Head to head: Ideogram V4.5 Edit vs Wan
Ideogram V4.5 Edit vs Wan
By Ryan Merket · Published
This matchup tests a familiar image-model tradeoff: precise layouts, object attributes and counting versus atmosphere and style fidelity. Across eight prompts, the differences show up in scenes ranging from a mirrored greenhouse to ukiyo-e storms.
Wan leads the saved aggregate, 65.1 to 61.0, and takes five task wins to Ideogram V4.5 Edit’s three, with no ties. The verdict is a **slight edge (not statistical)**—a close sample, not a runaway result. Wan’s strongest work is in controlled structure: it produced the cleaner centered aisle, clearer step-pyramids, and more consistent object colors and materials. It also handled the exact-count cups with a straighter, more evenly spaced row. Those wins point to an advantage on these prompts when clarity and specification-following matter most. Ideogram had the more persuasive results where mood and style carried the brief. Its greenhouse scene better integrated the mirror, diver and reflections into a misty cinematic setting, while its fishing-boat and storm-leviathan images were judged more faithful to the requested ukiyo-e scenes. Wan’s versions could be striking, but added details or cleaner graphic treatment sometimes pulled them away from the prompt’s central image. **Final call: Wan, by a slight edge on the saved task sample.** This assessment applies only to that saved task sample; it does not establish a general capability ranking.
Greenhouse mirror canal
Inside the abandoned tide-greenhouse of moon Nacre Delta, show a narrow canal running between chrome support arches and giant transparent seed-pods, with a waist-high antique mirror leaning against one arch and a glass skiff floating in the water; the scene must accurately render reflections in the mirror, rippling reflections on the canal, refraction through the seed-pods and skiff hull, and chrome highlights that correctly mirror nearby plants and lights; include a bioluminescent diver in a rust-orange suit passing through the scene so their figure appears consistently in all reflective and transparent surfaces, blue hour lighting with mist, precise optical realism, widescreen composition, 16:9.
OpenAI: GPT-6 Luna: Model A more fully captures the misty, cinematic greenhouse with chrome arches, transparent seed-pods, a leaning antique mirror, skiff, and diver, while integrating strong reflections across the scene. Model B has a clear composition and striking canal reflections, but feels less atmospheric and its mirror and diver are less convincingly integrated with the requested optical realism. (Second judge pass, order swapped — scores are the average of both: Model A more fully conveys the misty, biolumi | Anthropic: Claude Fable 5.1: Model A delivers a coherent tunnel of chrome ar…
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 B delivers a cleaner, more centered one-point perspective with an empty aisle and consistent-looking scale recession. Model A has appealing warm lighting, but the table, open book, and ladder interrupt the empty aisle and make the composition less focused. (Second judge pass, order swapped — scores are the average of both: Model B more precisely delivers the empty, centered one-point aisle, with clear consistent recession and warm overhead lighting. Model A is atmospheric and detailed, but | Anthropic: Claude Fable 5.1: Model B delivers a textbook, perfectly symmetri…
Hilda chart palette
Create a vintage travel poster for the desert moon resort 'Hilda Terminus' showing a ring-shadow crossing monumental step-pyramids and a tiny sand-skimming funicular descending toward a salt lake, using only the named Hilda chart palette: #1D3557 deep navy, #457B9D steel blue, #A8DADC pale cyan, #F1FAEE ivory, and #E63946 vermilion, with no other hues anywhere in the image; crisp poster shapes, simplified shadows, elegant retro typography area left blank, strong sunlit geometry, high graphic clarity, 16:9.
