Head to head: Kling Image vs Flux 3 Image
Kling Image vs Flux 3 Image
By Ryan Merket · Published
We put both image models through eight prompts spanning typography, anatomy, perspective, named art direction, and crowded still lifes. The results show how differently they handle exact prompt details versus atmosphere.
Flux 3 Image takes the matchup, 66.3 to Kling Image’s 55.8, with a tentative lean verdict. It won six tasks to Kling’s one, with one tie—a broad advantage, not a squeaker dressed up as a rout by one lucky image. The pattern is stronger prompt fidelity. Flux made the fishing boat, cresting wave, and dawn sun readable in the ukiyo-e scene; handled the five-color Hammershade palette with deliberate color blocking; and got the three-line poster text right, where Kling duplicated “MARKET.” It also made the rainy repair scene and bracelet-making hands easier to read, and kept the pantry still life’s requested objects distinct and accounted for. Kling’s win was a meaningful one: its watchmaker scene more clearly showed the repairer working on an alarm clock at an ironing-board workbench, with the requested props and soft window light. And perspective was a genuine tie. Kling’s empty aisle had the stronger vanishing point and cinematic depth; Flux was crisp and symmetrical, but put a table at the far end despite the empty-aisle brief. That split gets at the verdict: Kling can deliver atmosphere and a compelling composition, but Flux more consistently makes the specific things in the prompt visible and correct. **Our call: Flux 3 Image wins on dependable prompt execution; choose Kling for the watchmaker-style brief where its staging happens to land better.**
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 fully captures the prompt with a clearly readable fishing boat, a dramatic cresting wave, and a visible dawn sun, while retaining ukiyo-e outlines and stylized foam. Model A has a striking wave and fitting palette, but the boats are less legible and the dawn setting is less clearly conveyed. (Second judge pass, order swapped — scores are the average of both: Model B clearly depicts a fishing boat riding the cresting wave at dawn, with a visible sun, muted indigo-and-cream colors, an | Anthropic: Claude Fable 5.1: Model B presents a clear single fishing boat wi…
Watchmaker at the Ironing Board
A realistic editorial-style portrait of a middle-aged watch repairer standing at an ironing board converted into a makeshift workbench, clearly visible from mid-thigh up, carefully rethreading a tiny brass spring into an alarm clock with both hands in frame, showing anatomically correct fingers, natural wrist angles, and believable body proportions; the room contains a spray bottle, lint roller, folded pillowcases, and a green desk lamp, with soft morning side light from a laundry-room window and shallow depth of field emphasizing the face and hands.
OpenAI: GPT-6 Luna: Model A more clearly shows the watch repairer working on an alarm clock at an ironing-board workbench, with the requested laundry items, green lamp, and soft window light. Model B is realistic and includes several matching props, but its tighter crop and indistinct object being repaired make it less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model A more clearly shows the watch repairer working on an alarm clock with both hands, while retaining th | Anthropic: Claude Fable 5.1: Model A delivers the standing mid-thigh-up edit…
Hammershade Palette Radios
A precise product-still-life image of three stacked portable radios, a coiled extension cord, and a ceramic mug on a matte shelf, rendered using only the named Hammershade palette: Prussian blue, burnt sienna, mustard yellow, cream, and lamp black, with absolutely no other hues or stray color contamination; clean studio lighting, crisp edge definition, gentle cast shadows, and a disciplined 16:9 composition that makes palette fidelity immediately judgeable.
OpenAI: GPT-6 Luna: Model B makes the named palette more immediately judgeable, with deliberate color blocking and all requested objects clearly present; Model A is cleanly rendered but its vivid blue backdrop and orange radio feel less tightly controlled to the palette. Model B also has a more balanced, cohesive product-still-life composition. (Second judge pass, order swapped — scores are the average of both: Model B includes all requested objects in a balanced, crisp studio still life and stays closely within th | Anthropic: Claude Fable 5.1: Model B uses all five named colors with a convi…
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 the requested three centered lines accurately with a clean sans-serif treatment and subtle grain. Model A duplicates “MARKET” on a separate line, disrupting the requested layout despite its legible text and navy poster style. (Second judge pass, order swapped — scores are the average of both: Model B renders the requested wording accurately in three centered lines on a deep navy, subtly textured background. Model A duplicates “MARKET” on a separate line and uses a much harsher, b | Anthropic: Claude Fable 5.1: Model A duplicates the headline as 'NIGHT MARKE…
Rainy Umbrella Repair Diagram
A clean flat vector illustration of a compact umbrella repair kit spilled across a hallway runner—mini screwdriver, spare ribs, thread spool, wax pencil, instruction card, and a folded plum-colored umbrella—drawn in a stylish contemporary poster style with sharp shapes, minimal textures, and expressive negative space, conveying a quiet rainy-evening mood through simplified window-glow reflections and orderly composition, 16:9.
OpenAI: GPT-6 Luna: Model B makes the hallway runner, rainy window, plum umbrella, repair tools, and readable instruction card especially clear, with a strong quiet-evening mood. Model A is more faithful to the clean flat-vector style, but its setting feels more abstract and the rainy atmosphere is less apparent. (Second judge pass, order swapped — scores are the average of both: Model B more completely captures the rainy hallway-runner scene, with a rain-streaked window, clearly identifiable repair tools, and a le | Anthropic: Claude Fable 5.1: Model B nails more of the brief—plum umbrella,…
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 gives the hands and the threadwork greater clarity, with a plausible pinching grip and a natural candid feel. Model A is attractive and clearly on prompt, but the hands are smaller and less prominent in the composition. (Second judge pass, order swapped — scores are the average of both: Model B makes the bracelet-making action and both hands the clear focal point, with plausible grips on colorful threads and soft daylight. Model A is polished and shows both hands, but the hands and threa | Anthropic: Claude Fable 5.1: Model B shows a convincing, well-articulated ty…
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 more fully captures the long, empty aisle with a strong central vanishing point, consistent depth scaling, and warm overhead lighting. Model B is crisp and symmetrical, but the table at the far end slightly undermines the empty-aisle request. (Second judge pass, order swapped — scores are the average of both: Model B delivers a clean, centered one-point perspective with orderly shelves, clear depth scaling, and warm light. Model A is also convincing and more cinematic, but its wide-angle | Anthropic: Claude Fable 5.1: Model A delivers a convincing photorealistic on…
Midnight Pantry Negotiation
A cinematic 16:9 still-life scene on a narrow apartment kitchen counter at 2:13 a.m., where seven household objects are arranged as if in a tense negotiation: a dented turquoise kettle facing a striped ceramic sugar jar, a half-open tin of cardamom tea, a glass honey dipper laid across a chipped saucer, a folded grocery receipt, a pair of red oven mitts slumped like silent witnesses, and a tiny battery radio with one glowing dial; compose them in one coherent, well-balanced tableau with believable spatial interaction, subtle reflections on laminate, and moody cyberpunk neon spill from a magenta window sign mixing with a cool under-cabinet light.
OpenAI: GPT-6 Luna: Model B presents all seven requested objects clearly, with especially legible cardamom tea, honey dipper on the saucer, folded receipt, and glowing radio dial in a cohesive still life. Model A has stronger neon atmosphere, but Model B’s object details and arrangement are more precise and convincing. (Second judge pass, order swapped — scores are the average of both: Model B more faithfully includes the dented kettle, clearly paired oven mitts, honey dipper across a chipped saucer, and subtle cou | Anthropic: Claude Fable 5.1: Model A delivers the stronger cinematic mood wi…
Matchup powered by OpenRouter.