50 Gemini Nano Banana Prompts for Perfect AI Image

Good Gemini Nano Banana prompts are specific ones. The model follows long, detailed instructions closely, which means a vague prompt gets you a vague picture while a precise one lands close to what you actually had in mind.

It is also unusually strong at two things most image models fumble: rendering legible text inside an image, and editing an existing photo without redrawing everything around the edit. Nano Banana handles both well, and Nano Banana Pro is noticeably more reliable on typography. Two of the five sections below are built specifically around those strengths.

Below are 50 copy-ready prompts grouped into artistic styles, lighting and camera control, creative concepts, text and typography, and editing. Each is written as a full brief rather than a phrase, so you can swap the subject and keep the structure. If you would rather tighten a prompt before spending a generation on it, Chat Smith lets you draft it in chat first.

Gemini Nano Banana prompts are the text instructions you give Google's Gemini image model to generate or edit a picture. Nano Banana is the nickname the model picked up before its formal name stuck, and it is still the term most people search for. If you want the background on the model itself, we covered it in what is Gemini Nano Banana.

What separates a working prompt from a wasted generation is decisions, not adjectives. Beautiful and high quality tell the model nothing it can act on. A strong prompt fills four slots:

• Subject - what is in frame, with one or two identifying details.
• Scene and composition - where it is, the camera position, and where the empty space sits.
• Style and medium - photograph, oil painting, 3D render, woodblock print, and the era or tradition behind it.
• Technical details - lighting direction and hardness, lens, palette, aspect ratio. Lighting shifts the result more than anything else here.

Naming a movement, era or physical medium gets you a far more coherent look than asking for artistic. Describing the material the image is supposedly made of - canvas weave, woodgrain, grout lines - is what stops an AI art generator from defaulting to generic digital illustration.

1. Victorian animal portrait

A stoic grey wolf wearing a silk cravat and a monocle, seated in a high-backed leather chair, posed as a formal Victorian oil portrait. Dark umber background with a single warm window light from the left. Visible canvas weave and fine craquelure, muted palette of charcoal, oxblood and aged gold, head-and-shoulders framing, painterly detail in the fur.

2. Impressionist street cafe

A bustling Parisian street cafe at sunset, figures suggested rather than defined, wet pavement reflecting warm lamplight. Loose visible brushstrokes and broken colour in the late nineteenth-century Impressionist manner, palette of apricot, lilac and slate blue, no hard outlines anywhere, slightly elevated eye level looking down the street.

3. Baroque still life

A single peeled banana resting on a crushed crimson velvet cushion beside a tarnished silver goblet and a wilting tulip. Dramatic Baroque chiaroscuro with one candle as the only light source, deep black background with shadow detail retained, heavy fabric texture and sharp specular highlights on the metal, square composition.

4. Ukiyo-e woodblock print

A cat sitting beneath a blossoming cherry tree beside a still river, Mount Fuji small in the distance. Flat unmodulated colour fields, bold black outlines, visible woodgrain texture in the sky, traditional ukiyo-e palette of indigo, pale pink and cream, vertical composition with an empty band across the top.

5. Cyberpunk neon street

A sleek black motorbike racing through a rain-soaked futuristic street at night, motion blur on the background, magenta and cyan signage reflected across the wet asphalt. Dense layered signage in the mid-ground, low camera angle at wheel height, cinematic 2.35:1 crop, deep shadows, high colour saturation.

6. Art Nouveau poster

A stylised woman in profile surrounded by flowing lily stems and elaborate curved linework, framed by a decorative arched border. Flat colour with heavy ornamental outline in the Art Nouveau poster tradition, palette of sage, ochre and deep plum, vertical poster format, symmetrical, generous margin around the frame.

7. Low-poly 3D world

A smiling cartoon dog standing in a simplified angular landscape of faceted hills and triangular trees. Low-poly 3D render with flat shading and visible polygon edges, bright saturated palette of grass green, sky blue and warm yellow, soft ambient light with no hard shadows, three-quarter isometric-leaning perspective.

8. Sumi-e ink wash

A small fishing boat on a foggy lake with reeds in the foreground, most of the frame left as empty paper. Sumi-e ink wash on rice paper, gestural strokes with visible bleed and dry-brush texture, monochrome black ink with one faint grey wash for the mist, vertical composition, extreme negative space.

9. Stencil street art

A child releasing a kite that has become a flock of birds, sprayed onto a weathered concrete wall. Two-layer stencil technique with hard edges and slight overspray, black stencil with one red accent element, visible wall texture with cracks and old paint showing through, straight-on eye-level view.

10. Mosaic tile portrait

A portrait of a dog built entirely from small irregular glass tiles in varied tones, grout lines clearly visible between every tessera. Byzantine-style mosaic on a gold tile background, warm amber, cream and slate palette, front-facing composition, raking light so individual tiles catch highlights.

