Head to head: Kling Image vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
Kling Image vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
By Ryan Merket · Published
A prompt-following face-off tests these image models on anatomy, exact counts, legible text, constrained palettes and close-up food photography. The results show where reliable scene execution matters—and where a tighter artistic frame can still win.
Google’s Nano Banana 2 takes this matchup on execution, not just polish: it scores 67.0 to Kling Image’s 52.9, wins six of eight tasks, and earns a decisive 96% statistical verdict. Its advantage is clearest when the prompt sets specific requirements that are easy to miss or muddle. Nano Banana 2 gets the requested electric kettle and visible spiral infuser right, lays out exactly seven complete cups, and renders all three poster lines with the requested en dash. It also tracks the attic repair-bench details and the named pantry palette more faithfully. Across these tasks, Kling’s images can look polished, but polish doesn’t make up for a teapot-like infuser, cropped cups, or lighting that strays from the brief. Kling’s win is specific and earned: its jam-jar image commits to a true macro, with sharper sugar crystals and glossy fruit while breakfast elements fall away into bokeh. Nano Banana 2 keeps more of the kitchen in view, which weakens the close-up effect. The reading-nook prompt ends in a tie: both models produce inviting scenes, but the judges disagreed on whether either violated the no-plants constraint. **Final call: Google’s Nano Banana 2 is the stronger choice for prompts with concrete visual constraints; Kling Image is the better pick here for tight, tactile macro framing.**
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 gives a clear, tightly focused view of both hands with a plausible grip, while Model B more clearly conveys the candid bracelet-making activity and has a lively natural-light setting. Model B’s hand pose is somewhat crowded, but the overall scene feels more complete and engaging. (Second judge pass, order swapped — scores are the average of both: Model B more clearly shows the bracelet being worked with several colorful threads and has a natural candid outdoor feel, though the lower hand | Anthropic: Claude Fable 5.1: Model A shows both hands cleanly with a convinc…
Attic Repair Bench Perspective
An isometric-leaning but physically consistent 3D render of a narrow attic repair bench viewed from standing height, where a long wooden ruler, a compact sewing machine, a stack of five nesting storage tins, a dangling bare bulb, and a tiny battery radio all obey one coherent vanishing-point perspective and convincing relative scale; the bench recedes toward a small dormer window, floorboards and rafters align cleanly, nothing is stretched or toy-like, and a striped mug in the foreground is clearly larger than the pincushion and thread snips farther back, illuminated by dusty late-afternoon sunbeams and soft ambient shadow, 16:9
OpenAI: GPT-6 Luna: Model B more fully matches the requested repair-bench scene, with a long ruler on the bench, a clear five-tin stack, coherent perspective, and dusty sunlight; the foreground mug is also clearly larger than the sewing notions behind it. Model A is polished but more toy-like, and its ruler placement and lighting depart from the prompt. (Second judge pass, order swapped — scores are the average of both: Model B closely follows the requested attic-bench scene with convincing perspective, realistic r | Anthropic: Claude Fable 5.1: Model B delivers a physically consistent perspe…
Rainy Conservatory Kettle
A photoreal still life of a smoked-glass electric kettle on a chrome warming tray beside a rain-streaked conservatory window, half-filled with pale mint tea so the curved glass body shows accurate refraction of a submerged spiral infuser and the window grid behind it; include a tiny amber bottle, a folded linen towel, and three kumquats on the wet stone sill, with believable reflections in the chrome tray, faint mirror echoes in the kettle base, and distorted outdoor succulents visible through the droplets on the glass, lit by cool overcast daylight with one warm practical lamp glow from the left, composed as a calm editorial product scene, 16:9
OpenAI: GPT-6 Luna: Model B more closely matches the requested electric kettle and clearly shows the submerged spiral infuser, while retaining the rainy window, succulents, lamp, bottle, towel, and kumquats. Model A has strong reflections and atmosphere, but reads as a teapot with a basket infuser rather than an electric kettle. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the electric smoked-glass kettle, visible spiral infuser, window grid, and requested conserv | Anthropic: Claude Fable 5.1: Model B delivers a genuinely smoked-glass elect…
Exact counting
A clean overhead flat lay on white marble of EXACTLY seven identical espresso cups arranged in a single straight row, each clearly separate and individually countable, even soft lighting, 16:9.
