Head to head: Kling Image vs OpenAI: GPT-5.4 Image 2
Kling Image vs OpenAI: GPT-5.4 Image 2
By Ryan Merket · Published
We compare two image models across eight demanding prompts, from macro framing and hands to exact text, object counts, and crowded scenes. The results show where each model follows the brief—and where it lets details slip.
OpenAI: GPT-5.4 Image 2 wins decisively: **65.2 to 53.8**, taking seven of eight tasks with a tentative lean verdict. Kling Image’s only win came in camera and lens framing, where its widescreen macro shot better matched the requested aspect ratio and kept the wood-and-fingertip detail in focus. OpenAI’s edge is strongest when the prompt has a concrete checklist. It rendered all three poster lines correctly, showed the requested seven cups and 13 harmonicas, and made the friendship-bracelet action and foreground-to-background scale cues clearer. Those are meaningful wins in the kinds of image tasks where a polished look is not enough if the requested details are missing. It also took the ukiyo-e style and multi-subject scene comparisons, though not without tradeoffs. Across several prompts, its square framing missed a requested widescreen format; the harmonicas also skewed gold rather than silver. Kling, meanwhile, sometimes had the more faithful framing or atmosphere, but fell short on specifics such as the snare and the apparent harmonica count. **The call: OpenAI: GPT-5.4 Image 2 is the stronger all-around prompt follower in this matchup. Kling Image remains capable of a better-composed individual shot, but the seven-to-one task record—and the limited confidence—make this more than a narrow stylistic preference.**
Camera & lens framing
A macro close-up photograph of a luthier adjusting the turquoise enamel tuning peg on a hand-carved mandolin, with razor-sharp detail on the wood grain, fingertips, and tiny brass screw while the background dissolves into creamy bokeh from a true macro lens, cinematic window light from the left, intimate artisanal mood, 16:9.
OpenAI: GPT-6 Luna: Model A delivers the requested 16:9 macro framing, crisp wood and fingertip detail, and cinematic left-side light with creamy bokeh. Model B makes the adjustment to a turquoise peg more explicit, but its square framing misses the requested aspect ratio and feels less tightly focused on the action. (Second judge pass, order swapped — scores are the average of both: Model B more clearly shows the fingertips adjusting the turquoise enamel peg and includes the carved instrument detail and tiny brass | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 true-macro…
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 ukiyo-e tradition with a dramatic cresting wave, stylized foam, and a clear dawn palette; Model A has strong woodblock-like linework but lacks the dawn atmosphere. Model B’s composition is more dynamic and evocative, while both images are technically polished. (Second judge pass, order swapped — scores are the average of both: Model B strongly matches the ukiyo-e tradition, with a dramatic cresting wave, fishermen in the boat, a dawn sky, and the muted indigo-and- | Anthropic: Claude Fable 5.1: Model B delivers a towering cresting wave, a cr…
Perspective & scale
A realistic digital painting of a tram-sized outdoor concert stage being assembled in a narrow harbor alley, viewed from street level with strong one-point perspective leading to a single vanishing point between stacked speaker cases; in the foreground a coiled power cable, road cases, and a snare drum are correctly smaller than a midground upright bass on a dolly, while far in the distance two stagehands and a neon sign shrink naturally with depth, crisp dusk light, no warped lines, 16:9.
OpenAI: GPT-6 Luna: Model B more clearly depicts the requested scale hierarchy, with the cable and snare in front of a prominent upright bass, plus distant stagehands and a readable harbor-stage sign. Model A has stronger, cleaner one-point perspective and the requested widescreen framing, but its snare is absent and the key foreground-to-midground scale cues are less explicit. (Second judge pass, order swapped — scores are the average of both: Model B more directly matches the requested foreground-to-midground siz | Anthropic: Claude Fable 5.1: Model B nails the specified foreground trio (co…
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 depicts the friendship-bracelet-making process, with both hands in a plausible grip on the threads and a finished section visible. Model A better matches the requested 16:9 framing and has lovely soft light, but the bracelet-making action is less clear. (Second judge pass, order swapped — scores are the average of both: Model B more clearly shows a recognizable friendship bracelet in progress, with both hands using a plausible grip on the colorful threads. Model A has a stro | Anthropic: Claude Fable 5.1: Model A honors the 16:9 ratio, soft daylight, a…
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 correctly, including the requested 6–11pm range, with strong centered hierarchy and a cohesive minimalist look. Model A is also polished, but its middle line uses a dot instead of the requested dash and adds an unrequested divider. (Second judge pass, order swapped — scores are the average of both: Model B renders all three requested lines correctly and keeps the centered, minimalist navy-poster treatment. Model A is polished and well composed, but uses a dot inst | Anthropic: Claude Fable 5.1: Model B renders all three lines exactly as spec…
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 clearly countable cups in a straight row on white marble and matches the requested 16:9 format, though the contents vary and the shadows are somewhat strong. Model B has softer lighting and consistent cups, but its square format and extensive empty space make the composition less aligned with the prompt. (Second judge pass, order swapped — scores are the average of both: Model B shows exactly seven clearly separated, identical cups in a softly lit straight row, but it | Anthropic: Claude Fable 5.1: Both show exactly seven clearly separate, count…
Object counting
A watercolor illustration of a backstage rehearsal table in a tiny folk-music venue, showing exactly 13 distinct silver harmonicas laid out in a loose arc on a burgundy velvet cloth, every harmonica fully visible and separately countable, with a penciled set list reading 'Juniper Wake' nearby, soft amber dressing-room bulb light, no extra instruments, no duplicates hidden behind others, 16:9.
OpenAI: GPT-6 Luna: Model B presents exactly 13 clearly separated harmonicas in a deliberate loose arc, with the requested title legible and a convincing backstage atmosphere; its main deviations are the square format and gold-toned instruments. Model A has a stronger silver appearance and the requested wide format, but appears to show only 12 harmonicas and arranges them less clearly. (Second judge pass, order swapped — scores are the average of both: Model B shows exactly 13 separately visible harmonicas in a loo | Anthropic: Claude Fable 5.1: Model B renders 'Juniper Wake' correctly on a c…
Multi-subject scene composition
A lively watercolor illustration of an underground electro-swing jam inside a tiled midnight-blue subway platform: a trumpet player balancing on a luggage trunk, a tap dancer mid-step on the yellow safety line, a cellist seated on an overturned milk crate, a drummer using wire brushes on a compact kit, and two fascinated commuters filming on an old camcorder and a flip phone, all interacting in one coherent, well-balanced composition under cool fluorescent lights with a warm train arriving behind them, 16:9.
OpenAI: GPT-6 Luna: Model A presents the requested watercolor illustration with a lively, clear arrangement and all the major performers and filming commuters easy to identify. Model B captures the midnight-blue subway atmosphere and electro-swing setting especially well, but its darker, more crowded staging and modern-looking phone are a little less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested 16:9 watercolor-illustration style more close | Anthropic: Claude Fable 5.1: Model B renders every requested subject distinc…
Matchup powered by OpenRouter.