Head to head: Meta Muse Image Edit vs OpenAI: GPT-5.4 Image 2
Meta Muse Image Edit vs OpenAI: GPT-5.4 Image 2
By Ryan Merket · Published
Eight image-editing prompts put two models through a mix of composition, constraint-following, anatomy, text, and style tests. The comparison turns on how consistently each image fits the specific brief, not just how polished it looks.
Meta Muse Image Edit leads the saved results with an aggregate score of **65.7** to OpenAI: GPT-5.4 Image 2’s **62.1**. It won five tasks, OpenAI won two, and one was a tie—a clear result on this sample, not a statistical claim. Meta’s edge came in several briefs where matching the requested framing or scene details mattered: perspective and scale, negation, hands and anatomy, and the moonlit marsh. It also took the legible multi-line text task in the saved verdict. OpenAI’s wins were narrower: the Zorn-palette hellebore and the lynx scene. The tie on the spoonbill prompt is a reminder that a strong image can still miss a concrete requirement: the judges differed on how well each image met the requested bird count and panoramic framing. Across the set, the comparison rewards adherence to the brief as much as visual appeal. This assessment applies only to the saved task sample; it does not establish which model will perform better on other prompts. **Verdict: Meta Muse Image Edit wins this head-to-head, with a clear lead in these results.**
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 matches the requested wide, warmly lit photorealistic aisle, with shelves and books receding convincingly toward a centered vanishing point. Model B has especially strong one-point perspective, but its square framing and cooler lighting fit the prompt less well. (Second judge pass, order swapped — scores are the average of both: Model B has a strong centered vanishing point and clear scale recession, but its square framing misses the requested 16:9 format. Model A better mat | Anthropic: Claude Fable 5.1: Model A delivers a clean 16:9 one-point perspec…
Nine Spoonbills at Low Tide
A panoramic 16:9 wildlife photograph of exactly nine Eurasian spoonbills standing in a shallow tidal lagoon at dawn, each bird fully visible and individually countable, spaced apart in a loose zigzag across reflective water, with silver ripples, scattered eelgrass, and a misty sandbar behind them; ultra-realistic telephoto style, crisp feather edges, soft pink-gold sunrise light, low eye-level composition.
OpenAI: GPT-6 Luna: Model A better captures the requested panoramic, misty low-tide scene, with crisp birds, eelgrass, and a pleasing loose zigzag; however, it shows only about seven spoonbills. Model B has a lovely dawn palette but appears nearly square and shows about eight birds, with a more regular lineup. (Second judge pass, order swapped — scores are the average of both: Model B shows exactly nine countable spoonbills in reflective tidal water with dawn light and a sandbar, though its square framing misses th | Anthropic: Claude Fable 5.1: Both miss the required count of nine (Model A s…
Hellebore in Zorn Palette
A botanical still life of three hellebore blossoms in a hand-thrown stoneware vase, painted in classical oil style using only the Zorn palette: yellow ochre, vermilion, ivory black, and titanium white, with no other colors present anywhere in the image; dramatic side lighting, dark neutral backdrop, subtle tabletop shadow, finely controlled petals and leaves, elegant centered composition.
OpenAI: GPT-6 Luna: Model A makes the three hellebore blossoms and textured stoneware vase especially vivid, with dramatic lighting and finely resolved petals, though its cool gray foliage pushes the palette slightly. Model B has a restrained palette and elegant arrangement, but its more uniformly sepia treatment is less expressive and less clearly reflects the specified palette. (Second judge pass, order swapped — scores are the average of both: Model B better honors the restricted Zorn palette and presents three | Anthropic: Claude Fable 5.1: Model B delivers a convincing classical oil pain…
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 closely matches the requested 16:9 cozy nook, with warm afternoon light and none of the prohibited plants, lamps, or artwork. Model B also includes the requested furnishings and avoids prohibited items, but its square framing misses the specified aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model B includes the requested nook elements and avoids all prohibited objects, but its square framing misses the requested 16:9 format. Model A also respects the | Anthropic: Claude Fable 5.1: Model A delivers all required elements in a true…
Hands & anatomy
A natural candid photo of a person tying a friendship bracelet, both hands clearly visible with anatomically correct fingers and a plausible grip on the threads, soft daylight, shallow depth of field, 16:9.
OpenAI: GPT-6 Luna: Model A more closely matches the requested candid 16:9 photo, with both hands clearly visible in a plausible thread-tying grip and soft background blur. Model B also has convincing hands and a pleasant look, but its square framing departs from the requested aspect ratio and the bracelet-making action is less clear. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 candid photo, with both hands clearly visible in a believable grip and so | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame, conv…
Moonlit Marsh Without Birds
A serene 16:9 moonlit marsh landscape in realistic oil-painting style, with blackwater channels, reeds, cattails, water lilies, and drifting ground fog under a bright waxing gibbous moon, viewed from a low bank with reflections leading into the distance; the scene must contain no birds of any kind—no flying silhouettes, no perched shapes, no waterfowl, no feathers—only the empty wetland under silvery night light.
OpenAI: GPT-6 Luna: Model A more closely matches the requested 16:9 marsh scene, with clearly visible cattails, water lilies, low-bank foreground, drifting fog, and reflections leading into the distance; neither image contains birds. Model B is atmospheric and polished, but its near-square framing and less prominent cattails make it a slightly weaker fit. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested panoramic marsh, with prominent cattails, water lilies, d | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 format, a c…
Legible multi-line text
A minimalist event poster with three lines of crisp, correctly-spelled text stacked and centered: 'NIGHT MARKET' large on top, 'Fridays · 6–11pm' in the middle, 'Riverside Pier 4' at the bottom, on a deep navy background, clean sans-serif, subtle grain.
OpenAI: GPT-6 Luna: Model A renders all three lines correctly, but its spacing feels less balanced and its textured background is more pronounced. Model B has crisp, correctly spelled text, a restrained navy grain, and a cleaner, more cohesive centered layout. (Second judge pass, order swapped — scores are the average of both: Model B presents all three lines of crisp, correctly spelled text in a balanced centered layout on a deep navy, subtly grained background. Model A also renders the text accurately, but its te | Anthropic: Claude Fable 5.1: Both render all three lines correctly with clea…
Lynx Fur and Lichen Bark
A hyper-detailed oil painting of an Iberian lynx crouched on a twisted cork oak branch in a mossy ravine, the scene focused on exquisite fine texture: sharply rendered ear tufts, layered spotted fur, cracked bark, pale green lichen, tiny fern fronds, and damp stone surfaces, all crisp and unsmudged; cool north-light filtering through foliage, intimate close composition, naturalistic realism.
OpenAI: GPT-6 Luna: Model A more directly captures the intimate, crisp scene, with the lynx crouched on a twisted lichen-covered branch amid ferns and a cool ravine backdrop. Model B has a convincing lynx and atmospheric setting, but places it on a rocky ledge rather than the requested branch and feels less sharply detailed. (Second judge pass, order swapped — scores are the average of both: Model A more clearly shows the lynx crouched on a twisted, lichen-covered branch, with crisp fur, fern fronds, damp ravine de | Anthropic: Claude Fable 5.1: Model B delivers a more convincing Iberian lynx…
Matchup powered by OpenRouter.