Head to head: Wan vs Ideogram V4.5 Edit
Wan vs Ideogram V4.5 Edit
By Ryan Merket · Published
Eight image-generation and editing challenges put prompt fidelity, composition, and visual polish under the microscope. The results show how each model handles both tightly specified scenes and demanding spatial details.
Across the eight tasks, Wan’s edge came from doing more of what the prompts actually asked for: it won six tasks, while Ideogram V4.5 Edit won two. The limited confidence verdict makes this a clear win, not a photo finish. Wan was especially reliable on framing, style, and constraints. Its tea tin and umbrella-stall scenes better matched the requested flat-vector and watercolor treatments, while its library aisle hit the 16:9 composition and convincing depth. It also respected the reading-nook prompt’s explicit exclusions and kept more of the watch dial visible in the cashmere macro. That consistency wasn’t perfection. Wan added a second blue cylinder in the attribute-binding task, even as its overall arrangement and lighting were judged stronger. Ideogram’s best results were more than consolation: it delivered richer perfume-bottle reflections and caustics, and its isometric bedroom placed the desk, bed, rug, and lamp more precisely. The distinction is prompt execution. Ideogram can produce the more striking finish in a scene, but Wan more often preserves the requested format, style, and details together. **Winner: Wan, on a clear 6–2 task record and limited confidence.**
Midnight tea tin vector
A clean flat-vector editorial illustration of a dented cobalt-blue loose-leaf tea tin labeled "Morrow Finch No. 7" on a tiny kitchen windowsill at 2:13 a.m., half-open with a curl of bergamot steam and three spilled silver tea scoops beside it, mood wistful and sleepless, limited palette of ink navy, pale apricot, and mint, crisp geometric shadows from a single moonlit window, carefully balanced negative space, 16:9
OpenAI: GPT-6 Luna: Model A more closely matches the requested restrained flat-vector style, 16:9 framing, moonlit windowsill, and limited palette, with the label rendered clearly. Model B includes the time and a more literal tea spill, but its square, busier composition and extra colors stray further from the prompt. (Second judge pass, order swapped — scores are the average of both: Model B includes the labeled, open tin, visible tea leaves, three scoops with a spill, curling steam, and the 2:13 a.m. detail, thou | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 flat-vector…
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 fully matches the requested warm, photorealistic 16:9 library aisle, with a clear central vanishing point and convincing scale progression. Model B has strong one-point perspective, but its square framing misses the requested aspect ratio and its lighting feels less warmly overhead. (Second judge pass, order swapped — scores are the average of both: Model B has the clearest, most coherent one-point perspective, with shelves and books shrinking consistently toward a centered vanishin | Anthropic: Claude Fable 5.1: Model A delivers a clean one-point perspective…
Perfume vanity reflections
A photoreal morning vanity scene with a faceted glass perfume bottle named "Lune Harbor" beside a chrome eyelash curler and a half-filled tumbler of water on a pale stone countertop, viewed slightly above at three-quarter angle so the bottle refracts the window frame and a striped curtain, the chrome tool mirrors nearby objects with correct distortion, and soft sunbeams create layered caustics and transparent shadows through the glass, high realism, 16:9
OpenAI: GPT-6 Luna: Model A follows the requested 16:9 vanity setup closely, with the named faceted bottle, chrome curler, half-filled tumbler, striped curtain, and convincing sunlit glass effects. Model B has especially attractive reflections and lighting, but its square framing and less direct match to the requested view make it slightly less adherent overall. (Second judge pass, order swapped — scores are the average of both: Model B gives the perfume bottle stronger faceting and more vivid sunlit caustics, with | Anthropic: Claude Fable 5.1: Model A nails the 16:9 framing, striped curtain…
Car-free umbrella stall
A rainy evening neighborhood market scene centered on a cheerful consumer umbrella repair stall called "Drizzle Clinic" with dozens of hanging folded umbrellas in plum, ochre, and teal, wet cobblestones, puddles, hand-painted price tags, and bundled pedestrians in the background, but absolutely no cars, taxis, buses, bicycles, or scooters anywhere in the image, rendered as a soft watercolor illustration with precise storefront details and cinematic side lighting, 16:9
OpenAI: GPT-6 Luna: Model A more fully matches the requested soft watercolor, 16:9 neighborhood-market scene, with a detailed storefront, abundant folded umbrellas, wet cobblestones, and pedestrians; neither image contains vehicles. Model B has striking rainy lighting and a clearly legible main sign, but its near-photorealistic style and square framing depart from the prompt, and several smaller signs are garbled. (Second judge pass, order swapped — scores are the average of both: Model A better matches the request | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 soft waterc…
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 better matches the requested 16:9 tabletop scene, matte green cube, and soft studio lighting, though it adds a second blue cylinder. Model B uses only one blue cylinder, but misses the requested aspect ratio and has much harsher lighting and a less clearly separated arrangement. (Second judge pass, order swapped — scores are the average of both: Model B includes the requested objects in the correct colors and positions, while Model A adds a second blue cylinder despite the prompt specify | Anthropic: Claude Fable 5.1: Model A correctly binds all colors, keeps the c…
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 includes the requested reading nook elements and warm afternoon light while correctly avoiding plants, lamps, and wall art; it also uses the requested 16:9 framing. Model B is visually appealing, but prominently includes plants and framed artwork and does not match the requested aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model B is warm and polished but prominently includes many plants and framed artwork, directly violating the key negative constrai | Anthropic: Claude Fable 5.1: Model A fully respects the negation constraints…
Cashmere dial macro
An extreme close-up product image of a luxury wristwatch resting on folded graphite cashmere gloves, showing razor-sharp knitted fiber texture, brushed steel crown teeth, tiny engraved minute markers, faint dust specks on sapphire crystal, and a narrow stitched leather strap edge, shot in realistic studio photography with controlled raking light, shallow but accurate depth of field, and no smudging or melted detail
OpenAI: GPT-6 Luna: Model A shows the watch dial, engraved markers, and toothed crown clearly while retaining the graphite cashmere texture. Model B has exceptionally crisp knit and brushed metal, but its side-on angle hides nearly all of the dial and crown, weakening adherence to the requested details. (Second judge pass, order swapped — scores are the average of both: Model A shows more of the watch dial, fine minute markers, and toothed crown while retaining a clear close-up view of the graphite knit. Model B ha | Anthropic: Claude Fable 5.1: Model A delivers nearly every requested element…
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 clearly locates the bed, centered rug, and front-right lamp in a polished isometric room. Model A is clean and well composed, but its desk sits more to the side of the window and the illustration is less refined. (Second judge pass, order swapped — scores are the average of both: Model B places the bed along the left wall, the desk beneath the back-wall window, and the lamp in the front-right corner, with a polished, cohesive isom | Anthropic: Claude Fable 5.1: Both place the bed against the left wall, the d…
Matchup powered by OpenRouter.