AI Image Prompts: 40 Copy-Ready Prompts for Seedream, Nano Banana and GPT Image

40 copy-ready AI image prompts by style, with real side-by-side outputs from Seedream 5 Pro and Nano Banana 2 on Omnely.

AI Image Prompts: 40 Copy-Ready Prompts for Seedream, Nano Banana and GPT Image title card

Great ai image prompts are built like recipes: subject, setting, style, light, composition, and a short list of constraints. The 40 prompts below are organized by visual style, with eight of them run once each on Seedream 5 Pro and Nano Banana 2 at 1K so the behavior differences are visible rather than theoretical. The other 32 are copy-ready templates in the same format, ready to try on Seedream 5 Pro, Nano Banana 2, or GPT Image 2 inside Omnely.

Product photo of a ribbed glass perfume bottle with an amber liquid on a wet black slate slab, a few water droplets, single hard light from the upper left creating a long reflection, dark moody background, sharp focus, no text, no logo

product style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro kept the bottle smaller and the scene darker. Single example, 1:1 at 1K, generated on Omnely.*

product style image generated by Nano Banana 2 on Omnely

*Nano Banana 2 framed the bottle larger with a black cap and a stronger reflection. Single example, 1:1 at 1K, generated on Omnely.*

A model-agnostic prompt formula

Most image models, including Seedream 5 Pro, Nano Banana 2, and GPT Image 2, respond well to a plain-language description assembled from the same core blocks. The order matters less than the coverage:

  • Subject — what is in the frame and what it is doing or how it is arranged.
  • Setting — environment, surface, background, time of day.
  • Style — one visual direction (watercolor, cel-shaded anime, flat vector, cinematic photo, 3D clay, art deco poster, food photography, product photography).
  • Light — direction, hardness, color temperature, and any reflections or shadows.
  • Composition — framing, shot size, lens feel, depth of field, and subject placement.
  • Constraints — things to avoid, such as *no text*, *no logo*, *no people*, or a specific palette size.
  • Required text — if the image needs words, put them in quotation marks and spell them exactly.

Write one clear sentence or two. Avoid stacking three overlapping styles. Change one variable at a time when iterating.

Eight styles tested on Seedream 5 Pro and Nano Banana 2

Each prompt below was run once at 1K on each model. These are single examples, not a benchmark — a second run can look different. Both models are available inside Omnely using shared credits, and new users get free credits to try them.

1. Product photography

Product photo of a ribbed glass perfume bottle with an amber liquid on a wet black slate slab, a few water droplets, single hard light from the upper left creating a long reflection, dark moody background, sharp focus, no text, no logo

Both outputs rendered a clean amber ribbed bottle on wet slate with a single hard key light. In this single run, Seedream 5 Pro produced a smaller bottle in a darker overall scene, while Nano Banana 2 made the bottle larger, added a black cap, and gave a stronger reflection on the slate. Both respected *no text* and *no logo*.

2. Loose watercolor

Loose watercolor painting of a red fox curled asleep under a birch tree in fresh snow, soft wet-on-wet washes, white paper showing through, muted blues and warm orange, visible brush texture, no text

watercolor style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro used softer, airier washes. Single example, 1:1 at 1K, generated on Omnely.*

watercolor style image generated by Nano Banana 2 on Omnely

*Nano Banana 2 added more detail and background trees. Single example, 1:1 at 1K, generated on Omnely.*

Both models placed a sleeping fox under a birch and preserved visible brush texture. Seedream 5 Pro leaned softer and airier with more exposed paper. Nano Banana 2 added more background detail, including additional trees, which read as a denser scene.

3. Cel-shaded anime

Cel-shaded anime film still of a small red two-car train crossing a long seaside bridge at sunset, big puffy clouds, glittering sea, warm orange and teal palette, clean line art, no text, no characters

anime style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro went bold teal and orange with a straight bridge. Single example, 1:1 at 1K, generated on Omnely.*

anime style image generated by Nano Banana 2 on Omnely

*Nano Banana 2 drew an arched bridge with coastal cliffs and purple clouds. Single example, 1:1 at 1K, generated on Omnely.*

Both followed the red train on a sea bridge at sunset with no visible characters. Seedream 5 Pro held a bolder teal-and-orange palette with a straight bridge. Nano Banana 2 reinterpreted the bridge as arched with coastal cliffs and pushed the clouds toward purple, which changed the mood of the single output.

