perspective, with the shelves and lights converging clearly and the books scaling consistently into the distance. Model B is richly detailed and photorealistic, but its prominent foreground columns make the aisle feel less open and the perspective less immediately clear. (Second judge pass, order swapped — scores are the average of both: Model A has the clearer, more balanced one-point perspective, with the shelves and books receding consistent | Anthropic: Claude Fable 5.1: Model A delivers a textbook single-vanishing-po…
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 shows both hands clearly with a more convincing finger structure and plausible grip on the threads, making it the stronger fit for the hands-and-anatomy focus. Model B has a pleasing candid setting and strong shallow depth of field, but its fingers are less clearly defined around the bracelet. (Second judge pass, order swapped — scores are the average of both: Model B shows the person and a colorful bracelet in a pleasing candid setting, but the hands are less clearly focused and the gri | Anthropic: Claude Fable 5.1: Model A delivers large, clearly visible hands w…
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 B better matches the requested 16:9 framing and includes all the cozy reading-nook elements, though visible greenery outside the window conflicts with the no-plants constraint. Model A is attractive and well rendered, but shows more obvious foliage outside and does not match the requested aspect ratio as closely. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested 16:9 framing, while both images include visible vegetation through the win | Anthropic: Claude Fable 5.1: Both Model A and Model B satisfy the negation c…
Alpine Loom Detail
An ultra-detailed close travel scene inside a mountain guesthouse in the remote Vald Sere region: a handwoven indigo-and-rust blanket draped over a carved cedar chair beside a window, every wool fiber, tight weave, fringe knot, wood grain, and tiny brass tack rendered crisply; outside the window, sharply defined larch needles and distant rocky ridges remain visible but secondary, cool dawn light mixing with a small amber bedside lamp, high-resolution realism with meticulous texture fidelity, 16:9
OpenAI: GPT-6 Luna: Model B better captures the carved cedar chair, nuanced indigo-and-rust weaving, visible mountain landscape, and cool-dawn/amber-lamp contrast. Model A renders the blanket and wood crisply, but the brass tacks appear oddly placed on the blanket and the chair is less distinctly carved. (Second judge pass, order swapped — scores are the average of both: Model B more convincingly captures the richly draped indigo-and-rust blanket on a carved cedar chair, with striking cool dawn light balanced by an | Anthropic: Claude Fable 5.1: Model B delivers a genuinely carved cedar chair…
Cliffside Leap
A travel-adventure photograph of a solo parkour runner in a saffron windbreaker leaping between two narrow sea cliffs above the stormy turquoise Zhali Coast, captured mid-air with knees tucked and one hand reaching forward, loose gravel and a trailing scarf showing motion, waves exploding far below, late-afternoon sun cutting through mist from camera left, dramatic wide-angle composition with the horizon low and the athlete frozen sharply against streaked clouds, cinematic realism, 16:9
OpenAI: GPT-6 Luna: Model B more clearly captures the forward-reaching, knees-tucked leap, with dramatic late-afternoon light and a strong wide-angle view of the surf below. Model A also fits the scene well, but Model B’s extra coast sign is distracting and not requested. (Second judge pass, order swapped — scores are the average of both: Model A makes the leap clearly read as a crossing between two narrow cliffs, with flying gravel, turbulent water, and a dramatic storm-lit setting. Model B has a strong tucked pos | Anthropic: Claude Fable 5.1: Model B nails more of the prompt's specifics—kn…
Harbor Market Binding
A carefully staged still life in a small coastal market on the island of Brin Varo: a matte cobalt ceramic teapot, a brushed copper lantern, a translucent emerald glass bottle, a pale pink linen sunhat, and a jet-black rubber rain boot arranged on a weathered driftwood table, each object clearly separated so its own color and material are unmistakable and do not swap or bleed, soft overcast daylight, realistic travel editorial style, three-quarter view, 16:9
OpenAI: GPT-6 Luna: Model B presents all five objects clearly separated on a weathered table with a convincing, realistic editorial look, though the coastal setting is less explicit. Model A makes the seaside market more obvious, but its objects look more synthetic and the boot has distracting text artifacts. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested three-quarter travel-editorial view, with convincing material textures and clearly separated objects in | Anthropic: Claude Fable 5.1: Both renders deliver all five objects with corre…
Ferry Poster Text
A vintage-inspired poster mounted on a ferry terminal wall in Port Lenora, designed in clean mid-century typography and easy to read from a few meters away, featuring an illustrated moonlit bay and the exact headline text: "NORTH QUAY FERRY" with a smaller line beneath: "GATE 3 · 06:40", crisp legible lettering, balanced layout, screen-printed texture, warm indoor tungsten light with slight reflections on the glass frame, straight-on composition, 16:9
OpenAI: GPT-6 Luna: Model B more clearly depicts a moonlit bay with a ferry and presents the requested text prominently in a polished vintage poster, while Model A’s moon reads more like a sun and its poster is less visually compelling. Model A is also legible and well-framed, but Model B has the stronger composition and atmosphere despite some glass reflections. (Second judge pass, order swapped — scores are the average of both: Model B better captures the moonlit bay with a clearly illustrated ferry, exact legibl | Anthropic: Claude Fable 5.1: Both render the headline and gate line exactly,…
Matchup powered by OpenRouter.