Head to head: Wan vs Ideogram V4.5 Edit
Wan vs Ideogram V4.5 Edit
By Ryan Merket · Published
Wan and Ideogram V4.5 Edit are tested across image prompts that demand style control, precise composition, color limits, and careful handling of exclusions. The comparison focuses on how each model meets the details of the brief, not just on visual polish.
The saved aggregate scores are **58.5 for Wan** and **63.9 for Ideogram V4.5 Edit**, but the saved result is a tie on this task sample: Wan wins two tasks, Ideogram wins three, and three are ties. The score gap does not change that verdict. Ideogram’s clearest wins come on prompts where scene readability or constraint-following matters: its ukiyo-e fishing boat is more legible, its cliffside funicular scene better captures the requested interactions, and it keeps the reading nook clear of forbidden plants, lamps, and artwork. Wan stands out on the restricted-palette campsite and color-locked market still life, where it better respects the palette and the saffron wallet brief. The tied tasks show why neither model pulls away. Wan is stronger on some specifics in the ferry-deck and library-perspective scenes, while Ideogram brings stronger atmosphere or polish; the breakfast-light judges split on which image better fits the brief. Those are meaningful trade-offs, not a consistent advantage for either model. **Final call: Too close to call (not statistical).** This assessment applies only to the saved task sample; the aggregate scores are higher for Ideogram, but the saved head-to-head verdict remains a tie.
Windblown relief on ferry deck
Photorealistic travel portrait of a young woman on the open deck of the midnight ferry to Nólsoy, gripping a salt-speckled yellow raincoat collar and breaking into unmistakable relieved laughter after a storm passes; wet strands of hair stuck to her forehead, eyes half-squinting with emotional release, cheeks flushed from cold wind, distant harbor lights behind her, spray on the railing, cinematic overcast-to-clear transition lighting with a pale gap in the clouds catching her face, shallow depth of field, 16:9
OpenAI: GPT-6 Luna: Both images capture the wet-haired woman laughing in a yellow coat on a ferry deck, but Model A more clearly conveys the requested half-squinting relief and flushed cheeks. Model B has the stronger cinematic composition and moody harbor atmosphere, with the clearing in the clouds and shallow-focus lights adding polish. (Second judge pass, order swapped — scores are the average of both: Model A more clearly captures the windblown, rain-soaked release, with wet strands across the forehead, flushed | Anthropic: Claude Fable 5.1: Model A nails nearly every prompt beat—squintin…
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 clearly depicts a fishing boat riding a dramatic cresting wave, with strong ukiyo-e outlines, stylized foam, and a restrained indigo-and-cream palette. Model A has a convincing print-like treatment and dawn coloring, but the boat is partly obscured and less legible in the composition. (Second judge pass, order swapped — scores are the average of both: Model B more clearly depicts a fishing boat riding a single dramatic cresting wave, with a muted indigo-and-cream palette and convincing u | Anthropic: Claude Fable 5.1: Model B presents a single, clearly readable fis…
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 clearly delivers the precise centered one-point perspective, with orderly shelves and books that consistently diminish toward the vanishing point. Model B has appealing atmospheric warmth, but its more irregular shelf and book scale is a weaker match for the prompt’s explicit scale requirement. (Second judge pass, order swapped — scores are the average of both: Model A delivers the clearest centered one-point perspective, with the shelves and overhead lights converging cleanly and t | Anthropic: Claude Fable 5.1: Model A delivers a very clean, textbook one-poi…
Cliffside funicular reunion scene
Detailed watercolor illustration of a narrow cliffside station above the Bay of Kotor where five travelers and two local porters converge around a tiny red funicular car: one elderly man unfolding a hand-drawn map, a child pointing at a gull stealing a pastry, a porter balancing stacked teal suitcases, a woman kneeling to retie a bootlace, and another traveler leaning over the railing to photograph the boats below; layered terraces, laundry lines, cypress trees, and stone walls create depth, balanced triangular composition, lively but coherent interactions, bright post-rain Mediterranean light, 16:9
OpenAI: GPT-6 Luna: Model B more convincingly captures the tiny red funicular, stacked teal luggage, and the requested travelers’ interactions within a layered, atmospheric cliffside setting. Model A includes most of the same details and has a lively composition, but its larger car and more cartoon-like rendering feel less faithful to the watercolor prompt. (Second judge pass, order swapped — scores are the average of both: Model B follows the requested watercolor scene especially closely, clearly depicting the map | Anthropic: Claude Fable 5.1: Model B delivers a true cliffside funicular on…
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 B fulfills the requested nook while keeping plants, lamps, and artwork out of view. Model A is cozy and polished, but visible greenery outside the window and a cropped frame at the upper right violate the explicit exclusions. (Second judge pass, order swapped — scores are the average of both: Model B includes the requested reading nook elements in warm afternoon light and avoids visible plants, lamps, and artwork. Model A is cozy and well composed, but foliage is visible outside the window | Anthropic: Claude Fable 5.1: Model A breaks the negation constraints with a…
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 more closely follows the four-color restriction while depicting a fuller, unmistakable campsite, though its composition is busier and less strictly flat. Model B is cleaner and well balanced, but the white moon and stars and extra gray tones stray from the specified palette. (Second judge pass, order swapped — scores are the average of both: Model A depicts a more complete, engaging campsite with stronger visual detail, though its starry sky feels more like night than dusk and the subtle | Anthropic: Claude Fable 5.1: Model A stays tightly within the four-color pal…
Color-locked market still life
Studio-lit travel-themed still life in a crisp contemporary product-photography style: on a pale limestone slab sit exactly six souvenir objects from an imaginary desert rail journey, each with its own fixed color and material — a cobalt blue glazed ceramic espresso cup, a matte saffron-yellow leather passport wallet, a polished emerald-green glass compass, a brushed copper luggage tag, a chalk-white carved bone camel charm, and a deep crimson woven wool ticket pouch — arranged with clean spacing so every object is fully visible, neutral background, sharp shadows, strict attribute and color separation with no color bleeding or swaps
OpenAI: GPT-6 Luna: Model A more faithfully preserves the requested saffron wallet color and presents all six objects clearly on a distinct limestone slab, with strong color separation. Model B is also polished, but its wallet reads orange and the objects sit directly on a broad stone surface rather than a slab. (Second judge pass, order swapped — scores are the average of both: Model A matches the specified saffron-yellow wallet more closely and presents all six objects clearly on a distinct limestone slab. Model | Anthropic: Claude Fable 5.1: Model A nails the color lock — true cobalt cup,…
Mountain inn breakfast light study
Ultra-photorealistic interior of a small alpine guesthouse breakfast room in Val Müstair just after sunrise, viewed at table height: a brushed steel moka pot, a scarred walnut tabletop, a linen napkin, a half-peeled blood orange, a wool sweater draped over a chair, and the innkeeper’s hand reaching in to set down a ceramic cup; low golden sunlight from a frosted window creates believable soft shadows, subtle reflections, translucent citrus flesh, convincing skin texture, fabric weave, wood grain, and metal highlights, natural color grading, 16:9
OpenAI: GPT-6 Luna: Model A more clearly captures the requested warm, just-after-sunrise breakfast scene, with all key objects readable and the hand visibly setting down the cup. Model B has appealing, natural textures, but its cooler, moodier light and more crowded framing are less aligned with the prompt. (Second judge pass, order swapped — scores are the average of both: Model B better captures the requested intimate, just-after-sunrise atmosphere, with convincing steam, a blood-orange-like peeled citrus, rich t | Anthropic: Claude Fable 5.1: Model A hits nearly every prompt element cleanl…
Matchup powered by OpenRouter.