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 across crowded scenes, precise visual styles, typography, perspective, and close-up craft work. The comparison turns on which model more reliably puts the requested details in the frame—not simply which image looks polished.
Google: Nano Banana 2 (Gemini 3.1 Flash Image) takes the saved task sample, with an aggregate score of **64.0** to Kling Image’s **56.0**. It won six tasks, Kling won one, and one was a tie: a decisive editorial result for this sample, not a statistical claim. Nano Banana 2’s advantage was often literal prompt coverage. It more clearly staged the archive workers’ assigned actions, built the requested campsite and ukiyo-e fishing scene, and handled the bakery’s tiny braided saffron rolls more faithfully. It also came out ahead on the lab poster and one-point aisle, though the poster’s heading still broke across two lines and some labels were imperfect. Kling’s win was the relieved machinist: its closer three-quarter portrait better delivered the requested expression, calipers, shavings, and warm task light. The hands-and-anatomy task was a tie; Nano Banana 2 made the bracelet-making action legible, but neither image was an unambiguous success on the working hands and threads. **Verdict: Google: Nano Banana 2 (Gemini 3.1 Flash Image) wins this saved task sample, 64.0–56.0.** Its broader edge was matching specified actions and scene details; Kling’s strongest counterexample was a more intimate, prompt-faithful machinist portrait. This assessment applies to the saved task sample, not to image-model capability in general.
Archive team scene
A detailed stop-motion claymation scene of four records-office workers reorganizing a flood-damaged local shipping archive: one kneels by open crates of warped ledgers, one stands on a rolling ladder passing labeled boxes, one scans barcodes at a cart, and one points to a wall map with colored route strings; the room feels coherent and busy but balanced, with shelves, fans, drying papers, and a central aisle clearly readable, cinematic overhead-and-side lighting, wide composition, 16:9.
OpenAI: GPT-6 Luna: Model B more clearly depicts the four requested tasks, including a worker kneeling beside damaged ledgers and another pointing to a route map, while Model A has a strong cinematic aisle but misses or obscures some of those actions. Model B is busier and its labels are imperfect, but its flood-damaged archive details and role-specific activity align more closely with the prompt. (Second judge pass, order swapped — scores are the average of both: Model B more completely depicts the requested four | Anthropic: Claude Fable 5.1: Model B realizes all four specified actions—knee…
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-making process, with both hands working the threads in a natural outdoor candid scene. Model A has pleasing soft light and a strong close-up, but the action is less legible and the thread grip feels less convincing. (Second judge pass, order swapped — scores are the average of both: Model B has a convincing candid setting and clearly shows bracelet-making, though the working hands and threads are somewhat tangled visually. Model A presents both hands more | Anthropic: Claude Fable 5.1: Model B delivers a far more convincing candid sc…
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 campsite scene with a cohesive flat-vector style and a clear four-color palette, though it reads more as night than dusk. Model A includes the requested campsite elements but has a less coherent composition and appears to introduce extra color variation. (Second judge pass, order swapped — scores are the average of both: Model B more fully depicts a dusk campsite, with a tent, campfire, campers, and layered mountains arranged in a cohesive flat-vector s | Anthropic: Claude Fable 5.1: Model A sticks to a mostly strict four-color lo…
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 clearly depicts a fishing boat riding the cresting wave at dawn, with a strong ukiyo-e composition and convincing flat colors, bold outlines, and stylized foam. Model A has excellent wave detail and palette, but the boats are less clearly the central subject. (Second judge pass, order swapped — scores are the average of both: Model B clearly shows a fishing boat with its crew riding beneath a dramatic cresting wave, and its dawn sun, indigo-and-cream palette, bold contours, and styl | Anthropic: Claude Fable 5.1: Model B delivers a coherent ukiyo-e composition…
Relieved machinist
A close three-quarter portrait of a lathe operator in a small precision parts workshop at the exact moment of visible relief after a difficult measurement finally matches spec, eyebrows softened, mouth slightly open in a breathy laugh, shoulders dropping, a trace of stress still fading from the face; photorealistic style, metal shavings and calipers in the blurred background, cool industrial light mixed with warm task lamp glow, shallow depth of field.
OpenAI: GPT-6 Luna: Model A better matches the requested close three-quarter portrait, with a clear relieved expression, calipers, metal shavings, and the warm task lamp against cool workshop light. Model B conveys the laugh and workshop context well, but its wider composition and deeper focus miss the prompt’s intimate portrait and shallow depth of field. (Second judge pass, order swapped — scores are the average of both: Model B more clearly captures the breathy, relieved laugh and shows the operator checking a m | Anthropic: Claude Fable 5.1: Model A delivers the requested close three-quar…
Typography poster
A crisp editorial photo of a break-room corkboard in a municipal water-testing lab, centered on a freshly pinned safety poster with perfectly legible sans-serif text that reads exactly: "NIGHT SHIFT FILTER CALIBRATION" on the top line, "BAY 7" beneath it in larger type, and a small bottom strip reading "11:40 PM — WEAR NITRILE GLOVES"; realistic paper texture, pushpins, coffee mug and clipped lab checklist nearby, clean fluorescent lighting, straight-on composition, 16:9.
OpenAI: GPT-6 Luna: Model B has the stronger lab setting, balanced composition, and a clearly rendered bottom strip; Model A is crisp and readable but crops the surrounding context and lacks the requested strip treatment. Both split the long heading across two lines rather than keeping it on one line. (Second judge pass, order swapped — scores are the average of both: Model B more fully captures the municipal lab break-room setting, with the corkboard, clipped checklist, coffee mug, and cleanly legible requested wo | Anthropic: Claude Fable 5.1: Model A misspells the bottom strip as 'NITRTLE'…
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 B presents a more polished, warmly lit aisle with a clear central vanishing point and convincing scale recession. Model A also follows the perspective brief well, but its lighting and overall finish are less inviting. (Second judge pass, order swapped — scores are the average of both: Model B has a particularly clear, balanced one-point perspective, with shelf and book scale receding consistently toward the central vanishing point and warm overhead lighting. Model A also follows the prompt | Anthropic: Claude Fable 5.1: Both deliver a clean centered one-point perspec…
Baker hands close view
A realistic waist-up image of a bakery owner at a stainless-steel worktable shaping twelve tiny braided saffron rolls, both hands fully visible and anatomically correct with natural finger lengths and believable joints, a dusting of flour on the knuckles, relaxed shoulders and accurate arm proportions, apron lightly creased, trays and scale in the background; warm morning window light, documentary food-photography style, slightly angled composition, 16:9.
OpenAI: GPT-6 Luna: Model B better depicts tiny braided saffron rolls being shaped and includes bakery equipment in a documentary-style setting, though it shows two bakers and the hands are less prominent than requested. Model A has appealing warm light and a close view of one baker’s hands, but the baker is rolling a large sheet of dough and the finished rolls are not braided. (Second judge pass, order swapped — scores are the average of both: Model B more closely matches the waist-up bakery scene and shows a tray | Anthropic: Claude Fable 5.1: Model A delivers the close hands-focused view w…
Matchup powered by OpenRouter.