If a style prompt comes back looking like generic digital art, the fix is almost always to name the physical surface: paper grain, canvas weave, grout, woodgrain, print registration.

Lighting is the single highest-leverage line in any prompt. State the direction, the hardness and whether there is fill, and an AI image generator stops guessing. These ten are written as lighting and lens exercises, so swap in any subject you like.

11. Volumetric god rays

A single shaft of light cutting through a dark pine forest, dust motes and pollen suspended and clearly visible in the beam. Heavy atmospheric haze, near-black shadow everywhere outside the beam, cool green-grey palette with one warm highlight, vertical composition, camera low among the ferns.

12. Rim-lit silhouette

A figure standing in a doorway, defined only by a hard bright outline along the shoulders and hair, face and body left in near-total shadow. Strong rim light from directly behind with no fill at all, deep blue-black background, three-quarter body framing, high contrast, almost no detail in the shadows.

13. Softbox product lighting

A high-tech running shoe on a seamless pale grey backdrop, evenly lit with no harsh shadows. One large softbox above and slightly forward plus a white bounce card below for shadow fill, crisp product edges, clear material separation between mesh, rubber and knit, three-quarter hero angle, catalogue photography.

14. Chiaroscuro portrait

A face turned three-quarters to camera, half in bright light and half falling into deep shadow, the shadow terminator running down the bridge of the nose. Single hard light from high left with no fill, black background, tight head crop, rich skin texture, seventeenth-century chiaroscuro.

15. Underlighting

A face illuminated only from below by a low flickering source, exaggerated upward shadows across the cheekbones and brow, eyes lost in shadow. Sickly green-amber light against a black surround, slightly tilted camera, close crop, unsettling atmosphere without anything graphic.

16. Bokeh shallow focus

A single red rose in sharp focus with the background dissolved into large soft circular bokeh from string lights. 85mm lens at f/1.4 with the focal plane on the front petal edge only, warm amber bokeh against deep green, horizontal composition with the subject offset left.

17. Tilt-shift miniature

A busy harbour photographed so it reads as a miniature model set, a sharp band of focus across the middle with strong blur above and below. Tilt-shift effect with slightly boosted saturation and contrast, high vantage point looking down at roughly 45 degrees, crisp tiny detail inside the focus band.

18. Long exposure light trails

A nighttime bridge with car headlights drawn into continuous smooth streaks and the water below rendered glassy and featureless. 30-second exposure look, tripod-steady with no motion in the architecture, deep blue-hour sky retaining a gradient, wide composition using the trails as leading lines.

19. Caustic water light

Sunlight refracting through moving water and projecting rippling caustic patterns across a pale tiled pool floor, one pair of feet entering at the top of the frame. Shot from underwater looking up and forward, bright cyan and white palette, high clarity, square crop with the light pattern as the real subject.

20. Drone aerial in autumn

A straight-down aerial view of mixed autumn forest, the canopy forming a mosaic of crimson, ochre and remaining green with one narrow river cutting a pale line through it. Midday light for even exposure and minimal shadow, no horizon in frame, high-resolution flat-plane composition, natural saturation.

A useful habit: write the subject line once, then generate it four times changing only the lighting sentence. It teaches you more about the model in five minutes than reading a guide does.

These lean on Nano Banana's willingness to take an odd premise literally. The rule that matters here: one impossible element per prompt. Stack two or three and a text to image model tends to average them into mush.

21. Samurai on the moon

A chrome-plated samurai standing on grey lunar regolith, traditional layered armour fused with a mirrored sci-fi helmet, Earth small on the horizon behind. Hard unfiltered sunlight with a pitch-black sky and no atmospheric scatter, one long sharp shadow across the dust, full-body low-angle hero shot, cinematic concept art.

22. Detective cat

A tough-looking British Shorthair in a rumpled trench coat and fedora, seated behind a desk with a magnifying glass and scattered case files. Film noir lighting through venetian blinds casting hard stripes across the scene, monochrome with a faint amber cast, waist-up framing, smoke drifting through the light.

23. Crystal fox

A fox composed entirely of glowing multicoloured quartz crystal, mid-stride through fresh snow, its internal light making the snow beneath it glow. Hyper-detailed refraction and internal reflection in every facet, cool blue snow against warm prismatic light, side-profile action framing, shallow depth of field.

24. Bonsai whale

A full-size humpback whale floating serenely above a small bonsai garden on a wooden table, the scale deliberately impossible. Soft even daylight from a window with no dramatic shadow, muted sage, cedar and grey-blue palette, wide static composition, surrealist calm rather than spectacle.