4. 3D clay render

Soft clay 3D render of a cheerful little cactus in a terracotta pot wearing tiny round sunglasses, pastel pink background, rounded shapes, subtle fingerprint texture, soft studio light, toy-like, no text

3d clay style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro added cartoon eyes and a smile that the prompt did not ask for. Single example, 1:1 at 1K, generated on Omnely.*

3d clay style image generated by Nano Banana 2 on Omnely

*Nano Banana 2 kept a plainer cactus with round sunglasses. Single example, 1:1 at 1K, generated on Omnely.*

Both kept the terracotta pot, pastel pink background, and round sunglasses. Seedream 5 Pro added cartoon eyes and a smile to the cactus that the prompt did not request. Nano Banana 2 produced a plainer cactus with just the sunglasses, closer to the literal description.

5. Art deco poster with required text

Art deco travel poster of a palm-lined city gate at golden hour with geometric sun rays, limited palette of terracotta, teal and cream, bold sans-serif letters at the top read "VISIT MARRAKECH", flat print texture, spelling exactly as written

art deco poster style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro spelled VISIT MARRAKECH correctly above the gate. Single example, 1:1 at 1K, generated on Omnely.*

art deco poster style image generated by Nano Banana 2 on Omnely

*Nano Banana 2 spelled VISIT MARRAKECH correctly in a bold header. Single example, 1:1 at 1K, generated on Omnely.*

Both models spelled *VISIT MARRAKECH* correctly in this run. Seedream 5 Pro placed the title inside an arch above a palm-lined gate. Nano Banana 2 used a bigger, bolder header and emphasized the geometric sunburst rays. Text rendering can still fail on any model, so verifying spelling in each output is part of the workflow.

6. Cinematic portrait

Cinematic portrait of a generated, fictional middle-aged man in a yellow raincoat standing under a neon-lit awning on a rainy night street, teal and magenta reflections on wet pavement, 50mm lens look, shallow depth of field, film grain, no text

cinematic portrait style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro: a generated, fictional man in close framing, with no text. Single example, 1:1 at 1K, generated on Omnely.*

cinematic portrait style image generated by Nano Banana 2 on Omnely

*Nano Banana 2: a generated, fictional man in a wider street scene; it added neon sign text despite "no text". Single example, 1:1 at 1K, generated on Omnely.*

Seedream 5 Pro held a tighter portrait framing with the hood up and respected *no text*. Nano Banana 2 opened the frame into a wider street scene and added neon signs reading *NIGHT OWL ARCADE* and *NOODLE BAR* despite the constraint — a useful reminder that *no text* is not always obeyed and should be checked.

7. Flat vector

Flat vector illustration of a small wooden cabin by a mountain lake with pine trees and a rising moon, limited palette of five colors, clean geometric shapes, no gradients, no outlines, no text

flat vector style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro stayed very minimal with about five colors. Single example, 1:1 at 1K, generated on Omnely.*

flat vector style image generated by Nano Banana 2 on Omnely

*Nano Banana 2 used more than five colors and more detail. Single example, 1:1 at 1K, generated on Omnely.*

Seedream 5 Pro held a very minimal composition with roughly five colors, closer to the brief. Nano Banana 2 produced a richer illustration with more than five colors and extra scene detail, drifting from the palette constraint in this single run.

8. Food photography

Close-up food photo of a tall stack of blueberry pancakes with maple syrup dripping down the side and a pat of melting butter on top, fresh blueberries scattered on a white plate, soft morning window light, shallow depth of field, no text

food style image generated by Seedream 5 Pro on Omnely

*Seedream 5 Pro centered the stack on a white plate. Single example, 1:1 at 1K, generated on Omnely.*

food style image generated by Nano Banana 2 on Omnely

*Nano Banana 2 added a fork, a coffee cup and a wood table. Single example, 1:1 at 1K, generated on Omnely.*

Both produced tall blueberry pancake stacks with dripping syrup and a pat of butter. Seedream 5 Pro centered the stack tightly on a white plate. Nano Banana 2 added a fork, a coffee cup, and a wood table, treating the shot as a wider breakfast scene.