OpenAI: GPT-6 Luna: Model B more clearly depicts monumental step-pyramids and uses crisp, controlled poster geometry with a strong retro layout. Model A better suggests a ring crossing the peaks and a funicular descending toward the lake, but its pyramids read more like mountains and its scene is less graphically precise. (Second judge pass, order swapped — scores are the average of both: Model B more clearly uses monumental stepped pyramids and the restricted palette, with crisp vintage-poster geometry and strong | Anthropic: Claude Fable 5.1: Model B delivers true step-pyramids, strict adhe…
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 and Model B both place the correctly colored objects in the requested arrangement, including the duck on top of the cube. Model B looks more photorealistic and has stronger material detail, while Model A leaves more empty space and better matches the requested 16:9 framing. (Second judge pass, order swapped — scores are the average of both: Model B clearly places the matte green cube left of the glossy red sphere, with the blue cylinder behind both and the yellow duck sitting on the cube | Anthropic: Claude Fable 5.1: Both bind colors correctly and place the duck o…
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 A more directly depicts a fishing boat riding a single immense cresting wave, with a muted indigo-and-cream dawn palette and convincing ukiyo-e linework. Model B is also visually striking and strongly evokes the tradition, but its multiple boats and added details make the prompt’s central scene less immediate. (Second judge pass, order swapped — scores are the average of both: Model B captures the flat planes, bold outlines, and decorative foam especially well, but its crowded waves and pa | Anthropic: Claude Fable 5.1: Model A delivers a coherent single fishing boat…
Prismatic cargo binding
On the loading dais of the orbital bazaar ship Vesper-9, depict exactly seven trade objects arranged in a clean three-quarter view, each with its own unambiguous color-and-material pairing and no color bleeding between items: a cobalt-blue velvet cushion, a matte saffron ceramic amphora, an emerald-green glass lantern, a crimson lacquered wood coffer, a lilac brushed-aluminum helmet, a white marble astrolabe, and a black obsidian dagger with a copper wire hilt; behind them, a dim starport with soft haze and tiny dock lights, lit by a warm key light from the left and a cool rim light from the right, ultra-clear object separation, cinematic realism, 16:9.
OpenAI: GPT-6 Luna: Model B more closely matches the specified colors and materials, especially the lilac helmet and copper-wired dagger, while presenting all seven objects clearly. Model A has a moodier cinematic starport, but several object colors and materials are less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B matches the specified colors and materials more consistently, including the lilac helmet and white marble astrolabe, while presenting all seven obj | Anthropic: Claude Fable 5.1: Model A delivers all seven objects with convinc…
Exact counting
A clean overhead flat lay on white marble of EXACTLY seven identical espresso cups arranged in a single straight row, each clearly separate and individually countable, even soft lighting, 16:9.
OpenAI: GPT-6 Luna: Model B presents exactly seven clearly separated, matching cups in a straighter, more evenly spaced row, with a clean overhead view and soft lighting. Model A also has seven cups, but the alignment and cup presentation are less consistent. (Second judge pass, order swapped — scores are the average of both: Model B presents exactly seven clearly separated, identical espresso cups in a straight row on white marble with a clean overhead view. Model A also shows seven in a row, but the added saucers | Anthropic: Claude Fable 5.1: Both Model A and Model B render exactly seven i…
Ukiyo-e storm leviathan
A colossal ribbon-finned sky leviathan glides above the basalt spires of the remote world Iri-Kho while three kite-sail hunters cling to scarlet rigging in the foreground, rendered faithfully in Edo-period ukiyo-e woodblock style: bold contour lines, flat yet expressive color fields, stylized clouds and waves translated into wind currents, patterned garments, visible woodgrain texture, and composition inspired by dramatic vertical movement and asymmetry; dawn light breaking through storm bands, no modern painting techniques, just authentic ukiyo-e aesthetics adapted to this fantasy sci-fi subject.
OpenAI: GPT-6 Luna: Model A more fully captures the ribbon-finned leviathan, storm-driven vertical movement, basalt spires, hunters on scarlet rigging, and textured ukiyo-e woodblock feel. Model B has a clear dawn scene and stylized motifs, but its creature reads more like a fish and its cleaner, framed graphic style feels less faithful to the requested dramatic woodblock aesthetic. (Second judge pass, order swapped — scores are the average of both: Model A more fully captures the ribbon-finned sky leviathan, basal | Anthropic: Claude Fable 5.1: Model A delivers a far more authentic Edo-perio…
Matchup powered by OpenRouter.