Copy · paste · generate — no account needed to start
ChatGPT Image Prompts
Thirty prompts you can copy straight into the generator, grouped by what you are trying to make. Each one names the light, the framing and the finish, because a prompt that only names the subject leaves the model to guess the rest.
What the prompts above produce
Real output from this workspace. Every picture is AI-generated, so your result will differ with your wording and the model you pick — treat these as the shape of the outcome, not a fixed one.
Type inside the picture
What makes a prompt work here
A usable prompt is closer to a shot list than a sentence. The five parts below are what the strong ones on this page all have, and dropping any one of them is where most disappointing results come from.
Close-up or wide, eye level or above, and the focal length if it matters. It fixes the composition before the model invents one — and if the composition is right but the picture is too small, the image upscaler is the second half of that job.
What must stay the same — a face, a product, a colour. This is what makes a series look like a series.
Draft at 1K while the prompt is still moving. Re-run the keeper at 4K once it is not. If what you want is a change to a picture you already have, that is the image editor rather than a new prompt.
Creative engine
What running a prompt involves
Paste it, pick a resolution, generate. The estimate updates before you commit and a failed run costs nothing.
Subject, stated plainly
Who or what is in frame, in ordinary words. Elaborate adjectives buy less than a clear noun.
Light
Where the light comes from and how hard it is. This single clause changes more about the result than any style word.
Framing and lens
Close-up or wide, eye level or above, and the focal length if it matters. It fixes the composition before the model invents one.
Portrait prompts for ChatGPT image generation
Faces are where prompt accuracy shows first — wrong lighting or a lost detail reads instantly. Each of these names the light, the lens and the framing, because those three do most of the work.
Window-light studio portrait
A close-up portrait of a woman with freckles beside a north-facing window, soft directional light falling across one cheek, shallow depth of field, natural skin texture, muted warm palette, shot on 85mm.
Hard-sun editorial
An editorial portrait in hard midday sun against a plaster wall, sharp shadow cutting diagonally across the frame, high contrast, saturated skin tones, no fill light.
Environmental portrait, artisan
A woodworker in their workshop, half-lit by a doorway, tools and sawdust visible behind, hands in focus and face slightly turned, documentary framing.
Classic 35mm film look
A candid portrait on 35mm film, visible grain, slightly warm cast, subject mid-laugh, background falling out of focus at f/2.
Profile against gradient
A clean profile shot against a seamless gradient backdrop, rim light tracing the jaw and nose, everything else falling to near-black.
Group of four, one light
Four people arranged at different depths under a single overhead source, each face readable, the falloff carrying the depth.
Poster and typography prompts
This is the group that separates image models. Legible words inside a picture is a specific capability, and a prompt that names the type treatment gets far closer than one that just says 'with text'.
Minimalist film poster
A minimalist film poster, single centred object on a flat colour field, large condensed sans-serif title across the lower third, small credit block beneath, generous margins.
Retro travel print
A mid-century travel poster of a coastal town, flat colour separations, four inks only, the town name in bold slab serif across the sky.
Product ad with headline
A commercial product shot on a seamless backdrop, the product lit from behind, a short headline set in clean grotesque type across the top, brand mark bottom-right.
Concert bill
A gig poster with stacked band names in decreasing type sizes, high-contrast duotone photograph behind, date and venue in a boxed footer.
Infographic panel
A flat-design infographic panel explaining a three-step process, numbered circles, short labels, consistent icon weight, plenty of white space.
Menu board
A cafe menu board in chalk lettering on dark slate, three sections with prices aligned right, small illustrated flourishes between sections.
Product and food prompts
Commercial work is mostly about the light and the surface. These name both, plus what is happening physically — splash, condensation, steam — because that is what makes a still look shot rather than rendered.
Chilled can, splash
A chilled drinks can suspended mid-air in a ring of splashing liquid, fruit slices frozen around it, studio strobes, deep saturated background.
Condensation macro
A macro shot of condensation beading on cold glass, backlit so each droplet carries a highlight, the liquid inside softly out of focus.
Food on a wooden table
A dessert in a glass bowl on a worn wooden table, window light from behind, syrup running down the side, warm interior falling out of focus.
Flat-lay cosmetics
A flat-lay of cosmetics on a soft pink ground, products arranged loosely around a central item, handwritten script overlaid, top-down, even shadowless light.