32 more ai image prompts by style

The templates below follow the same subject–setting–style–light–composition–constraints pattern. They were not run for this article; use them as starting points and iterate one variable at a time. All of them work as image prompts for Seedream 5 Pro, Nano Banana 2, or GPT Image 2 inside Omnely.

Product

Product photo of a matte black ceramic coffee mug on a pale oak table beside roasted beans, soft window light from the right, gentle steam, shallow depth of field, clean minimal background, no text, no logo
Product photo of white leather sneakers floating against a pastel blue paper backdrop, soft shadow below, even studio light, straight-on catalog angle, crisp focus, no text, no logo
Product photo of a stainless steel wristwatch on a folded navy linen cloth, dramatic single light from the upper right, deep shadows, dark background, macro detail on the dial, no text, no logo
Product photo of a glass jar of golden honey with a wooden dipper above it, drips of honey mid-air, warm soft light, light cream background, shallow depth of field, no text, no logo

Watercolor

Loose watercolor painting of a Tuscan hilltop village at dawn, cypress trees along a winding road, soft washes of ochre and sage, visible paper grain, white edges, no text
Loose watercolor painting of a rainy Parisian cafe terrace with yellow awnings, wet reflections on cobblestones, muted palette, blurred figures, white paper showing through, no text
Loose watercolor painting of a hummingbird hovering by a pink magnolia branch, delicate wet-on-wet washes, soft pastel background, visible brush texture, no text
Loose watercolor painting of a quiet fishing harbor at sunset, small wooden boats, warm orange and dusty blue washes, loose linework, paper grain visible, no text

Anime

Cel-shaded anime film still of a quiet countryside bus stop in summer, cicadas implied, bright blue sky with towering cumulus clouds, warm palette, clean line art, no text, no characters
Cel-shaded anime film still of a cozy bedroom window at night with a cat curled on the sill, warm lamp light inside, cool blue city lights outside, soft shading, no text, no characters
Cel-shaded anime film still of a cherry blossom path along a canal in spring, falling petals, soft pink and teal palette, clean line art, no text, no characters
Cel-shaded anime film still of a rainy convenience store at night from across the street, glowing white signage, wet reflections, cool blue and warm yellow contrast, clean line art, no text, no characters

3D clay

Soft clay 3D render of a smiling avocado character sitting on a lemon-yellow background, rounded shapes, subtle fingerprint texture, soft studio light, toy-like, no text
Soft clay 3D render of a tiny camper van parked beside a pine tree on a pastel green hill, rounded wheels and windows, soft shadows, cheerful palette, toy-like, no text
Soft clay 3D render of a round owl perched on a crescent moon against a lavender background, soft studio light, rounded shapes, matte surface, toy-like, no text
Soft clay 3D render of a stack of three pastel donuts with sprinkles on a mint background, rounded shapes, soft shadow, subtle texture, toy-like, no text

Art deco poster

Art deco travel poster of a streamlined ocean liner at sunset with geometric waves, limited palette of navy, gold and cream, bold sans-serif letters at the top read "ATLANTIC CROSSING", flat print texture, spelling exactly as written
Art deco travel poster of a mountain railway winding through snow peaks, geometric fir trees, limited palette of teal, white and bronze, bold letters at the bottom read "ALPINE EXPRESS", flat print texture, spelling exactly as written
Art deco travel poster of a tropical lighthouse at dawn with geometric sun rays, limited palette of coral, cream and deep blue, bold letters at the top read "CAPE HORIZON", flat print texture, spelling exactly as written
Art deco travel poster of a city skyline with a stylized zeppelin, geometric clouds, limited palette of maroon, gold and ivory, bold letters at the bottom read "VISIT CHICAGO", flat print texture, spelling exactly as written

