Head to head: Meta Muse Image Edit vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
Meta Muse Image Edit vs Google: Nano Banana 2 (Gemini 3.1 Flash Image)
By Ryan Merket · Published
A close image-editing matchup turns on more than getting the requested objects into the frame: composition, style, and how clearly the scene reads all matter. The results split across detailed spatial briefs, controlled visual styles, and atmospheric scenes.
Meta Muse Image Edit takes the saved verdict, but only by a **slight edge—not a statistical one**: its aggregate score is **64.5** to Nano Banana 2’s **61.6**. The task record is similarly close, with Meta winning four tasks, Google two, and two ties. Meta’s clearest advantage came in scenes where placement and the main action needed to read immediately. It was stronger on the attribute-binding tabletop, the desert airstrip’s layered depth, and the woodblock-style fishing-boat scene; it also won the funicular-market task despite missing the requested widescreen format. That last result is a reminder that Meta’s detail sometimes came at the expense of format or compositional discipline. Google had the better result on the rainy coach stop and the restricted-palette campsite, where atmosphere and a clean, controlled vector treatment counted. The harbor-tram and hands-and-anatomy tasks were ties: each model made important parts of the brief legible, but neither had an unambiguous overall advantage. **Final call: Meta Muse Image Edit, by a slight editorial edge on this task sample.** Its wins on spatial relationships and prompt-specific action outweigh Google’s strengths in two distinct visual treatments, but the narrow score gap and two ties make this a close verdict—not a broad claim of capability superiority.
Harbor Tram Positions
A crisp editorial illustration of a compact electric harbor tram yard at blue hour, 16:9: a mustard-yellow tram numbered 204 is parked exactly to the left of a teal baggage cart; a red motor scooter is directly in front of the baggage cart; a stack of three silver shipping trunks sits on top of a low wooden pallet between the tram and the scooter; behind the tram stands a narrow glass shelter with a timetable board; a white seagull is perched on the shelter roof; and a mechanic in navy coveralls kneels to the right of the scooter tightening a wheel bolt. Compose the scene so every named object is fully visible and the specified left/right, in front of, on top of, between, and behind relationships are unmistakable.
OpenAI: GPT-6 Luna: Model B keeps the tram, shelter, cart, scooter, trunks, seagull, and mechanic fully visible while making their requested positions easy to read. Model A is richly detailed, but the tram is cut off at the edge and the crowded foreground makes the arrangement less clear. (Second judge pass, order swapped — scores are the average of both: Model B makes the requested left/right and front/behind relationships clear, with the tram, shelter, cart, scooter, trunks, seagull, and kneeling mechanic all ful | Anthropic: Claude Fable 5.1: Model A satisfies every spatial relationship wi…
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 closely matches the soft studio tabletop scene, with all four objects clearly identifiable in the requested positions and colors. Model B also gets the object relationships right, but its rustic, cluttered setting and less polished presentation are less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model B follows the requested 16:9 framing and object relationships, while Model A presents the same color and placement bindings in a cleane | Anthropic: Claude Fable 5.1: Model A nails every binding: matte green cube l…
Funicular Market Arrival
A richly detailed vintage travel poster showing a hillside funicular station above the fictional port town of Brinza, 16:9: a cream-and-coral funicular car has just arrived as six distinct travelers interact in front of it—a florist passing a bundle of orange lilies to a cyclist in a tan cap, two children pointing toward the sea, a ticket inspector clipping a stub for an elderly woman with a striped suitcase, and a porter wheeling a brass-handled trunk down the platform. Add a street musician with a small accordion near a poster kiosk, gulls circling beyond the railing, terraced houses below, and a coherent, balanced composition where all subjects feel part of one lively transport scene.