OpenAI: GPT-6 Luna: Model A appears to contain seven cups, but the cups at both edges are cropped, making the row less clearly countable and the composition less polished. Model B shows exactly seven complete, distinct cups in a straight, evenly spaced row on white marble, with clean overhead lighting. (Second judge pass, order swapped — scores are the average of both: Model B clearly shows exactly seven separate, matching espresso cups in a straight row on white marble with soft lighting. Model A also suggests sev | Anthropic: Claude Fable 5.1: Model A shows seven cups but the first and last…
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 renders all three lines clearly and uses the requested en dash in the time, while Model A appears to use a dot separator instead. Both are polished and centered, but Model B’s restrained poster treatment and accurate text better match the prompt. (Second judge pass, order swapped — scores are the average of both: Model B renders all three lines with the requested wording and punctuation, in a clean centered layout on a subtly textured navy poster. Model A is also legible, but changes the | Anthropic: Claude Fable 5.1: Model A renders the text legibly but replaces t…
Hammershade Pantry Palette
A stylized still life illustration of pantry objects arranged on a countertop—two ceramic canisters, a ridged glass oil bottle, a loaf wrapped in paper, a checked dishcloth, and a wooden spoon—rendered strictly in the named palette 'Hammershade' consisting only of oxblood, mustard ochre, soot black, chalk cream, and muted sage, with no other colors anywhere in the image; use crisp gouache-like texture, graphic shapes, and side-lit morning illumination that preserves the palette without introducing blue, purple, orange, or metallic off-tones
OpenAI: GPT-6 Luna: Model B better follows the named palette and the crisp, graphic illustration style, while clearly including all requested pantry objects; its bread still introduces warm brown-orange tones beyond the palette. Model A has a pleasing still-life arrangement, but its strongly brown and orange lighting departs more substantially from the strict palette. (Second judge pass, order swapped — scores are the average of both: Model B includes the requested pantry objects in a crisp, graphic gouache style a | Anthropic: Claude Fable 5.1: Model B uses all five named palette colors (oxb…
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 has a particularly inviting, well-balanced reading nook with the chair, books, and mug clearly arranged by the window; Model A also follows the scene well and has lovely warm light. Both show foliage outside the window, which conflicts with the strict no-plants requirement, but neither includes a lamp or wall artwork. (Second judge pass, order swapped — scores are the average of both: Model B more clearly reads as an inviting reading nook, with an open book and glasses on the armchair al | Anthropic: Claude Fable 5.1: Both outputs correctly omit plants, lamps, and…
Macro Jam Jar Framing
A true macro close-up photograph of the mouth of an open blackberry jam jar on a breakfast table, shot with a 100mm macro lens at very shallow depth of field so the crystalline sugar on the sticky rim and a tiny silver spoon nicking the surface are razor sharp while the toast rack, butter dish, and tiled kitchen beyond dissolve into creamy circular bokeh; natural sunrise light from a nearby window, intimate food-editorial styling, tight framing that fills the image with the jar opening and glossy fruit texture, 16:9
OpenAI: GPT-6 Luna: Model A better delivers the tight macro framing, sharply detailed sugared rim and glossy jam, with breakfast elements softened into bokeh. Model B is appealing but frames the jar more widely and leaves more of the kitchen and table in focus. (Second judge pass, order swapped — scores are the average of both: Model A gives the rim crystals and glossy blackberry texture greater macro clarity, with a more immersive, tightly framed composition. Model B includes more of the breakfast setting and a sp | Anthropic: Claude Fable 5.1: Model A delivers a true tight macro framing whe…
Matchup powered by OpenRouter.