Head to head: AuraFlow vs ImagineArt 1.5 Pro Preview
AuraFlow vs ImagineArt 1.5 Pro Preview
This matchup wasn’t subtle: one model consistently handled prompt discipline better across the board, while the other mostly coasted on surface appeal. The interesting part is where the loser still looked good enough to make some calls feel closer than the score suggests.
AuraFlow came into this one with some visual charm, but charm is not the same thing as control. On the aggregate, **ImagineArt 1.5 Pro Preview wins decisively, 68.2 to 43.8, with 99% confidence**. The task ledger is even harsher: **ImagineArt takes 6 wins, AuraFlow takes 0, with 2 ties**. That is not a stylistic preference result; it is a reliability result. The pattern is clear. ImagineArt was better at the boring, crucial stuff that separates a pretty image generator from a dependable one: **attribute binding, negation, counting, exact text, and style fidelity**. It put the duck where it belonged, respected the "no plants, no lamps, no artwork" reading-nook constraint far better, got much closer to exactly nine blue pots, and crushed the poster task where AuraFlow fell into the usual text-gibberish trap. It also delivered the stronger ukiyo-e interpretation, with period-appropriate forms and a more faithful palette instead of drifting into generic poster energy. Even in the two ties, ImagineArt often sounded like the more convincing image-maker. In **Reflections & glass**, it better sold the chrome teapot, side light, and reflected details. In **Wheelbarrow Sprint**, it more convincingly captured motion, puddle splash, storm light, and the seedling detail AuraFlow missed. Those were scored as ties, but they read like warnings for AuraFlow: when the prompt asks for physical plausibility or specific scene evidence, it too often settles for a polished approximation. AuraFlow’s best defense is that it can look good. Several notes describe it as attractive, polished, or energetic. But too often that polish covered prompt misses: the wrong object relationships, broken negation, bad counting, weaker action depiction, and text errors that are fatal in any design-oriented workflow. **ImagineArt 1.5 Pro Preview is the better tool because it follows instructions more consistently, and in this test it did so by a wide margin.**
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.
Model B better matches the prompt with a photorealistic close-up, dramatic side light, and more convincing reflections of the window and apples in the chrome teapot and water glass. Model A is clean and attractive, but the glass appears underfilled/empty and the required reflected apple/window details in both reflective surfaces are less clearly and correctly rendered. (Order-swapped judge pass: Model B better matches the prompt with a more convincing chrome teapot, polished dark marble, dramatic side light, and clearer reflections of the window and apple in the chrome and surrounding surfaces. Model A is attractive and photorealistic, but the water glass reflection behavior is less convincing and the scene feels less controlled compositionally.)
Wheelbarrow Sprint
A candid action photograph of a young horticulturist sprinting across a rain-darkened community garden path, one boot splashing through a puddle as a red wheelbarrow full of marigold seedlings bucks sideways and loose potting soil arcs through the air; capture the subject mid-stride with convincing motion, flying jacket hem, bent knees, and a tense grip on the handles, late-afternoon storm light breaking through clouds, shallow depth of field, dramatic diagonal composition, cinematic realism, 16:9
Model B better matches the prompt with a more convincing candid action photograph, visible puddle splash, storm light breaking through clouds, and a stronger sense of wheelbarrow motion. Model A is striking and polished, but it feels more stylized, misses the seedling detail, and the pose/wheelbarrow interaction looks less physically convincing. (Order-swapped judge pass: Model B matches the prompt more completely with visible marigold seedlings, stronger sprinting motion, dramatic diagonal energy, and a more cinematic rain-storm garden setting. Model A is realistic and well lit, but it misses the seedlings and feels less dynamic and less shallow in depth of field.)
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.
Model B follows the requested spatial relationships much better: the matte green cube is left of the glossy red sphere, the blue cylinder is behind, and the yellow duck is on top of the cube. Model A has stronger polish, but the duck is incorrectly on the sphere rather than on the cube, which is a major attribute-binding failure. (Order-swapped judge pass: Model A follows the attribute bindings much better: the matte green cube is left of the glossy red sphere, the blue cylinder is behind, and the yellow duck is on top of the cube. Model B is cleaner and more stylized, but it breaks the key relation by placing the duck on the red sphere instead of on the cube.)
