Head to head: Wan vs Ideogram V4.5 Edit
Wan vs Ideogram V4.5 Edit
By Ryan Merket · Published
Eight image tasks put Wan and Ideogram V4.5 Edit through workshop, interior, object-placement, and counting briefs. The comparison turns on how each handles specific constraints as well as overall composition.
Ideogram V4.5 Edit leads the saved aggregate, 60.9 to Wan’s 56.0, with 4 task wins to 2 and 2 ties. The gap supports a **slight edge (not statistical)**, not a decisive separation; this assessment applies to the saved task sample only. Ideogram’s clearest wins came on the negation brief, where it excluded forbidden plants, lamps, and artwork, and the hands-in-glass-studio scene, where it kept the artisan’s face and both working hands in frame. It also took the loom close-up and bakery-layout verdicts, though the bakery comparison was close: Wan’s timer placement was more faithful, while Ideogram better captured the requested pre-dawn mood and seed-jar placement. Wan’s two wins were more narrowly defined but meaningful. It delivered the cleaner, unobstructed one-point-perspective aisle, and gave the object-binding scene stronger material contrast with the blue cylinder clearly behind the other objects. The spatial-layout and nine-seed-jars tasks were ties: both models missed parts of the briefs, in different ways. **Final call: Ideogram V4.5 Edit, by a slight editorial edge on this saved task sample.** Its advantage comes from more wins across varied briefs, while Wan remains the stronger pick for clean perspective and crisp object relationships.
Hands in glass studio
A documentary-style 16:9 photograph of a stained-glass artisan at work in a narrow workshop, shown from mid-thigh up at a large oak bench, clearly visible face and full upper body, natural body proportions, both hands fully in frame and unobstructed: her left hand steadies a cobalt-blue panel while her right hand uses a brass fid tool, with all ten fingers distinct and anatomically correct, short trimmed nails, realistic knuckles and wrist angles; scraps of colored glass, a soldering iron, pattern sketches, and a lead came roller scattered across the bench; warm late-afternoon window light from the left, crisp focus, believable skin texture, candid editorial realism.
OpenAI: GPT-6 Luna: Model B better matches the requested framing, with the artisan’s face and upper body fully visible, both hands working on the blue panel, and a convincing, well-equipped workshop. Model A has clear hand placement and warm light, but the face is cropped and the scene omits some of the specified workshop details. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested mid-thigh-up workshop scene, with the artisan’s face, both hands, stained-glass pa | Anthropic: Claude Fable 5.1: Model A delivers the hand-focused brief most cl…
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 closely follows the prompt, with a clean, empty aisle, warm overhead lighting, and strong, balanced one-point perspective. Model B is atmospheric, but the ladder and book on the floor conflict with the empty-aisle request. (Second judge pass, order swapped — scores are the average of both: Model A delivers a clean, balanced one-point-perspective aisle with consistently receding shelves and an unobstructed floor. Model B has rich warm lighting, but the ladder and book on the floor co | Anthropic: Claude Fable 5.1: Model A delivers a clean, truly empty, long ais…
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 is warm and polished, but includes visible greenery outside the window and a framed picture, violating two explicit exclusions. Model B includes the requested chair, books, tea, and window while keeping plants, lamps, and artwork out of the frame. (Second judge pass, order swapped — scores are the average of both: Model B includes the requested reading nook elements and warm window light while avoiding the explicitly forbidden plants, lamps, and artwork. Model A is invitingly composed, b | Anthropic: Claude Fable 5.1: Model A violates the negation constraints with…
Loom detail closeup
An ultra-detailed macro-style 16:9 image of a textile conservator repairing an antique jacquard tapestry on a restoration loom, rendered in hyperreal digital painting: sharply resolved fabric weave, frayed silk threads, tiny couching stitches, tarnished brass loom hardware, lint on black velvet sleeves, stray fibers catching cool task-lamp light, and handwritten thread tags pinned along the edge; no smudging, no painterly blur, every material surface crisply differentiated; shallow depth of field centered on the repair area, with the conservator’s gloved fingertips entering the frame at the lower right.
OpenAI: GPT-6 Luna: Model A more clearly delivers the crisply resolved repair area, cool task light, tarnished loom hardware, and densely textured tapestry, while Model B has a more dramatic close-up composition but relies on stronger blur and warmer lighting than requested. Both include handwritten tags and gloved hands, though the tag text is not reliably legible in either image. (Second judge pass, order swapped — scores are the average of both: Model B more fully captures the requested macro-style restoration s | Anthropic: Claude Fable 5.1: Model B follows the macro-style brief far more…
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 presents the blue cylinder behind the other objects and has stronger, more vivid material contrast, while preserving the duck-on-cube arrangement. Model B follows the object placement well but leaves substantially more empty space and has softer, less defined rendering. (Second judge pass, order swapped — scores are the average of both: Model A more convincingly renders the requested materials and studio tabletop scene, with the blue cylinder clearly positioned behind the ot | Anthropic: Claude Fable 5.1: Model A delivers a convincingly photorealistic…
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 keeps the rug centered, while Model A’s desk sits noticeably left of the window. Model A has a cleaner flat-vector isometric look, but Model B better matches the requested spatial arrangement overall. (Second judge pass, order swapped — scores are the average of both: Model B places the desk beneath the window and includes the requested left-side bed, centered round rug, and front-right lamp. Model A has a polished isometric compo | Anthropic: Claude Fable 5.1: Model A has a clean, consistent isometric box b…
Nine seed jars
A clean watercolor illustration of a seed archivist cataloging EXACTLY 9 distinct glass specimen jars on a steel table, each jar fully visible and individually countable, arranged in an irregular but non-overlapping layout; every jar contains a different heirloom seed type with a handwritten label and slightly different fill level, and there are no extra jars anywhere in the room; the archivist in a moss-green lab coat stands behind the table making notes on a clipboard; pale skylight, soft shadows, whitewashed archive walls, meticulous composition, 16:9.
OpenAI: GPT-6 Luna: Model A has a more meticulous, legible arrangement with distinct seed types, labels, and one archivist, but it appears to show only eight table jars and adds three jars on a shelf. Model B avoids extra jars elsewhere, yet its crowded overlapping display is difficult to count and it depicts two archivists rather than one. (Second judge pass, order swapped — scores are the average of both: Model A has a clearer, individually countable arrangement and a single archivist, but shows eight jars on the | Anthropic: Claude Fable 5.1: Model A shows only eight jars on the table plus…
Bakery layout logic
A cinematic 16:9 scene inside a pre-dawn bakery kitchen, in realistic illustration style, showing a precise arrangement of objects: a copper mixing bowl is on top of a blue digital scale; a tray of crescent rolls is to the left of the scale; a glass jar of poppy seeds is behind the mixing bowl; a folded striped towel is between the tray and the scale; a wooden rolling pin is to the right of the scale; and a small brass timer is in front of the scale but not touching it. The baker stands in the background near the oven, warm amber practical lighting, eye-level composition, every listed object clearly visible and placed exactly as described.
OpenAI: GPT-6 Luna: Model A follows the requested object layout especially clearly, with the timer separated in front of the scale and the baker near the oven; Model B is more cinematic but places the timer beside the scale rather than clearly in front. Model B has moodier lighting, while both images are polished and show the listed objects well. (Second judge pass, order swapped — scores are the average of both: Model B better matches the requested pre-dawn cinematic mood and places the seed jar behind the bowl, w | Anthropic: Claude Fable 5.1: Both place the bowl on the blue scale, rolls le…
Matchup powered by OpenRouter.