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 across fashion details, exact counts, text rendering, and scene direction. The comparison turns on how each model balances precise prompt-following with polished composition.
The saved scores are **64.3 for Meta Muse Image Edit** and **65.3 for Google: Nano Banana 2 (Gemini 3.1 Flash Image)**. That one-point gap sits alongside a **2–3–3 split** in task wins, losses, and ties—not a decisive separation in this sample. Meta’s strongest results came in hands and anatomy, where the thread-working grip felt natural, and counted raincoats, where all 11 coats were clearly separated in a clean studio composition. Google took brocade macro suit, exact counting, and glove shop portrait; its espresso cups were both countable and filled, while the brocade looked more like a refined catalog image than damaged fabric. The models tied on legible multi-line text, golden-hour tailor, and attribute binding. Those ties reflect different trade-offs: restrained grain versus heavier texture, a more evocative tailoring scene versus clearer signage, and lifelike materials versus roomier object placement. **Final call: Too close to call (not statistical). Keep the saved result as a tie; Google’s slightly higher aggregate does not make this a clear win. This assessment applies only to the saved task sample.**
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 has a natural, convincing close-up of both hands working the colorful threads, with especially strong soft daylight and shallow depth of field. Model B shows the person and the nearly finished bracelet more clearly, but its hand-and-thread interaction looks slightly less natural and the background is busier. (Second judge pass, order swapped — scores are the average of both: Model A gives the hands more prominence and shows a natural, plausible grip as the person works the threads, with | Anthropic: Claude Fable 5.1: Model A delivers crisp, anatomically convincing…
Brocade macro suit
An extreme close-up fashion still life of a midnight-teal brocade suit jacket draped over a walnut chair, showing razor-sharp fabric weave, metallic thread, hand stitching, horn buttons, stray fibers, and subtle lint without any smudging, fine-detail luxury catalog photography, controlled studio lighting raking across the surface from the left, dark neutral background
OpenAI: GPT-6 Luna: Model B better captures the refined luxury-catalog look, with a clear drape over a walnut chair, controlled left-side lighting, and crisp brocade detail. Model A has impressive macro texture, but its excessive loose fibers and rough stitching make the jacket look damaged rather than subtly linted. (Second judge pass, order swapped — scores are the average of both: Model B presents a refined luxury-catalog close-up with the midnight-teal brocade, metallic threads, horn button, walnut chair, and c | Anthropic: Claude Fable 5.1: Model A checks more literal boxes (multiple hor…
Exact counting
A clean overhead flat lay on white marble of EXACTLY seven identical espresso cups arranged in a single straight row, each clearly separate and individually countable, even soft lighting, 16:9.
OpenAI: GPT-6 Luna: Model A has exactly seven clearly separated cups in a straight row, but the cups appear empty. Model B also has exactly seven evenly arranged cups and visibly contains espresso, matching the prompt more closely while maintaining a clean overhead composition. (Second judge pass, order swapped — scores are the average of both: Model B more clearly depicts exactly seven identical espresso cups in a straight, individually countable row, with espresso visible in each cup. Model A also has seven cups | Anthropic: Claude Fable 5.1: Both Model A and Model B deliver exactly seven 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 B presents all three lines crisply and correctly, with balanced centering and restrained grain that better suits the minimalist brief. Model A is also legible and well composed, but its noticeably heavier texture is less subtle. (Second judge pass, order swapped — scores are the average of both: Model B presents all three lines crisply and correctly, centered on a deep navy background with restrained grain and balanced negative space. Model A also renders the text correctly and has strong | Anthropic: Claude Fable 5.1: Both render all three lines correctly spelled an…
Counted raincoats
A clean editorial product photograph of exactly 11 distinct raincoats hanging separately on a brushed steel rolling rack in a fashion studio, each coat a different cut and color with visible spacing so every single garment is fully countable, polished concrete floor, pale dove-gray backdrop, crisp softbox lighting, front-facing symmetrical composition, ultra-realistic commercial apparel photography, 16:9
OpenAI: GPT-6 Luna: Model A presents exactly 11 distinct, colorful raincoats in a polished, nearly symmetrical studio composition, with each garment easy to count. Model B also appears to have 11 coats, but they are more tightly crowded and the visible studio background makes the presentation feel less clean and editorial. (Second judge pass, order swapped — scores are the average of both: Model A presents exactly 11 distinct, differently colored raincoats in a clean, polished studio setup, with varied cuts and a s | Anthropic: Claude Fable 5.1: Both Model A and Model B deliver exactly 11 cou…
Golden hour tailor
A cinematic street-fashion scene at golden hour: a young tailor in a saffron trench coat adjusting a mannequin outside a tiny shop called Vesper Hem on a sloping cobblestone lane, long amber shadows, sun flaring off the shop window, warm nostalgic mood, realistic editorial photography with shallow depth of field, composed from curb height, 16:9
OpenAI: GPT-6 Luna: Model A more fully captures the requested cinematic golden-hour mood, with a saffron-coated tailor actively fitting a pinned mannequin, a glowing cobbled lane, and a striking sun flare. Model B has a clearer shop sign and a pleasing street scene, but its lighting and shallow-focus editorial look are less evocative, and the window lettering is cluttered. (Second judge pass, order swapped — scores are the average of both: Model A more strongly captures the intimate tailoring action, saffron coat, | Anthropic: Claude Fable 5.1: Model A delivers a dramatic saffron trench and s…
Glove shop portrait
A realistic portrait of a woman in a small bespoke glove shop, seated three-quarter view at a wooden fitting table, both hands clearly visible as she gently stretches a pearl-gray leather glove over her left hand while her right hand holds the glove opening, all ten fingers anatomically correct and naturally posed, accurate wrists and proportions, soft window light, elegant contemporary fashion photography, 16:9
OpenAI: GPT-6 Luna: Model A more clearly captures the active glove-fitting gesture, with the pearl-gray glove and bespoke workshop details rendered in warm, polished fashion photography. Model B shows the seated shop setting more fully, but the gesture feels more like adjusting an already-worn glove and is less visually engaging. (Second judge pass, order swapped — scores are the average of both: Model B more directly matches the seated three-quarter portrait, with the pearl-gray glove fitted on her left hand as he | Anthropic: Claude Fable 5.1: Model A delivers a richly detailed shop with a…
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 has especially convincing photorealistic materials and soft studio lighting while correctly placing the duck on the green cube, with the red sphere to the right and blue cylinder behind. Model B also follows the arrangement well and gives the objects more breathing room, but Model A looks more polished and lifelike overall. (Second judge pass, order swapped — scores are the average of both: Model B presents all four objects clearly in the requested left-to-right arrangement, with the duc | Anthropic: Claude Fable 5.1: Both models bind every color to the correct obj…
Matchup powered by OpenRouter.