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.
Model B closely matches the requested cozy reading nook and correctly avoids plants, lamps, and visible artwork or picture frames, while Model A clearly violates the negation constraints with multiple plants and a framed picture. Both are aesthetically pleasing, but B adheres far better to the prompt. (Order-swapped judge pass: Model A closely matches the requested cozy reading nook and correctly avoids plants, lamps, and artwork, though it includes an extra mug and feels slightly cluttered. Model B violates the negation constraints with multiple plants and a framed picture, despite having pleasant lighting and a clean composition.)
Nine Blue Pots
A crisp studio-style product photograph of a moss-green metal shelf in a potting shed holding EXACTLY nine distinct cobalt-blue ceramic plant pots, each separated and fully visible for easy counting, with a different small herb or succulent in every pot; no extra pots anywhere in the scene, straight-on eye-level composition, clean neutral plaster wall behind, bright diffused daylight, high detail, 16:9
Model B is much closer to the requirement of exactly nine blue pots on a green shelf, while Model A shows far too many pots and a different shelf setup. B still misses the straight-on studio product-photo feel and uses stronger directional light, but overall adheres far better to the prompt. (Order-swapped judge pass: Model A is much closer to the prompt, showing a moss-green shelf with exactly nine cobalt-blue pots and a clean studio-like product-photo feel, though the lighting is harsher than requested and some pots are not perfectly separated for easy counting. Model B fails the core count requirement with far more than nine pots and has a less product-focused composition despite a straight-on view.)
Seed Saver Portrait
A detailed oil painting portrait of an elderly seed saver seated at a wooden potting bench in a glass greenhouse, shown from mid-thigh up with both hands fully visible as she delicately opens a dried lunaria seed pod over a ceramic tray; emphasize anatomically correct hands with clearly formed fingers and thumbs, natural shoulders and arm proportions, calm upright posture, expressive face, weathered denim apron, soft morning light filtered through greenhouse panes, richly observed botanical surroundings, three-quarter view
Model B better matches the prompt with a clear three-quarter view, visible hands actively opening a dried seed pod at a wooden potting bench in a richly detailed greenhouse. Model A is appealing but more front-facing, the tray rests on her lap rather than a potting bench, and the seed-pod action is less clearly depicted. (Order-swapped judge pass: Model A better matches the prompt with a three-quarter mid-thigh portrait in a greenhouse, both hands visible opening a dried seed pod over a ceramic tray, and more natural anatomy. Model B is appealing but deviates from the specified action and view, showing a frontal pose with a different seed-handling setup and less convincing hand anatomy.)
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.
Model B is much more faithful to the ukiyo-e woodblock tradition, with flatter color planes, more period-appropriate boats, and a cohesive muted indigo/cream palette. Model A has strong energy but breaks adherence with modern-looking fishing boats and a more poster-like, less authentic treatment. (Order-swapped judge pass: Model A is much more faithful to the ukiyo-e woodblock print prompt, with flat color planes, bold outlines, stylized foam, and a muted indigo/cream dawn palette. Model B has strong graphic composition, but the modern motorboats and brighter palette break prompt adherence and period/style fidelity.)
Rare Fern Fair Poster
A graphic design close-up of a cream paper poster pinned to a corkboard at a botanical society hall, with elegant but highly legible dark green typography that clearly reads: "RARE FERN FAIR" on the top line and "JUNE 14" below it; the text must be spelled exactly as written and easy to read, surrounded by tasteful illustrated fiddlehead fern motifs, art nouveau inspired layout, soft indoor tungsten lighting, sharp focus, minimal visual clutter
Model B closely matches the prompt with clearly readable exact text, elegant dark green typography, tasteful fern motifs, and an art nouveau poster pinned in warm indoor light. Model A has appealing styling but fails the exact text requirement and includes multiple misspellings and extraneous gibberish, which significantly hurts adherence. (Order-swapped judge pass: Model A closely matches the prompt with correct, highly legible text, cream paper, dark green art nouveau typography, fern motifs, and a pinned poster under warm indoor lighting. Model B has major text rendering errors ("RAE FERNN FAIR" and gibberish filler text), weaker adherence to the corkboard setting, and a less convincing poster composition.)
Matchup powered by OpenRouter.