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
Eight image-generation tasks put these models through garden scenes, tightly specified compositions, exact counting, and named visual styles. The comparison rewards not just attractive images, but how reliably each one follows the brief.
Google: Nano Banana 2 takes the matchup, scoring **66.0 to Kling Image’s 57.3**, with an **limited confidence** verdict. It won five tasks to Kling’s two, with one tie—a clear lead, built on stronger results across varied and demanding briefs rather than a single standout image. Nano Banana 2 was more reliable when a prompt demanded several things at once. It made the rooftop garden feel lived-in, rendered a legible bracelet-tying moment, stayed closer to a strict four-color palette, and more convincingly evoked ukiyo-e. Its seed-packet poster was also the more complete design, though the “MOONFLOWER” headline didn’t match the specified datura. Kling’s wins were precise ones: its potting-shelf image had the cleaner, more controlled product arrangement, and its overhead espresso lineup showed exactly seven cups with coffee visible. Those are meaningful advantages for prompts where orderly presentation or countable objects matter. But they didn’t offset Nano Banana 2’s broader consistency. The golden-hour allotment was a tie: Kling had richer sunset light, while Nano Banana 2 better conveyed the requested medium-wide documentary scene. **Verdict: Google: Nano Banana 2 wins on range and brief-following; Kling Image is the sharper pick for clean product layouts and exact-count compositions.**
Golden Hour Allotment
A documentary-style image of an allotment garden at golden hour, 16:9: an older gardener in a rust-red sweater tying tomato vines to bamboo canes, long warm sunlight skimming across dew-dulled leaves, backlit dust motes, deep stretched shadows between soil rows, and a distant line of sheds glowing amber; the lighting should clearly convey the last fifteen minutes before sunset, calm and reflective mood, medium-wide composition.
OpenAI: GPT-6 Luna: Model A more strongly conveys the final warm, low sunlight with glowing sheds, deep row shadows, and a focused documentary moment of the gardener tying vines. Model B has an appealingly broad allotment view, but its brighter sky and less pronounced late-sunset lighting feel slightly less reflective. (Second judge pass, order swapped — scores are the average of both: Model B more fully captures the requested medium-wide documentary scene, with the gardener tying the vines, clear soil-row shadows, | Anthropic: Claude Fable 5.1: Model A delivers a more cohesive golden-hour lo…
Color-Bound Potting Shelf
A meticulously arranged still life on a greenhouse potting shelf, 16:9, in a clean editorial product-photography style: a matte cobalt-blue ceramic watering can, a brushed copper hand trowel, a pale mint-green glass mister, a raw terracotta seed tray, and a folded mustard-yellow waxed-canvas apron; each object must keep its own exact color and material with no color swapping or bleeding; front-facing composition, soft diffused morning light, sharp detail, neutral backdrop of blurred greenhouse panes.
OpenAI: GPT-6 Luna: Model A presents all five objects with the requested colors and materials in a polished, clean editorial arrangement, while Model B is more cluttered and adds an unrequested label to the apron. Model B has a convincing greenhouse setting, but Model A’s softer, more controlled composition better matches the prompt. (Second judge pass, order swapped — scores are the average of both: Model A has a cleaner, more front-facing product arrangement and a polished editorial look, while keeping the reques | Anthropic: Claude Fable 5.1: Model A follows the brief closely with a matte…
Bench With No Blossoms
Documentary photojournalism, 16:9: a community rooftop garden in early spring with three galvanized raised beds, damp potting soil, hand-lettered plant markers, a green watering can tipped on its side, and a cedar potting bench under a cloudy sky; deliberately show that there are NO flowers or blossoms anywhere in the scene, only seedlings, bare stems, and leafy starts; eye-level composition, crisp natural light, realistic textures.
OpenAI: GPT-6 Luna: Model B more clearly conveys a lived-in rooftop garden, with the tipped watering can, three raised beds, and a potting bench all visible; neither image shows blossoms. Model A is cleaner, but its can appears upright and its bench reads more like a simple plank than a potting bench. (Second judge pass, order swapped — scores are the average of both: Model B more fully matches the requested rooftop garden, with three galvanized beds, a tipped green watering can, a cedar potting bench, cloudy skies | Anthropic: Claude Fable 5.1: Model B satisfies more of the brief—the green w…
Midnight Pollinator Poster
Flat vector illustration of a moon-garden scene for a seed-packet poster: a white datura flower opening beside dark teal foliage while a single hummingbird hawk-moth hovers in front of it; mood is hushed, dreamy, and slightly mysterious; crisp geometric shapes, limited palette, elegant negative space, subtle grain texture, clean poster composition.
OpenAI: GPT-6 Luna: Model B more fully reads as a finished seed-packet poster, with a clear datura, hovering hawk-moth, moonlit palette, and polished typography, though its “MOONFLOWER” headline is slightly at odds with the specified flower. Model A has strong vector shapes and a lovely flower, but its inset landscape treatment feels less like a cohesive poster composition. (Second judge pass, order swapped — scores are the average of both: Model B more fully delivers the seed-packet poster brief, with a clear datu | Anthropic: Claude Fable 5.1: Model B delivers a convincing seed-packet poste…
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 presents a more complete, balanced mountain campsite at dusk, with strong visual detail and a palette that stays much closer to the requested four colors. Model A is cleanly vector-styled, but its gradient sky and vivid red/yellow fire depart from the palette restrictions, and the campsite feels less balanced. (Second judge pass, order swapped — scores are the average of both: Model B more fully delivers the dusk campsite scene and stays close to the specified four-color palette, with a | Anthropic: Claude Fable 5.1: Model A breaks the four-color restriction with a…
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 B more clearly shows the bracelet being tied onto a wrist, with a natural candid setting and convincing interaction between both people. Model A has a strong close-up of the hands and soft depth of field, but the tying action is less legible. (Second judge pass, order swapped — scores are the average of both: Model B more clearly depicts a candid bracelet-tying moment outdoors, with both working hands and the recipient’s wrist visible in soft daylight. Model A has a clean close-up and plau | Anthropic: Claude Fable 5.1: Model A renders two large, well-formed hands pi…
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 captures ukiyo-e woodblock conventions, with stylized foam, flat indigo-and-cream planes, and a dawn seascape; its dramatic wave and layered boats make a stronger composition. Model A clearly shows a boat riding a cresting wave, but its gradients and cleaner modern illustration feel less faithful to the named tradition. (Second judge pass, order swapped — scores are the average of both: Model B more convincingly evokes ukiyo-e through its dramatic curling wave, stylized | Anthropic: Claude Fable 5.1: Model B is far more authentic to the ukiyo-e tr…
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 shows exactly seven distinct, matching espresso cups in a straight overhead row, with espresso visible and a polished marble setting. Model B also has seven cups, but they are empty and their handles nearly touch the neighboring cups, making the separation less clear. (Second judge pass, order swapped — scores are the average of both: Model A shows exactly seven clearly countable espresso cups in a straight overhead row, with espresso visible and a polished marble setting. Model B also h | Anthropic: Claude Fable 5.1: Both Model A and Model B deliver exactly seven…
Matchup powered by OpenRouter.