Head to head: Meta Muse Image Edit vs Google: Nano Banana Pro (Gemini 3 Pro Image)
Meta Muse Image Edit vs Google: Nano Banana Pro (Gemini 3 Pro Image)
By Ryan Merket · Published
Eight image-editing and generation tasks put perspective, reflective detail, object placement, text rendering, and scene fidelity under the microscope. The results turn on which model follows tightly specified visual instructions more consistently—not simply which produces the more polished image.
This was a broad test of composition as much as image quality: from a centered library aisle and chrome reflections to a precisely staged climbing-gear layout. Meta Muse Image Edit won all eight tasks, finishing at 66.7 to Nano Banana Pro’s 62.9; the judges’ verdict was decisive, with a tentative lean. That’s a consistent advantage across varied prompts, not a single standout image carrying the result. Muse’s edge was clearest when the prompt dictated where things had to go and what had to remain visible. Its library shelves receded more cleanly, the climbing gear followed the requested overhead arrangement, and the velodrome scene brought the bike, mechanics, rider, tools, and lap counter together without helmets. In the sandbar relay, it also supplied the small setting details—shells, gulls, wet sand, and the referee on a red crate—that made the scene feel faithful rather than merely lively. Nano Banana Pro had real strengths: its Rookery Cup poster rendered the key text and design well, and its spatial-layout image centered the rug more precisely. But those wins on individual details didn’t overturn the overall judgment. The poster was less focused, and in the room-layout comparison Muse’s lamp placement better matched the specified front-right position. On attribute binding, Muse showed the blue cylinder more clearly; on reflections, it made the apple and window reflections easier to read. **Verdict: Meta Muse Image Edit wins, 8–0.** Nano Banana Pro can make polished images and occasionally nails a particular compositional choice, but Muse was more dependable at turning detailed instructions into the requested scene.
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 the centered one-point view, with both shelf rows and the aisle converging cleanly while the books diminish consistently into the distance. Model B is atmospheric and convincing, but its projecting shelf structures make the perspective less clean and the aisle feel less uninterrupted. (Second judge pass, order swapped — scores are the average of both: Model A presents a clearer, more symmetrical one-point perspective, with the shelves and books receding consistent | Anthropic: Claude Fable 5.1: Model A delivers a clean, centered one-point pe…
Reflections & glass
A photorealistic close-up of a chrome teapot on a polished dark marble counter next to a glass of water; the window and a red apple on the counter must be correctly reflected in BOTH the chrome and the water's surface, dramatic side light, 16:9.
OpenAI: GPT-6 Luna: Model A more clearly captures the dramatic close-up and shows the window and red apple reflected in the chrome, with convincing glass and marble reflections. Model B is polished too, but the apple reflection is less evident and the composition feels less focused on the requested reflective details. (Second judge pass, order swapped — scores are the average of both: Model A more clearly shows the apple reflected in the chrome and on the polished counter, alongside strong window reflections and co | Anthropic: Claude Fable 5.1: Model A clearly reflects both the window grid a…
Cliff Gym Gear Layout
A realistic 16:9 overhead still life of climbing gear arranged on a weathered wooden platform at dawn: a coiled orange rope in the center, a chalk bag directly to the left of the rope, a pair of lime climbing shoes directly to the right of the rope, a silver carabiner on top of the rope coil, a folded topo map beneath the rope, a blue water bottle behind the rope, and a headlamp precisely between the chalk bag and the rope; cool early-morning light, crisp shadows, meticulous spatial layout.