Cinematic portrait

Cinematic portrait of a generated, fictional elderly fisherwoman in a blue oilskin jacket on a windy dock at dawn, cool overcast light, 85mm lens look, shallow depth of field, film grain, no text
Cinematic portrait of a generated, fictional young barista in an apron behind a wooden counter, warm tungsten lights, soft bokeh of shelves behind, 50mm lens look, film grain, no text
Cinematic portrait of a generated, fictional middle-aged scientist in a white coat beside softly glowing lab glassware, moody teal rim light, shallow depth of field, 85mm lens look, film grain, no text
Cinematic portrait of a generated, fictional street musician under warm string lights at dusk, blurred crowd in the background, 35mm lens look, shallow depth of field, film grain, no text

Flat vector

Flat vector illustration of a bicycle leaning against a bakery window with a basket of bread, limited palette of five colors, clean geometric shapes, no gradients, no outlines, no text
Flat vector illustration of a hot air balloon drifting over rolling hills at sunrise, limited palette of five colors, clean geometric shapes, no gradients, no outlines, no text
Flat vector illustration of a desk with a laptop, plant and coffee cup seen from above, limited palette of five colors, clean geometric shapes, no gradients, no outlines, no text
Flat vector illustration of a lighthouse on a rocky cliff above calm waves, limited palette of five colors, clean geometric shapes, no gradients, no outlines, no text

Food

Close-up food photo of a cast iron pan with sizzling garlic butter shrimp and lemon wedges, steam rising, warm overhead light, dark wooden table, shallow depth of field, no text
Close-up food photo of a bowl of ramen with a soft-boiled egg, scallions and nori, chopsticks lifting noodles, warm side light, dark ceramic bowl on a stone table, shallow depth of field, no text
Close-up food photo of a slice of lemon tart with torched meringue on a white plate, fresh mint leaf, soft daylight from the left, pale marble surface, shallow depth of field, no text
Close-up food photo of a rustic sourdough loaf, cut open to show the crumb, flour dusted on a dark wooden board, warm window light, shallow depth of field, no text

Short tips for better outputs

  • Text in images — always wrap required words in quotation marks and spell them exactly. State capitalization if it matters. Leave room in the composition for the letters.
  • "No text" is a request, not a guarantee — as the cinematic portrait example showed, models sometimes add signage anyway. Check every output and re-run if needed.
  • Reference images — Seedream 5 Pro accepts up to ten reference images and is useful when a specific object, pose, or color needs to carry across generations. Nano Banana 2 and GPT Image 2 also support references in the Omnely app.
  • Iterate one variable at a time — change the lighting, or the palette, or the camera angle, but not all three at once, so it is clear which change caused which difference.
  • Trim the prompt — if two phrases conflict (e.g. "no outlines" and "clean line art"), pick one. Overlong prompts tend to drift.

For ChatGPT-specific prompt phrasing, see ChatGPT image prompts. If the goal is to transform a photo rather than generate from scratch, see AI prompts for photos.

FAQ

Do the same prompts work across Seedream, Nano Banana, and GPT Image?

Mostly yes — a well-structured prompt covering subject, setting, style, light, composition, and constraints transfers between them. The outputs will differ in framing, palette interpretation, and text rendering, as the eight side-by-side examples above show. Pick the model whose single-run behavior matches the brief best, then iterate from there.

How to get legible text in an image

Put the exact text in quotation marks, note capitalization, place the text instruction near the end of the prompt, and reserve clear space in the composition (for example, "bold sans-serif letters at the top read…"). Verify spelling in each output; regenerate if a letter is wrong rather than trying to fix it in prose.

Which Omnely image model is a good starting point?

For a tight, literal interpretation of the prompt, Seedream 5 Pro is a strong starting point and accepts up to ten reference images. For richer, scene-building outputs that often add context, try Nano Banana 2. For projects involving posters, UI mockups, or text-forward layouts, GPT Image 2 is worth a comparison run. All three live in one workspace on Omnely, so the same prompt can be tested against each.

Copy the product prompt at the top of this article and run it on Seedream 5 Pro inside Omnely to see how the ribbed glass bottle, wet slate, and hard upper-left light come together in a single frame.

For prompts organized around physical media, see AI art prompts. To animate a finished image, use the AI video prompts guide; the AI prompts overview also covers chat workflows.