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
Kling Image and OpenAI: GPT-5.4 Image 2 face a varied image-generation test, from exact object placement and legible text to perspective, palette, and emotional expression. The prompts also test whether each model respects requested framing and style—not just whether the main subject appears.
OpenAI: GPT-5.4 Image 2 leads the saved task sample with an aggregate score of **64.0**, against **51.5** for Kling Image. It won five tasks, Kling won two, and one was a tie. The difference came from several distinct strengths, not uniform wins across the brief types. Kling’s clearest successes were tasks where composition and framing were tightly specified. It delivered the requested 16:9 reading nook and bookstore aisle, while OpenAI’s square images missed that requirement. Kling also won the bittersweet-victory prompt: OpenAI supplied the chef’s face and tense, joyful expression, while Kling omitted the face and failed to make the pastry read as a croquembouche. OpenAI was more dependable on relationships between objects and on details that must read precisely. It placed the bakery items more faithfully, kept closer to the restricted palette, rendered all three poster lines correctly, and built the rooftop noodle-cart scene with its requested people, props, and downhill perspective. The tasting lineup was a tie: Kling better matched the flat-vector, 16:9 brief; OpenAI depicted some of the vessels more accurately but departed from the requested style and framing. **Verdict: OpenAI: GPT-5.4 Image 2 is the clear result on this saved task sample, with Kling’s framing advantage showing where OpenAI still misses explicit format constraints. This is an editorial assessment of the saved task sample, not a claim of statistical confidence or a verdict on every use case.**
Negation
A cozy reading nook with an armchair, a stack of books, and a mug of tea by a window — with absolutely NO plants, NO lamps, and NO artwork or picture frames anywhere in the frame. Warm afternoon light, 16:9.
OpenAI: GPT-6 Luna: Model A closely matches the requested 16:9 reading nook, with the armchair, books, tea, and warm window light while avoiding all prohibited items. Model B also includes the key objects and avoids plants, lamps, and artwork, but its square framing misses the requested aspect ratio and its book text appears less consistent. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested 16:9 format and presents all the requested elements while avoiding plants, lam | Anthropic: Claude Fable 5.1: Both Model A and Model B correctly omit plants,…
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 closely matches the requested 16:9 framing while showing a warm, empty aisle with convincing one-point perspective and clear size recession. Model B has especially strong symmetry and consistent scale, but its square framing misses the specified aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model B has exceptionally clean, centered one-point perspective and crisp detail, but its square framing misses the requested 16:9 format. Model A matches the forma | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame with…
Bakery counter position logic
Whimsical but precise 16:9 bakery-window scene in vibrant flat-vector design: on a striped café counter, place a tall pistachio layer cake exactly in the center; a tiny espresso cup to the left of the cake; a glass jar of biscotti to the right of the cake; a strawberry tart directly in front of the cake; a brass cake server behind the cake; a lemon macaron resting on top of the espresso cup; a folded mint-green receipt lying between the espresso cup and the strawberry tart; and a sugar cube positioned behind the jar of biscotti but still visible. Camera is straight-on at counter height so all relationships read clearly, with sunny morning light and neat graphic shadows.