OpenAI: GPT-6 Luna: Model A follows the requested overhead layout more precisely, with the rope centered and the surrounding gear clearly positioned as specified. Model B is appealing but uses a more oblique, diagonal arrangement that is less faithful to the meticulous spatial layout. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested overhead layout more precisely, with the rope centered and the other gear clearly arranged in the specified positions. Model B has appea | Anthropic: Claude Fable 5.1: Model A delivers a true overhead view with ever…
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 presents all four correctly colored objects in the requested arrangement with especially clear materials and polished studio lighting, though its framing is less wide than 16:9. Model B matches the aspect ratio, but its image is grainier and the blue cylinder is more heavily obscured. (Second judge pass, order swapped — scores are the average of both: Model A shows the blue cylinder more clearly behind the cube and sphere while preserving the requested color and placement relationships. | Anthropic: Claude Fable 5.1: Both outputs bind every color to the correct obj…
Rookery Cup Poster
A clean graphic design poster pinned to a cork wall in a community rowing club, rendered as a realistic close-up product shot with even soft lighting and sharp focus; the poster uses bold modern sans-serif typography and clearly legible text that reads exactly: "ROOKERY CUP" on the top line, "SAT 14 OCT" beneath it, and "PIER 9" at the bottom; navy, cream, and orange color palette, a simple icon of crossed oars, tidy margins, no extra words, no misspellings.
OpenAI: GPT-6 Luna: Model A presents a sharp, close-up poster with all three requested lines clearly legible, crossed oars, and the specified palette. Model B also renders the key text and design well, but its wider, busier scene includes distracting extra signage and less emphasis on the poster. (Second judge pass, order swapped — scores are the average of both: Model A gives the poster stronger visual prominence and cleanly renders all three requested lines with a polished navy, cream, and orange design. Model B | Anthropic: Claude Fable 5.1: Model A delivers a tight close-up product shot w…
No-Helmet Velodrome Pit
A cinematic 16:9 sports photograph of a bustling indoor velodrome pit area just before a midnight endurance race: a silver track bicycle on a repair stand, two mechanics in teal coveralls adjusting the chain, a rider in a maroon skinsuit stretching beside a crate of spare wheels, tool cases, floor reflections, and a glowing lap counter in the background — but deliberately NO helmets anywhere in the image, not on people, not hanging, not on shelves, not on the bike; crisp high-contrast arena lighting, shallow depth of field, realistic documentary style.
OpenAI: GPT-6 Luna: Model A more completely matches the requested pit scene, with a clearly featured silver bike on a stand, two teal-clad mechanics, a maroon rider stretching beside spare wheels, tools, reflective flooring, and a glowing lap counter; no helmets are visible. Model B captures the main elements too, but its composition is less focused and the pit details are less prominent. (Second judge pass, order swapped — scores are the average of both: Model A presents the bicycle on its repair stand, the two te | Anthropic: Claude Fable 5.1: Model A delivers every requested element vividl…
Sandbar Relay Tangle
A dynamic watercolor illustration of an eccentric beach sport scene at low tide: five athletes in a made-up relay game called duneball interacting in one coherent composition — one player in a saffron jersey leaping to catch a lime-green ball, one kneeling teammate drawing a boundary line in wet sand with a driftwood stick, two opponents in slate-blue uniforms lunging from opposite sides, and a barefoot referee on a red crate blowing a whistle — with a striped rescue tower, scattered shells, and distant gulls, balanced triangular composition, bright windy afternoon light, 16:9.
OpenAI: GPT-6 Luna: Model A more completely captures the requested low-tide details, including the barefoot referee on a red crate, scattered shells, gulls, and the striped rescue tower, while keeping the action clear in a cohesive watercolor scene. Model B is lively and well composed, but its setting is sparser and the referee’s placement and flying hat make the scene feel slightly less faithful. (Second judge pass, order swapped — scores are the average of both: Model A more fully realizes the specified five-play | Anthropic: Claude Fable 5.1: Model A nails the low-tide wet-sand setting and…
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 A places all requested furniture clearly, especially the floor lamp near the front-right corner, while Model B’s lamp sits farther along the right wall. Model B centers the rug more precisely and has a polished look, but Model A is slightly stronger on the specified spatial layout. (Second judge pass, order swapped — scores are the average of both: Model B follows the requested layout especially closely, with the rug centered on the floor and the lamp clearly occupying the front-right corn | Anthropic: Claude Fable 5.1: Model A places every element as specified—bed o…
Matchup powered by OpenRouter.