Head to head: Wan vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
Wan vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
By Ryan Merket · Published
Wan and Google’s Nano Banana 2 face off across eight image-generation tasks, from a ukiyo-e scene and rainy transit imagery to close-up product detail. The comparison tests how well each model balances prompt fidelity, composition, and visual polish.
Google’s Nano Banana 2 (Gemini 3.1 Flash Image) leads the saved results with an aggregate score of **64.5** to Wan’s **55.4**. It won seven tasks, tied one, and lost none—a decisive result on this task sample, not a claim of statistical certainty or a general ranking of image models. The wins came from varied, concrete strengths: Google made the fractured mirror, wet canopy, and approaching tram read more clearly; gave the monorail scene a more convincing elevated curve and roof action; and handled the gearshift macro with sharper subject focus and softer gauge bokeh. Its harbor scene also communicated the amphibious taxi more directly, while its bracelet image better conveyed hands actively tying it. The sole tie was **Named art style**. Wan’s muted indigo-and-cream palette and simpler focal composition had real appeal, while Google’s stronger woodblock cues—including bold outlines and stylized foam—made the scene feel more like ukiyo-e. In **Restricted palette**, Wan showed tighter color discipline, but Google’s campsite was clearer and more balanced. Neither model was flawless: the reading-nook outputs drew concern over greenery despite the no-plants constraint. **Verdict: Google’s Nano Banana 2 wins this matchup on the saved task sample, with stronger scene readability and prompt execution across the seven decided tasks.**
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 convincingly evokes a traditional ukiyo-e print, with a dramatic cresting wave, stylized foam, and a clear dawn setting. Model A better matches the muted indigo-and-cream palette and has a simpler focal composition, but Model B’s richer print-like scene is more compelling overall. (Second judge pass, order swapped — scores are the average of both: Model B more fully delivers the ukiyo-e woodblock look, with bold outlines, stylized foam, a prominent fishing boat and fishermen, and a | Anthropic: Claude Fable 5.1: Model A keeps strictly to the requested muted in…
Chrome Tram Canopy
An ultra-detailed realistic scene inside the glass canopy of the Aurelian Loop tram stop during a light drizzle, featuring a polished chrome tram nose approaching slowly, transparent side panels beaded with rain, and a fractured mirror advertisement column reflecting commuters, signal lights, and the tram at correct angles; layered refractions through wet glass, subtle double reflections in puddles, cool overcast light with red route LEDs punctuating the scene, centered perspective, 16:9
OpenAI: GPT-6 Luna: Model B more fully captures the fractured mirror advertisement, rain-beaded glass, commuters, red signals, and reflective tram, with convincing detail and legible route text. Model A has stronger centered symmetry, but its mirror column is not visibly fractured and its signage is less coherent. (Second judge pass, order swapped — scores are the average of both: Model B more fully captures the polished approaching tram, rain-beaded canopy, and fractured mirror column reflecting commuters and sign | Anthropic: Claude Fable 5.1: Model B delivers a genuinely fractured mirror c…
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 and Model B both depict a cozy reading nook without visible lamps, artwork, or picture frames, though both show greenery outside the window despite the strict no-plants requirement. Model B has a more complete, balanced scene with a prominent window, armchair, tea, and multiple book stacks, giving it the compositional edge. (Second judge pass, order swapped — scores are the average of both: Model B has a fuller, inviting reading-nook composition with the requested chair, books, tea, and | Anthropic: Claude Fable 5.1: Both Model A and Model B correctly omit plants,…
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 more directly shows both hands of one person plausibly pinching and guiding the bracelet threads, with convincing anatomy and soft daylight. Model B has an attractive candid composition, but appears to show another person tying the bracelet onto a wrist, making it slightly less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B more directly shows a bracelet being tied around a wrist, with both people’s hands clearly engaged in the task an | Anthropic: Claude Fable 5.1: Model B reads as a genuinely candid scene with…
Restricted palette
A flat-vector illustration of a mountain campsite at dusk using ONLY four colors — cream, burnt orange, teal, and charcoal — with no gradients and no stray off-palette colors, balanced composition, 16:9.
OpenAI: GPT-6 Luna: Model B more clearly depicts a balanced mountain campsite at dusk, with a cohesive restricted palette and polished flat-vector styling. Model A has a suitable palette and graphic style, but its textured cream background, unusual central shapes, and crowded arrangement weaken adherence and clarity. (Second judge pass, order swapped — scores are the average of both: Model B presents a clear, polished mountain campsite at dusk with a balanced, layered composition and a convincing restricted palette | Anthropic: Claude Fable 5.1: Model A keeps a tighter four-color discipline (…
Monorail Sparks
A midnight maintenance monorail labeled Kestrel-9 bursts out of a rain-slick service tunnel on a steep elevated curve, captured mid-action as orange brake sparks scatter beneath the wheels and a mechanic in a lime safety harness leaps between two roof panels, coat whipping sideways in the wind; cinematic realism, cyberpunk neon accents from violet station signs and teal puddle reflections, hard backlight through mist, motion-rich composition with a low tracking angle and diagonal energy, 16:9
OpenAI: GPT-6 Luna: Model A captures the tunnel exit, rain, sparks, and Kestrel-9 labeling, but the train’s path feels less curved and the mechanic’s jump reads as a leap off the side. Model B has a stronger elevated curve, more dynamic cinematic depth, and a mechanic moving across the roof panels, though the tunnel-exit action is slightly less explicit. (Second judge pass, order swapped — scores are the average of both: Model B more fully captures the prompt with the rain-slick elevated curve, tunnel exit, vivid b | Anthropic: Claude Fable 5.1: Model B nails the core action—mechanic in a lim…
Harbor Vector Calm
A clean vector illustration of a compact amphibious taxi boat named Mossfin idling beside a foggy basalt harbor at blue hour, with rounded geometric forms, limited palette of sea-glass green, cream, slate, and one warm amber cabin light; serene slightly melancholic mood, crisp negative space, simplified ripples, gull silhouettes, and neatly stacked lobster crates creating a balanced poster-like composition
OpenAI: GPT-6 Luna: Model B more clearly conveys an amphibious taxi boat through its track-like hull, and its misty basalt harbor, amber cabin light, gulls, and neatly grouped crates make a cohesive, atmospheric scene. Model A has a strong poster-like layout and clear Mossfin branding, but its harbor details feel busier and the amphibious quality is less apparent. (Second judge pass, order swapped — scores are the average of both: Model B more directly captures the foggy basalt harbor, amphibious taxi boat, amber c | Anthropic: Claude Fable 5.1: Model B delivers the key distinctive content—an…
Lens on Gearshift
Macro close-up photograph of the worn brass-and-resin gear selector inside a restored 1963 snow tractor, shot with a true macro lens at extremely shallow depth of field so only the engraved shift numbers and a tiny crack in the amber resin knob are tack sharp while frost crystals on the dashboard gauges melt into creamy bokeh; cold dawn light entering from the left, intimate off-center framing, tactile textures of brushed metal, chipped paint, and stitched leather
OpenAI: GPT-6 Luna: Model B better captures the extreme macro focus, with the amber resin knob and legible shift pattern crisp against creamy, frosty gauge bokeh. Model A has strong wintry textures, but its engraved markings are garbled and the gauges remain too distinct for the requested shallow depth of field. (Second judge pass, order swapped — scores are the average of both: Model B better captures the intimate macro framing, shallow depth of field, cracked amber resin, worn materials, and softly blurred frosty | Anthropic: Claude Fable 5.1: Model B renders a convincing shift-pattern engr…
Matchup powered by OpenRouter.