OpenAI: GPT-6 Luna: Model B follows the requested object relationships more closely, especially the macaron on the cup and the receipt between the cup and tart, though it departs from the flat-vector 16:9 style. Model A better matches the specified style and aspect ratio, but several requested placements are incorrect. (Second judge pass, order swapped — scores are the average of both: Model B places nearly all requested items in their specified relationships, including the macaron on the cup and the receipt betwee | Anthropic: Claude Fable 5.1: Model B nails nearly all the positional logic (…
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 A matches the requested 16:9 format and clearly conveys dusk, but uses noticeable extra brown and peach tones and gradients. Model B has a cleaner restricted-palette vector style and a well-balanced campsite scene, though its square format misses the 16:9 requirement. (Second judge pass, order swapped — scores are the average of both: Model B is a polished, balanced campsite scene that largely stays within the four requested colors, though it misses the 16:9 aspect ratio. Model A matches t | Anthropic: Claude Fable 5.1: Model B nails the core restricted-palette requi…
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 en dash in the time, while Model A appears to use a dot in its place. Model A is especially crisp and spacious, but Model B’s centered layout and subtle vintage grain still suit the poster brief well. (Second judge pass, order swapped — scores are the average of both: Model B renders all three lines with the requested wording and punctuation, including the en dash, in a balanced centered layout on a subtly grained navy background. Model A | Anthropic: Claude Fable 5.1: Model A renders the middle line incorrectly as '…
Bittersweet victory grin
Painterly close-up portrait of a young pastry chef in a flour-dusted navy apron standing in a competition kitchen, holding a slightly lopsided croquembouche on a silver stand and conveying unmistakable relieved triumph after a near disaster: glossy eyes, raised cheeks, tense joyful smile, a faint exhale, and eyebrows that show disbelief turning into pride. Warm spotlight from above, blurred stainless-steel background, shallow depth of field, realistic skin texture, expressive hands, contemporary editorial food-magazine style, horizontal 16:9 composition.
OpenAI: GPT-6 Luna: Model B captures the chef’s glossy-eyed, tense joyful relief and pride alongside a recognizable croquembouche in a flour-dusted competition kitchen. Model A omits the chef’s face and emotional expression, and its pastry does not resemble a croquembouche, despite clear hands and a stand. (Second judge pass, order swapped — scores are the average of both: Model B captures the relieved, emotional triumph with a flour-dusted navy apron, expressive face and hands, and a visibly lopsided croquembouche | Anthropic: Claude Fable 5.1: Model A omits the face entirely—the core emotio…
Color-locked tasting lineup
Studio tabletop product illustration, vibrant flat-vector design, 16:9: arrange six distinct drink vessels in a clean row on a pale stone counter, each object with its own exact color and material that must stay perfectly bound to that object: a frosted cobalt-blue glass bottle, a matte saffron-yellow ceramic mug, a polished copper cocktail shaker, a translucent emerald-green gummy-bear-shaped cup, a milky white porcelain teapot, and a clear smoked-gray acrylic tumbler. Each vessel contains a different visible garnish or liquid level, but the specified color and material belong to the vessel itself only; no color bleeding, swapping, or shared finishes. Bright softbox lighting, crisp shadows, minimal modern packaging-shot composition.
OpenAI: GPT-6 Luna: Model A better matches the requested 16:9 flat-vector illustration and presents all six vessels in a clean, balanced row. Model B depicts the gummy-bear cup and tumbler more accurately, but its photorealistic style and square framing depart from the prompt. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested vibrant flat-vector style and 16:9 tabletop lineup, while Model B presents the objects as polished photorealism in a square frame. Model | Anthropic: Claude Fable 5.1: Model A honors the flat-vector style and 16:9 fr…
Rooftop noodle cart perspective
Cinematic 16:9 street-scene illustration of a tiny rooftop ramen cart on a steep hillside in Valparaíso at blue hour, viewed from eye level down a narrow lane with one clear vanishing point: in the foreground, a giant enamel bowl of miso ramen on a folding counter with chopsticks resting across it; midground, the vendor, a dented kettle, and two customers on stools; farther uphill, strings of lanterns, stacked crates of limes, and painted row houses receding consistently into the distance; in the far background, a funicular car and harbor lights. Relative scale must feel physically correct at every depth, with straight architectural lines converging naturally and no warped bowls, stools, or buildings. Rich cinematic lighting with warm lantern glow against cool evening shadows, highly detailed editorial travel-poster style.
OpenAI: GPT-6 Luna: Model B includes the vendor, two customers, kettle, limes, funicular, harbor lights, and a convincing downhill perspective in a richly lit scene. Model A better matches the requested 16:9 format, but its oversized bowl disrupts the physical scale and several requested cart details are less clear. (Second judge pass, order swapped — scores are the average of both: Model B brings the giant ramen bowl into the foreground and includes the vendor, customers, kettle, limes, funicular, and harbor withi | Anthropic: Claude Fable 5.1: Model B delivers nearly every requested element…
Matchup powered by OpenRouter.