Head to head: Google: Nano Banana Pro (Gemini 3 Pro Image) vs Ideogram V4.5 Edit
Google: Nano Banana Pro (Gemini 3 Pro Image) vs Ideogram V4.5 Edit
By Ryan Merket · Published
We compared two image models on eight prompts spanning precise text, object counts, anatomy, restricted palettes, and editorial-style scenes. The results show how well each model follows the brief when composition and small details matter as much as visual polish.
Google’s Nano Banana Pro wins this matchup, and not by a photo finish: it scored **65.2 to Ideogram V4.5 Edit’s 53.5**, taking **seven of eight tasks** with **limited confidence**. Its edge came from repeatedly honoring the whole brief—not just making an attractive image—including aspect ratio, layout, subject details, and exact wording. That consistency showed up across very different jobs. Google rendered all three lines of the requested text legibly and exactly, produced seven cups in the specified row, and kept the campsite palette tighter. It also better captured the greenhouse gardener’s tearful joy and the vintage potting-shed still life, where Ideogram drifted to square framing or muddier object details. Google wasn’t flawless: its macro “ladybug” reads more like a beetle, and the seed-saving scene partly obscures the fingers and crops the gardener’s head. But Ideogram’s one win—the empty library aisle—was a genuine strength: its centered perspective and receding shelves were cleaner, even though it missed the requested widescreen format. Across the rest of the set, Ideogram more often traded prompt fidelity for a polished image. **Final call: Nano Banana Pro is the stronger choice for prompt-led image generation.** Ideogram delivers a standout perspective scene, but Google’s broader reliability on text, counts, framing, and scene-specific details makes this a decisive result.
Macro ladybug on thyme
An ultra-detailed macro close-up photograph of a tiny jade-green ladybug standing on the fuzzy purple flower of creeping thyme, shot with a true macro lens so the insect’s faceted eyes and pollen grains are razor sharp while the background melts into creamy circular bokeh, diagonal composition, dewdrops catching early morning sun, extremely shallow depth of field, botanical realism, 16:9.
OpenAI: GPT-6 Luna: Model B reads more clearly as a ladybug, the prompt’s defining subject, and keeps the macro flower-and-bokeh mood despite its non-jade coloring and square framing. Model A has stronger wide composition and crisp detail, but depicts a distinctly different kind of beetle. (Second judge pass, order swapped — scores are the average of both: Model A has the requested jade-green coloring, crisp macro detail, dewdrops, shallow depth of field, and 16:9 framing, though its insect reads more like a beetle | Anthropic: Claude Fable 5.1: Model A delivers a jade-green beetle with razor…
Seed-saving hands anatomy
A naturalistic full-body image of a lanky community gardener seated sideways on a wooden stool in a sunlit seed library, carefully splitting open a striped pod over a paper envelope with both hands fully visible and anatomically correct — five distinct fingers on each hand, believable knuckles, relaxed wrists, accurate forearm length, natural shoulder alignment, proportional legs and torso — scattered seed packets, labeled jars, and drying herbs around the room, soft window light, documentary realism, 16:9.
OpenAI: GPT-6 Luna: Model A better matches the requested wide, full-body seated scene and richly detailed seed-library setting, though its fingers are partly obscured and the gardener’s head is cropped. Model B depicts the seed-splitting action and hands more clearly, but its tight framing omits the full-body composition and much of the requested environment. (Second judge pass, order swapped — scores are the average of both: Model A better captures the wide seed-library setting, sideways seated pose, stool, jars, | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 documentary…
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 presents all three requested lines centered and legibly, with the specified wording, clean sans-serif typography, and subtle grain on deep navy. Model B is bold and readable, but breaks the title into two lines, changes the middle separator, and adds a more distressed decorative treatment than requested. (Second judge pass, order swapped — scores are the average of both: Model A follows the requested three-line centered layout with the exact wording and clean, legible sans-serif text. Mo | Anthropic: Claude Fable 5.1: Model A renders all three lines exactly as spec…
Joyful greenhouse triumph
A realistic editorial-style portrait of a middle-aged orchid breeder in a misty hilltop greenhouse at dawn, clutching a clay pot with a single rare speckled cymbidium bloom and convincingly showing sudden, unmistakable tearful joy after years of failed crosses — wide open smiling mouth, raised cheeks, wet eyes, tense delighted hands, posture leaning forward in disbelief — rows of tagged seedlings and foggy glass panels behind her, soft golden sidelight, crisp facial detail, natural skin texture, 16:9.
OpenAI: GPT-6 Luna: Model A more fully captures the requested tearful, unmistakable joy, speckled orchid, tagged seedlings, and dawn-lit greenhouse in the specified widescreen portrait format. Model B has a compelling joyful expression and polished composition, but lacks visible tears and is square rather than 16:9. (Second judge pass, order swapped — scores are the average of both: Model A more fully captures the requested tearful, unmistakable joy and includes the speckled bloom, tagged seedlings, misty greenhous | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 editorial f…
Bound colors in potting shed
A clean still-life scene inside a potting shed in the style of a vintage travel poster: exactly four gardening objects arranged on a weathered bench under bright side light — a matte cobalt-blue metal watering can, a glossy saffron-yellow ceramic seed tray, a deep burgundy leather gardening glove, and a pale mint-green glass spray bottle — each object clearly separate, each color bound only to its own specified object and material, no color swapping, simplified bold shapes, elegant poster composition, 16:9.
OpenAI: GPT-6 Luna: Model A better matches the requested 16:9 vintage travel-poster still life, with all four objects clearly recognizable and their specified colors largely kept distinct; its added poster text is a minor deviation. Model B has polished illustration work, but the seed tray is less clearly a seed tray, the spray-bottle trigger introduces extra colors, and the square framing misses the requested format. (Second judge pass, order swapped — scores are the average of both: Model A better captures the vi | Anthropic: Claude Fable 5.1: Model A delivers a true 16:9 vintage travel pos…
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 A better follows the requested 16:9 format and presents a clear mountain campsite in a mostly restricted palette. Model B is polished, but its square format and broader range of hues make it less faithful to the prompt. (Second judge pass, order swapped — scores are the average of both: Model A better matches the requested 16:9 flat-vector campsite scene and keeps its palette more controlled, while Model B is square and uses pronounced extra sunset hues. Both have appealing, balanced compo | Anthropic: Claude Fable 5.1: Model A holds tightly to the four-color constra…
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 shows exactly seven clearly separated, identical cups in a straight row on white marble and matches the requested 16:9 format. Model B also has seven cups, but its square framing and lower placement of the row adhere less closely to the requested composition. (Second judge pass, order swapped — scores are the average of both: Model A presents exactly seven clearly separated cups in a straight row and matches the requested 16:9 overhead marble layout. Model B shows espresso in the cups, b | Anthropic: Claude Fable 5.1: Model A shows exactly seven identical empty whi…
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 B more cleanly delivers the empty, centered one-point-perspective aisle, with books and shelves receding consistently toward the vanishing point. Model A has the requested 16:9 framing and warm light, but chairs and ladders clutter the aisle; Model B’s square framing misses the requested aspect ratio. (Second judge pass, order swapped — scores are the average of both: Model B more directly fulfills the empty, centered one-point-perspective aisle, with books and shelves receding convincingl | Anthropic: Claude Fable 5.1: Model A delivers the requested 16:9 frame, conv…
Matchup powered by OpenRouter.