25. Deep sea jellyfish city

A bioluminescent city built inside the translucent bell of an enormous ancient jellyfish, tiny lit windows visible through the membrane, tendrils trailing down into black water. The city lights as the only light source, deep blue-black surround with teal and violet glow, wide concept-art composition.

26. Tree of knowledge

An enormous ancient oak with hundreds of open books suspended among its branches, pages faintly glowing, a spiral staircase wound around the trunk. Late-afternoon light filtering through the canopy, warm gold against deep green shadow, low-angle view from the base looking up, magical realism.

27. The weight of time

A colossal hourglass slowly crushing the peak of a stone mountain, sand pouring upward instead of down. Overcast diffuse light with no clear sun direction, desaturated palette of slate, bone and dull ochre, wide static composition with the hourglass off-centre, quiet surrealism rather than drama.

28. Digital serenity

A person sitting cross-legged on a beach where the sand is flowing luminous data and the sea is scrolling code. Soft dawn light plus emissive glow from the ground, palette of deep navy, cyan and warm skin tones, wide composition with the figure small in frame, tranquil sci-fi.

29. City on a turtle

An entire walled metropolis built on the shell of an enormous sea turtle wading through shallow ocean, tiny boats around its feet. Golden-hour side light with long shadows across the water, warm amber city against cool blue sea, wide low-angle composition, fantasy matte painting.

30. Crayon ocean

An ocean made of melted vibrant crayon wax caught mid-flow, tiny folded paper boats sailing on the ridges, wax texture and colour bleeding visible at every wave. Macro-scale shot so the wax reads as material rather than water, saturated primary palette, shallow depth of field, playful surrealism.

Notice that each of these still specifies lighting and framing. The concept is the hook, but the technical lines are what stop the output looking like a stock illustration of the idea.

Legible in-image text is where Nano Banana pulls ahead of most rivals, and Nano Banana Pro further still. Three things make it work: put the exact string in capitals, name the typeface class rather than a specific font, and say where in the frame it sits. The same discipline applies when you use an AI logo generator for a wordmark.

31. Minimalist coffee logo

A clean circular logo for a coffee brand on an off-white background. Render the text THE DAILY GRIND in a bold geometric sans-serif, letter-spaced, arranged around the inside of the circle, with a simple bean glyph centred. Two colours only: espresso brown and cream. Flat vector look, no gradients, no drop shadow, text perfectly legible and correctly spelled.

32. Gothic film title

A stone gargoyle perched on a rain-slick rooftop at night. Render the word MIDNIGHT across the lower third in ornate weathered blackletter, pale grey against the dark sky. Hard blue moonlight from behind the gargoyle, heavy negative space above the text, vertical poster composition, lettering sharp and correctly spelled.

33. Neon window sign

A photorealistic vintage neon sign mounted on a dark brick wall. Render the word OPEN in flowing script neon tubing, electric blue with a faint pink outer glow and visible mounting hardware. Night scene with the sign as the only light source, glow spilling onto the brick, uneven tube brightness suggesting a flicker, straight-on framing.

34. Chalkboard menu

A rustic black chalkboard in a cafe. Render the heading LATTE ART in flowing white chalk script with three short menu lines beneath in smaller hand-lettered capitals. Visible chalk dust, smudges and uneven pressure, warm side light from a window, wooden frame in shot, all text legible with accurate spelling and no invented characters.

35. Retro travel poster

A mid-century travel poster of a futuristic ship descending into an ancient redwood forest. Render PLANET X in a wide condensed serif across the top and VISIT THE OUTER RING in small caps at the bottom. Flat screen-print colour separation with slight registration offset, limited palette of teal, cream and burnt orange, vertical composition, crisp text.

36. Vaporwave grid poster

A digital poster with a receding wireframe grid and floating chrome geometric solids. Render AESTHETICS 1989 in a wide italic sans-serif with a chromatic-aberration edge, centred on the horizon line. Palette limited to hot pink, teal and black, sun as a simple gradient disc, symmetrical 4:5 composition, text perfectly spelled.

37. Product packaging label

A matte black cylindrical tin standing on grey concrete. Render the label text COLD BREW No. 4 in a small refined serif, centred, with a thin rule above and below, plus 250 ML in tiny caps at the base. One soft light from upper left with gentle falloff, the label wrapping naturally with the cylinder curve, product photography, all lettering sharp and correct.

38. Weather infographic card

A clean weather card graphic showing stylised rain over a simplified city outline. Render 30% CHANCE OF RAIN in a medium-weight geometric sans-serif, left-aligned, with the numeral noticeably larger than the words. Flat two-tone palette of deep slate and pale sky blue, generous padding, 16:9 composition, text legible at small size.