Perfume on dark water
A faceted perfume bottle standing on still dark water, coloured smoke coiling around the stopper, single hard light from behind.
Packaging on seamless
Packaging standing on a seamless backdrop, three-quarter angle, soft box left and a rim light right, faint reflection under the base.
Character and illustration prompts
Consistency is the ask here: the same face across several frames. Naming the identity anchors — hair, marking, wardrobe — matters more than naming the art style.
Storybook character sheet
An illustrated character sheet of a young witch with a black cat, four poses across one frame, painted border, lantern light, consistent face and costume in every pose.
Isometric room
An isometric cutaway of a small workshop, warm interior lighting, dozens of small readable props, clean line weight, muted palette.
Watercolour animal study
A loose watercolour study of a fox, wet edges and bloom in the wash, white paper showing through, minimal linework.
Ukiyo-e wave
A woodblock print in the ukiyo-e tradition: a breaking wave, flat colour areas, visible carving texture, a small boat for scale.
3D clay figure
A soft clay-render figure of a chef, rounded forms, matte surface, gentle studio lighting, pastel background, subtle contact shadow.
Comic panel
A single comic panel: two figures on a rooftop at dusk, heavy ink shadows, halftone shading, speech balloon left empty.
Scene and landscape prompts
Wide scenes reward prompts that name the atmosphere and the time of day, then one specific detail to anchor scale.
Misty gorge
A gorge in heavy fog, a thin waterfall on the far wall, layered ridges receding into white, a single bird for scale.
Cliffside villa
A white villa built into limestone karst above still green water, late afternoon light, small boat wake crossing the frame.
Neon alley, rain
A narrow alley after rain, neon signage reflected in standing water, steam from a vent, one figure walking away from camera.
Desert minimal
A single figure on a pale dune under a wide flat sky, tiny in frame, everything else empty, high-key exposure.
Interior at golden hour
A living room at golden hour, low sun striping the floor through blinds, dust visible in the beam, furniture mostly in shadow.
Forest canopy from below
Looking straight up through a dense canopy, sun breaking through in shafts, leaf detail sharp at the edges of the light.
Questions about writing image prompts
No. GPT Image 2 is the OpenAI model that draws inside ChatGPT, and it runs here directly — no Plus account, and no account here either for your first pictures.
Because the parts you leave out get invented. Naming light, framing and finish removes three whole categories of guesswork, which is why a longer prompt is usually the more predictable one.
Yes, and it is worth doing. The same wording lands differently on Nano Banana 2, Seedream and Qwen Image 3, and running one prompt across two models is the fastest way to learn what each is good at.
Put the identity anchors in every prompt — hair, marking, wardrobe — and change only the clause describing the pose or setting. For tighter control, upload the first result as a reference and edit from there.
Twenty credits at 1K, forty at 2K, eighty at 4K. The estimate shows before you generate, and a failed run is refunded.
Paid plans carry commercial use rights. The prompts on this page are examples to edit rather than anything we claim ownership of.
One credit balance for every image model
Credits are shared across everything Veida AI runs. You see the cost before each run, and a failed job is never charged.
Free
What's included
- Generate before you sign up — the free engine runs in your browser
- 200 credits the moment you create an account
- 40 more every day you check in — about 1,400 a month, roughly 70 images
Secure checkout by Stripe. Card details never touch our servers.
Plus
What's included
- Access to GPT Image 2, Nano Banana 2, Seedream and Qwen
- Up to 4K output
- Watermark-free downloads
Secure checkout by Stripe. Card details never touch our servers.
Pro
What's included
- Everything in Plus
- Higher image concurrency — run several prompts at once
- No throttle on the generator queue
Secure checkout by Stripe. Card details never touch our servers.
Max
What's included
- Everything in Pro
- The highest concurrency tier on the site
- Skip soft-rate limits entirely
Secure checkout by Stripe. Card details never touch our servers.
Charges appear as “VEIDA AI” on your card statement. You can cancel any time from Settings → Billing; cancellation takes effect at the end of the period you already paid for. Operator details, the full model list, and the refund window are on the about page.
Where to take a prompt next
The models these were written against, and the pages that go deeper on each.
Pick a prompt and run it
Copy any prompt above, paste it into the generator, and change one clause at a time. That is the whole method.
