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 varied image-generation tasks, from precise spatial layouts to stylized scenes and portraits. The comparison turns on how each handles composition, visual style, and specific prompt details.
The aggregate scores are **59.8 for Wan** and **61.1 for Ideogram V4.5 Edit**, but the saved result is a tie on this task sample. Wan won 3 tasks, Ideogram won 4, and 1 was tied. The editorial strength is **Too close to call (not statistical)**: Ideogram’s small aggregate edge does not overturn the saved tie. Ideogram’s clearest wins came in scenes where overall composition and style had to work together. Its gouache reunion better balanced the characters, lanterns, mountain mist, and signage; its dune poster had a more cohesive travel-poster look and a stronger sense of motion; and its funicular scene made the descent through the hillside town read more clearly. It also placed the bedroom desk directly beneath the window, where Wan missed that alignment. Wan was more dependable on exact visual constraints. It handled the object relationships in the attribute-binding task more convincingly, kept the library aisle empty and in one-point perspective, and better conveyed stunned relief in the salt-flat portrait. The restricted-palette campsite was a tie: Wan offered fuller detail, while Ideogram’s composition was cleaner. **Final call: a tie, with Ideogram’s narrow aggregate lead offset by Wan’s stronger precision on several constraint-heavy prompts.** This assessment applies only to the saved task sample; it does not establish a broader capability advantage for either model.
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 more clearly and convincingly renders the requested object colors, materials, and spatial relationships, with the duck sitting atop the cube and the cylinder behind the other objects. Model B follows the arrangement too, but its washed-out, sparse presentation and softer material definition make it less effective overall. (Second judge pass, order swapped — scores are the average of both: Model B and Model A both correctly place the yellow duck on the green cube, with the red sphere to t | Anthropic: Claude Fable 5.1: Model A nails every binding: a truly matte gree…
Monsoon Platform Reunion
A richly composed scene at the remote Khera Pass railway platform during a summer monsoon: seven distinct travelers and vendors interacting under one long corrugated shelter—a botanist protecting specimen tubes, twin schoolgirls sharing a mango ice pop, a tea seller pouring from a brass kettle, a porter balancing three indigo trunks, a tourist unfolding a soaked paper map, and a station dog waiting hopefully—arranged in one coherent, balanced composition with puddle reflections, luggage, signage, and mountain mist tying the group together; painterly gouache travel-poster style, cool rainlight with warm lantern accents, 16:9
OpenAI: GPT-6 Luna: Model B more convincingly captures the requested painterly gouache style, with a balanced scene, mountain mist, lanterns, and clear Khera Pass signage; it also includes all the key characters and interactions. Model A is richly detailed and strongly adheres to the setup, but its more graphic rendering and misspelled station sign make it a little less successful overall. (Second judge pass, order swapped — scores are the average of both: Model B presents all requested characters and props in a co | Anthropic: Claude Fable 5.1: Model B delivers a genuinely painterly gouache…
Dune Board Cutback
A desert adventure scene showing a lone sandboarder carving a sharp cutback down the apricot dunes of Erg Nacera, one hand skimming a spray of glittering sand while a striped scarf whips backward and the board throws a crescent plume, capturing the exact instant of dynamic balance and speed; stylized contemporary travel poster art, long telephoto composition compressing layered dunes, low sun creating dramatic rim light and elongated shadows, vivid sky, 16:9
OpenAI: GPT-6 Luna: Model B better captures the contemporary travel-poster style, vivid low-sun palette, layered dunes, and flowing scarf and sand plume; Model A has a more prominent rider but the scarf appears detached and the foreground props distract from the scene. Model B’s sweeping composition also conveys the carve and speed more effectively. (Second judge pass, order swapped — scores are the average of both: Model B more fully captures the stylized travel-poster look, compressed layers of apricot dunes, and | Anthropic: Claude Fable 5.1: Model B delivers a cohesive stylized travel-pos…
Cliffside Funicular Geometry
A wide-angle architectural travel illustration of the cliffside town of Puerto Zafiro with a red funicular descending toward a tiny harbor 180 meters below, steep stair lanes, stacked pastel inns, handrails, balconies, laundry lines, and moored fishing skiffs all drawn at correct relative scale with a single coherent vanishing-point perspective so the hillside feels believable from foreground terrace café to distant breakwater; crisp vibrant flat-vector design, bright noon light, ultra-clean composition, 16:9
OpenAI: GPT-6 Luna: Model B makes the steep funicular route, stacked pastel hillside town, and harbor read more clearly as one dramatic descent, while Model A offers a charming café foreground and lively details but a less convincing hillside perspective. Model B’s large empty sky and extra funicular car are minor deviations from the prompt. (Second judge pass, order swapped — scores are the average of both: Model B more convincingly connects the steep hillside, funicular, terrace café, and harbor in a coherent, ri | Anthropic: Claude Fable 5.1: Model B depicts a true rail funicular descendin…
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 precisely delivers a clean, empty one-point-perspective aisle with consistent scale and warm overhead lighting. Model B is atmospheric, but the ladder and open books on the floor detract from the requested empty aisle, and its lighting is less consistently warm. (Second judge pass, order swapped — scores are the average of both: Model A more cleanly fulfills the empty, symmetrical one-point-perspective aisle, with consistent recession and warm overhead lighting. Model B is atmospher | Anthropic: Claude Fable 5.1: Model A delivers a clean, empty aisle with shel…
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 B places the desk directly beneath the back-wall window and gives the room a more balanced, polished composition; Model A is clean and readable but its desk is offset from the window. Both include the requested bed, centered round rug, and front-right floor lamp in an isometric style. (Second judge pass, order swapped — scores are the average of both: Model B places the desk directly beneath the back-wall window and clearly includes the bed along the left wall, centered round rug, and fron | Anthropic: Claude Fable 5.1: Model B places the bed on the left wall, a cent…
Salt-Flat Farewell
A close-up cinematic travel portrait of a weather-beaten ferry captain on the blinding white salt flats of Laguna Caracol, gripping a faded navy cap to his chest as he receives unexpected news, his face unmistakably showing stunned relief with wet eyes, a trembling half-smile, and tightened brow; hyperreal editorial photography style, wind-chapped skin and tiny salt crystals visible, distant mirage horizon softly blurred, low golden sunset sidelight sculpting every expression detail, shallow depth of field, 16:9
OpenAI: GPT-6 Luna: Model A more closely captures the requested stunned relief, with wet-looking eyes, a trembling half-smile, visible salt crystals, and warm sunset sidelight. Model B has a strong emotional portrait and salt-flat setting, but its expression reads more pained than relieved and the lighting is less golden. (Second judge pass, order swapped — scores are the average of both: Model A more clearly conveys the requested stunned relief through the wet-looking eyes and trembling half-smile, with especially | Anthropic: Claude Fable 5.1: Model A nails the hyperreal brief with visible…
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 uses the requested palette and dusk details well, but its campsite is crowded and some objects feel visually tangled. Model B is cleaner and more balanced, with a clear mountain-campsite scene, though its sparse details and faint extra-toned outlines make the restricted-palette adherence slightly less certain. (Second judge pass, order swapped — scores are the average of both: Model A presents a fuller, more balanced campsite scene with a cohesive four-color graphic style, while Model B | Anthropic: Claude Fable 5.1: Model A sticks rigorously to the four-color pale…
Matchup powered by OpenRouter.