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
Eight image prompts put composition, object relationships, style, and fine visual detail to the test. The matchup also exposes a recurring tradeoff: satisfying the requested frame versus making the scene more polished or legible.
OpenAI: GPT-5.4 Image 2 is the saved winner on this task sample, with an aggregate score of **65.1** to Kling Image’s **53.8** and five task wins to three. The editorial strength is a **slight edge**, not a statistical claim—and the scores don’t erase the areas where Kling did better. Kling won hands & anatomy, reflections & glass, and attribute binding. In those prompts, it more reliably honored the requested 16:9 framing and made key objects or actions easy to read. OpenAI’s square crops cost it points even when its lighting or reflections were striking. OpenAI took spatial layout, macro thimble and thread, the Bauhaus spice tin poster, greenhouse jar reflections, and eleven pantry pears. Its advantages were often specificity and presentation: a more cohesive isometric room, a better-rendered macro scene, a more convincing poster, clearer mirror reflections, and pears that were easier to count. The wins weren’t spotless: it missed the requested frame in several cases, and the pear-count judgment was not uniform across the notes. **Verdict: OpenAI: GPT-5.4 Image 2 gets the slight edge on this saved task sample, chiefly by making several complex scenes and details more legible. Kling Image is the better fit in this set when requested framing and straightforward spatial fidelity matter most; this assessment applies only to the saved task sample.**
Spatial layout
A clean isometric illustration of a bedroom: a bed against the LEFT wall, a round rug centered on the floor, a desk under the WINDOW on the back wall, and a floor lamp in the FRONT-RIGHT corner. Flat-vector style, consistent perspective.
OpenAI: GPT-6 Luna: Model A follows the requested flat-vector style closely, while Model B presents the same key furniture placements in a more polished, cohesive isometric composition. Model B is more detailed and visually appealing, though its realistic shading and textures are less faithful to the flat-vector request. (Second judge pass, order swapped — scores are the average of both: Model B places all requested furniture clearly and creates a polished, cohesive isometric bedroom with the bed against the left w | Anthropic: Claude Fable 5.1: Model B places every item correctly (bed headbo…
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 better matches the requested 16:9 candid composition, soft daylight, and shallow depth of field, with both hands clearly visible and a plausible grip. Model B shows the bracelet-making action more clearly, but its square, tightly cropped framing is less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B shows the bracelet-making action more clearly, but its square crop misses the requested 16:9 framing. Model A has a more polished wide com | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame, soft…
Macro thimble and thread
An extreme macro close-up of a brass sewing thimble lying on a walnut tabletop beside a threaded needle and a loose coil of teal thread, shot with a true macro lens at very shallow depth of field so the engraved dimples on the thimble are razor sharp while the thread falls into creamy bokeh; low side light from a nearby window reveals tiny scratches, lint fibers, and warm wood grain, cinematic 16:9 framing.
OpenAI: GPT-6 Luna: Model B better captures the requested lying brass thimble, clearly threaded needle, and detailed teal fibers in a true macro look, though it misses the 16:9 framing. Model A has appealing warm light and a strong wide composition, but the thimble stands upright and the needle does not appear threaded. (Second judge pass, order swapped — scores are the average of both: Model B better captures the extreme macro view, sideways thimble, crisp dimples, and teal thread falling into bokeh, though its sq | Anthropic: Claude Fable 5.1: Model A correctly delivers 16:9 framing, sharp…
Reflections & glass
A photorealistic close-up of a chrome teapot on a polished dark marble counter next to a glass of water; the window and a red apple on the counter must be correctly reflected in BOTH the chrome and the water's surface, dramatic side light, 16:9.
OpenAI: GPT-6 Luna: Model A better matches the requested close-up and 16:9 framing, with the teapot, water glass, apple, and window all clearly present. Model B has especially striking reflections and dramatic light, but its square framing is less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 framing and more clearly shows the teapot, water glass, apple, and their reflections. Model B has more dramatic side lighting, but its squ | Anthropic: Claude Fable 5.1: Model A delivers the required 16:9 frame with t…
Bauhaus spice tin poster
A Bauhaus-style still life poster of a kitchen shelf featuring a cobalt moka pot, three labeled spice tins reading "sumac," "caraway," and "fenugreek," a striped tea towel, and a sliced blood orange, rendered with strict geometric simplification, bold primary shapes, asymmetrical typography blocks, clean black outlines, and flat cream-red-blue-yellow palette; raking morning light casts crisp diagonal shadows across the composition, 16:9.
OpenAI: GPT-6 Luna: Model B has stronger Bauhaus poster design, more deliberate geometric composition, and clearer asymmetrical typography while including all requested objects and legible labels. Model A matches the requested widescreen format, but its muted background shelf and less cohesive poster treatment make it feel less aligned with the style. (Second judge pass, order swapped — scores are the average of both: Model B feels more like a designed Bauhaus poster, with strong asymmetrical typography, bold geome | Anthropic: Claude Fable 5.1: Model B delivers a true Bauhaus poster with cor…
Attribute binding
A photorealistic tabletop scene: a matte green cube to the LEFT of a glossy red sphere, a small blue cylinder BEHIND both, and a single yellow rubber duck sitting ON TOP of the cube. Colors must stay on the correct objects. Soft studio light, 16:9.
OpenAI: GPT-6 Luna: Model A follows the requested left/right placement, puts the duck on the cube, and clearly shows the cylinder behind the objects in the requested 16:9 tabletop scene. Model B is polished and also gets the object relationships right, but its square framing misses the specified aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model A follows all object, color, and spatial relationships while matching the requested 16:9 format; Model B also binds the objects correct | Anthropic: Claude Fable 5.1: Both Model A and Model B bind all colors correc…
Greenhouse jar reflections
A sunlit still life on a kitchen windowsill with a tall ribbed glass jar half-filled with water, a chrome egg cup, a small round hand mirror, and a sprig of rosemary leaning against a fogged pane; the scene must show accurate reflections in the chrome and mirror, realistic refraction and magnification through the jar water and ribbed glass, faint double reflections in the window, and soft caustic light patterns on the painted sill, late-afternoon golden light, 16:9.
OpenAI: GPT-6 Luna: Model B more clearly shows the mirror reflecting the scene, convincing reflections in the chrome, and bright caustic patterns on the sill, while retaining the requested objects and warm window light. Model A better matches the 16:9 format, but its mirror reflection and caustic effects are less distinct. (Second judge pass, order swapped — scores are the average of both: Model B has especially appealing golden light, clear caustics, and convincing mirror reflections, but its square framing misses | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame, a ge…
Eleven pantry pears
A neatly arranged pantry-shelf still life showing exactly 11 distinct Seckel pears, each fully visible and individually countable, placed among a blue enamel colander, a folded receipt, a jar of star anise, and a ceramic salt cellar; warm tungsten under-cabinet lighting, straight-on composition with no overlap that hides any pear, natural textures and subtle shadows, vintage travel poster color mood, 16:9.
OpenAI: GPT-6 Luna: Model B presents the pears in a clearer, individually countable arrangement and includes the requested pantry props, though it appears to have about ten pears rather than exactly eleven and uses a square format. Model A better matches the 16:9 shelf scene and folded receipt, but several pears overlap or crowd together, making the requested count difficult to verify. (Second judge pass, order swapped — scores are the average of both: Model B clearly presents exactly 11 individually visible pears | Anthropic: Claude Fable 5.1: Model B delivers exactly 11 clearly separated, c…
Matchup powered by OpenRouter.