Composition, light, lenses, text that spells right and edits that keep what you love, for GPT Image 2.5, Nano Banana 2 and Flux 3. With 12 copy-ready templates.
Product hero shot of a matte green glass perfume bottle on wet black stone, soft rim light from behind, 85mm, shallow depth of field, label reads “VERDE” in a thin serif, no other text, 4:5.
Describe one moment, not a category: the subject with its materials and details, the framing and aspect ratio, the setting, the light source and direction, a lens or medium, and the style. Put any text in quotation marks with the font and placement, keep it short, and rule out extra text and logos. When editing, change one thing per turn and say what must stay the same. GPT Image 2.5 is strongest with text and layouts, Nano Banana 2 with edits and consistency, Flux 3 with natural, descriptive sentences.
Pick a medium, describe your idea, tap one option per row. Every part of the prompt is color-coded, scored, and matched to the model that handles it best.
Close-up of a ceramic coffee cup on a marble counter, 35mm lens, soft three-point studio light, photorealistic, crisp detail, square 1:1 format.
Specific materials, framing and light turn a generic picture into a usable one.
Choose the aspect ratio for where the image will live, then frame inside it. Use the ratio picker, or name it at the end of your prompt.
Say where it comes from and what color it is. These phrases work across models.
A lens or medium tells the model how to render depth, texture and style.
Where most image prompts go wrong, and how to avoid it.
A minimalist café poster with a single coffee cup illustration. The headline reads “BREW BETTER” in an elegant serif at the top center, once. Cream background, plenty of space. No other text. 4:5.
Change only the jacket to the leather jacket from image 2. Keep the person, pose, face, background and lighting exactly the same.
Pick by the job, then use its habits.
Best for: posters, packaging, infographics, UI mockups and anything with text.
Product photo of a kraft-paper coffee bag standing on a light oak counter, morning window light from the left, steam from a cup beside it. The bag label reads “MORNING BLEND” in a bold condensed sans-serif, centered, once. Clean, warm, commercial. No other text, logos or watermarks. 4:5.
Best for: photo edits, combining several images, and keeping a character or product consistent.
Place this exact character from the reference in a new scene: walking along a rainy Tokyo street at night, neon signs reflected in puddles. Keep the face, hair and jacket identical. Vertical 9:16.
Best for: photoreal scenes and clean commercial images from natural language.
A close-up photograph of a glass of iced coffee on a sunlit café table, condensation running down the glass, a striped paper straw, soft morning light from behind making the ice glow, shallow depth of field, a small card next to it that reads “Cold Brew”.
Best for: artistic looks, illustration, mood boards and style exploration.
Three common mistakes, and the fix.
beautiful perfume bottle, aesthetic, high quality, 4k
No material, setting, light or framing. You’ll get a generic stock look.Product hero shot of a faceted green glass perfume bottle with a gold cap on wet black stone, soft rim light from behind, droplets on the glass, 85mm, shallow depth of field, dark luxurious mood. Vertical 4:5, no text.
poster that says summer sale 50 percent off with nice fonts
Copy isn’t quoted, font and placement are vague, and nothing rules out extra text.Bright summer sale poster with a lemon-yellow background and a single sliced lemon at the bottom. Headline reads “SUMMER SALE” in a bold rounded sans-serif at the top, and “50% OFF” larger below it. No other text. 4:5.
make the background better and the person nicer and change the colors
Three vague edits at once, and nothing protected.Change only the background to a softly blurred city park in autumn. Keep the person, pose, clothes and face exactly the same.
Swap in your product or subject and run them on any image model on Creatide.ai.
Product hero shot of a white sneaker floating at a slight angle against a solid pastel orange background, soft studio light from above, a gentle shadow below, crisp detail on the laces and stitching, minimal and clean. Square 1:1, no text.
A woman’s hand holding a matte black water bottle on a mountain trail at sunrise, warm backlight, mist in the valley behind, shallow depth of field, natural colors, authentic outdoor lifestyle photo. Vertical 4:5.
Minimalist concert poster: a single neon pink circle glowing over a dark blue city skyline. Headline at the top reads “NIGHT SIGNALS” in a bold condensed sans-serif. Below it, smaller: “LIVE · NOV 14”. Plenty of empty space, grainy print texture. No other text. 2:3.
Top-down flat lay of a rustic breakfast table: avocado toast, a soft-boiled egg, fresh berries and black coffee on linen, soft window light from the left, natural colors, editorial food photography. Square 1:1.
Close-up portrait of an elderly fisherman with a weathered face and a knitted wool cap, looking past the camera, overcast seaside light, 85mm lens, shallow depth of field, muted documentary color.
Wide shot of a Japandi-style living room with a low oak sofa, linen cushions and a paper floor lamp, late afternoon light through sheer curtains, calm neutral palette, architectural photography. 16:9.
Character sheet of the same young explorer in three views: front, side and back, on a plain light grey background, consistent outfit (olive jacket, red scarf, brown boots), clean flat lighting, stylized 3D look. Labels under each view: “FRONT”, “SIDE”, “BACK”.
Change only the background to a snowy mountain cabin porch at dusk with warm lights. Keep the person, their pose, clothing and the lighting on their face exactly the same.
A flat vector logo concept for a coffee roastery called “Ember & Oak”: a simple flame shape formed from an oak leaf, in deep orange on cream, with the name below in a warm serif. Centered, plenty of margin, no other text.
A clean infographic titled “How Cold Brew Works” with four numbered steps in a horizontal row: grind, steep 12 hours, filter, serve over ice. Simple line icons, beige and brown palette, readable sans-serif labels. Landscape 16:9. No other text.
A floating island with a waterfall pouring into the clouds below, ancient stone ruins on top, golden-hour light breaking through, birds in the distance, painterly concept-art style, epic scale. Wide 21:9.
YouTube thumbnail: a surprised man pointing at a giant glowing smartphone on the right, bright yellow background, bold outline around him. Big text on the left reads “I TRIED IT” in heavy white sans-serif with a black stroke. 16:9, no other text.
Straight answers to the questions people search most. Checked on October 21, 2026.
Describe one specific moment: subject with materials and details, framing and aspect ratio, setting, light source and direction, lens or medium, and style. Add any text in quotes and finish with constraints.
Quote the exact words, name the font style and placement, keep the copy short, and close with “no other text.” GPT Image 2.5 is the most reliable for text-heavy layouts.
Change one thing per turn and say what must stay: “Change only the sky to sunset. Keep the house and lighting identical.” Feed the approved result back in for the next change.
Upload a reference, say what it controls, and repeat key details like outfit and hair in every prompt. Nano Banana 2 is strong at this.
Sentences. Newer models like Flux 3 and GPT Image 2.5 understand natural descriptions better than comma-separated tag lists.
Pick it for where the image will live: 1:1 for feeds, 4:5 for Instagram ads, 9:16 for Stories, 16:9 for YouTube and slides. Set it with the ratio picker or name it in the prompt.
GPT Image 2.5 for text and layouts, Nano Banana 2 for edits and consistency, Flux 3 for photoreal scenes, Midjourney for artistic looks. All are on Creatide.ai on one balance.
Yes. Use Motion or any image-to-video model on Creatide.ai, and describe only what should move.
Upscale takes images and video up to 8K.
Paid generations on Creatide.ai include commercial rights.
Model names are trademarks of their owners. Tips are based on each model’s public documentation and guides, checked on October 21, 2026. Models update often, so results can vary.
Start generatingUnlock premium AI creation, faster rendering, commercial rights and watermark-free exports whenever your next idea is ready.
Start generatingPremium models and faster renders, ready whenever your next idea is.