39. Book cover typography

A matte hardback book cover sitting on a plain surface. Render the title THE LONG QUIET in a large high-contrast serif across the upper half and A NOVEL in small letter-spaced caps beneath, separated by one thin horizontal rule. Cream stock with deep green ink, a soft raking light showing the paper tooth, straight-on cover view.

40. Storefront awning

A narrow city storefront with a striped fabric awning. Render BLOOM AND CO in painted script across the awning valance and FLORIST in small caps on the window glass below. Overcast diffuse daylight with no harsh shadow, palette of dusty rose, cream and dark green, straight-on eye-level composition, both text elements sharp and correctly spelled.

Keep strings short. Accuracy drops off quickly past about four words, so for anything longer than a headline it is still safer to generate the image clean and set the type yourself afterwards.

These start from an existing picture rather than a blank canvas, which is where Nano Banana is genuinely useful rather than merely competent. The pattern to copy: state what must stay identical before you state what changes. Working image to image, most bad edits come from the model rebuilding something you never asked it to touch.

41. Character consistency across scenes

Using the character in the reference image, generate the same person standing at a rain-lit bus stop at night. Keep the face, hair length, build and clothing exactly as they appear in the reference. Change only the environment and the lighting, which should be hard blue-white from a single overhead lamp.

42. Day to night conversion

Take the supplied afternoon photograph and re-light it as a late-night scene. Keep every object, position and framing identical. Replace the daylight with cool moonlight plus one warm sodium streetlamp from the left, add practical window lights in the background buildings, and deepen the shadows without losing detail inside them.

43. Seasonal change

Convert the supplied summer park scene to deep winter. Keep the composition, the paths and the buildings unchanged. Replace the foliage with bare branches, add fresh untouched snow on every horizontal surface, cool the white balance, and flatten the light to overcast diffuse with no visible sun.

44. Background replacement

Keep the subject in the supplied image exactly as is, including the edge detail in the hair. Replace the cluttered background with a seamless mid-grey studio backdrop and adjust the subject's edge lighting to match a single softbox from the upper left. Do not alter the pose, skin tone or clothing.

45. Outfit change

In the supplied portrait, change the subject's casual t-shirt to a well-tailored charcoal linen blazer over a white shirt, with natural drape and a correct shoulder fit. Keep the face, pose, hair and background completely unchanged, and match the fabric lighting to the existing key light direction.

46. Aspect ratio extension

Extend the supplied 16:9 image outward to a 9:16 vertical frame. Generate plausible new content above and below that continues the existing scene, matching grain, colour grade and perspective exactly. Do not alter or rescale the original area, and keep the horizon line consistent across the new region.

47. Object removal

Remove the parked car and the roadside signage from the supplied street photograph. Reconstruct the road surface, kerb and wall behind them so the repair is undetectable, matching the existing texture, shadow direction and colour. Leave every other element, including the pedestrians, exactly as it is.

48. Focus adjustment

Take the supplied group photograph and reduce the depth of field. Keep the second person from the left in sharp focus and blur everything else progressively by distance, with natural circular bokeh. Match the blur falloff to an 85mm lens at f/1.8 and preserve the original colour grade.

49. Style transfer with structure kept

Re-render the supplied photograph of a quiet street as a thick impasto oil painting: heavy visible brushwork, slightly exaggerated colour, softened edges. Keep the composition, perspective and every structural element in exactly the same position. Do not add or remove any objects.

50. Product mockup blend

Place the supplied matte black product box onto a polished marble surface in a studio setting. Match the box's perspective to a slightly elevated three-quarter camera, generate a correct contact shadow and a soft reflection in the marble, and light it with one large softbox from the upper left. Keep the box artwork and proportions untouched.

• Change one variable at a time. Rewrite the lighting, the style and the lens together and you learn nothing about which change helped.
• Lead with the subject. Opening words carry more weight, so put what the image is of first and technical settings last.
• Cap the palette. Naming two or three colours produces far more cohesive images than asking for vibrant, which fills the frame with everything at once.
• Say what must not change. For any edit, list the protected elements before the change. It is the single biggest difference between a clean edit and a redraw.
• Reference a tradition, not a living artist. Naming a movement, era or medium gets you the look without copying a specific person's style.

Chat Smith gives you Nano Banana alongside other image models in one place, which matters because the same prompt lands differently on each. Nano Banana Pro is the one to reach for when the image contains text; the others often read better for painterly styles. Running a prompt through two and comparing takes a minute and usually settles the question faster than another rewrite.

Because it is a chat interface first, you can also draft the prompt before spending a generation on it. Paste one of the blockquotes above, ask for three variants with different lighting, then generate only the version you like. For a consistent set, keep the conversation open and change one line per message so the chat history doubles as your prompt library. If the underlying technology is new to you, our explainer on generative AI is a good starting point.