OpenAI: GPT-6 Luna: Model A is richly detailed and includes nearly every requested interaction, but its square format misses the specified 16:9 layout. Model B uses the requested wide composition and brings the travelers, funicular, sea, and town together clearly, though its oversized headline is somewhat distracting. (Second judge pass, order swapped — scores are the average of both: Model B fits the requested wide poster format and stages the specified interactions clearly, including the flower handoff, ticket in | Anthropic: Claude Fable 5.1: Model A delivers every requested element (lilie…
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 B makes the friendship bracelet and the tying action especially clear, with a lively candid composition and plausible hand positions, though it includes another person. Model A is a strong, natural close-up with convincing hands, but the bracelet-making action is less immediately legible. (Second judge pass, order swapped — scores are the average of both: Model A keeps both hands and the threadwork as the clear focal point, with a plausible close-up grip and soft daylight. Model B more exp | Anthropic: Claude Fable 5.1: Model A delivers a clean, hand-focused candid w…
Desert Airstrip Depth
A highly realistic wide-angle view of a remote desert airstrip at sunrise, 16:9, with strict linear perspective: in the immediate foreground a small orange fuel tug occupies the lower right corner; farther back and centered on the tarmac stands a twin-prop cargo plane with its nose facing slightly left; much farther away near the hangar, a white pickup truck appears correctly smaller; and on the horizon a row of four wind-bent utility poles recedes toward a single vanishing point. A mechanic on a ladder reaches the plane’s right engine, stacked fuel drums sit beside the hangar, and long low shadows stretch consistently across the ground. Relative sizes must be believable and the perspective must feel precise, with no warped aircraft geometry or mismatched scale.
OpenAI: GPT-6 Luna: Model A more closely matches the requested placement and depth, with the orange tug prominent in the lower-right foreground, the cargo plane centered, and the truck, hangar, mechanic, drums, and receding poles all clearly visible. Model B has appealing sunrise lighting, but the plane sits farther left and the foreground tug is less prominent, making the composition and scale relationships less faithful. (Second judge pass, order swapped — scores are the average of both: Model A conveys the depth | Anthropic: Claude Fable 5.1: Model A delivers the requested layered depth mo…
Rainy Coach Stop Realism
Photorealistic nighttime scene at a rural coach stop after rain, 16:9: a dark green intercity bus labeled 'North Vale 7' idles beside a wet concrete curb while warm interior lights glow through slightly fogged windows; a stainless-steel luggage rack reflects the amber bus lights, rainwater beads on the painted metal body, the rubber tires are damp and matte, a cedar bench under the shelter shows soaked grain and darker patches, a passenger in a wool coat holds a black umbrella with realistic wet fabric sheen, and the bus driver’s face and hands are naturally lit by dashboard light. Headlights create believable reflections in puddles, the shelter glass catches faint glare, and all materials—painted metal, wet concrete, wood, skin, fabric, glass—must look physically convincing.
OpenAI: GPT-6 Luna: Model A more closely matches the requested scene, with the clearly rendered “North Vale 7” label, warm fogged windows, reflective luggage rack, wet bench, umbrella-carrying passenger, and dashboard-lit driver. Model B has a convincing rainy-stop atmosphere, but its bus labeling is less faithful and the overall scene is less focused. (Second judge pass, order swapped — scores are the average of both: Model B more completely captures the requested stop, with a clearly visible umbrella-holding pass | Anthropic: Claude Fable 5.1: Model A renders the text and most elements but…
Restricted palette
A flat-vector illustration of a mountain campsite at dusk using ONLY four colors — cream, burnt orange, teal, and charcoal — with no gradients and no stray off-palette colors, balanced composition, 16:9.
OpenAI: GPT-6 Luna: Model B delivers a polished, balanced flat-vector campsite scene with a clear dusk atmosphere and a tightly controlled palette. Model A includes the requested elements and colors, but its rough, oversized foreground shapes feel less balanced and less like clean flat-vector artwork. (Second judge pass, order swapped — scores are the average of both: Model B delivers a polished, balanced flat-vector campsite at dusk and appears to stay within the four-color palette. Model A also uses the requested | Anthropic: Claude Fable 5.1: Model A holds strictly to the four-color palett…
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 towering cresting wave at dawn, with a strong muted indigo-and-cream palette and expressive woodblock-style foam. Model B is also beautifully rendered in the ukiyo-e tradition, but its several boats and more distant wave make the central action less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model A more clearly conveys a fishing boat caught beneath a dramatic cresting wave at dawn, with the ri | Anthropic: Claude Fable 5.1: Model A follows the literal brief tightly (sing…
Matchup powered by OpenRouter.