AI Image Prompt Generator That Writes Model-Ready Prompts
Type what you want to see, pick a style, lighting, camera and aspect ratio, and the generator writes a clear, structured prompt for the model you choose. Copy it as plain text or JSON, or send it to the model on Revana with the price shown before anything is made.
Your prompt for Nano Banana Pro · 39 words
A product photo of a ceramic coffee mug with a chipped rim, on a wooden table by a window, with space above for a headline. Soft window light, with a warm neutrals color palette. Close-up shot. Vertical 4:5 composition.
Run it on Revana: $0.04 an image on 1K, quoted before anything is made. Every size and price
# One Idempotency-Key per job: resend with the same $KEY after a timeout and nothing is charged twice
KEY=$(uuidgen)
curl https://api.revana.si/v1/generations \
-H "Authorization: Bearer $REVANA_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: $KEY" \
-d '{
"model": "nano-banana-pro",
"input": {
"prompt": "A product photo of a ceramic coffee mug with a chipped rim, on a wooden table by a window, with space above for a headline. Soft window light, with a warm neutrals color palette. Close-up shot. Vertical 4:5 composition.",
"aspect_ratio": "4:5",
"resolution": "1K"
},
"max_cost_usd": 0.04
}'
The generator runs in your browser and is free. Prices as of November 2026; an image that fails costs $0.00.
The six parts of a strong image prompt
被写体
Start with who or what is in the picture, with one or two concrete details, and put it first so the model weights it most.
Weak: a nice mug
Strong: a ceramic mug with a chipped rim
Setting and composition
Say where the subject is and how it sits in the frame. The aspect ratio belongs here too: a 9:16 story and a 16:9 banner need different framing.
Weak: on a table
Strong: centered on a white table, space above for a headline, vertical 4:5
Style or medium
Name one look: product photo, watercolor, 3D render or flat vector. Mixed styles give muddled pictures, so the generator keeps one per prompt.
Weak: photo, oil painting and anime
Strong: watercolor illustration
Lighting and colour
Soft window light, golden hour or a hard studio flash change the mood more than any adjective. Add a short palette when the picture must match a brand.
Weak: beautiful lighting
Strong: soft window light, warm neutral palette
Camera and lens
Close-up, overhead or an 85mm portrait lens set distance and depth of field. They matter for photos; the generator hides them for illustrations.
Weak: professional camera
Strong: 85mm portrait lens, shallow depth of field
Text in the picture
Put the exact words in quotes, keep them short and say where they go. Text-first models such as Nano Banana Pro and Ideogram V3 render it most reliably.
Weak: a poster with some sale text
Strong: the exact text "Summer Sale" as a bold headline at the top
Write for the model you will use
Models read prompts differently, which is why the generator changes its wording when you switch models. Full sentences suit Nano Banana, GPT Image and Seedream; FLUX does well with photographic terms; Midjourney prefers short phrases and an --ar ratio. Model names belong to their makers.
Nano Banana Pro
Accurate text in pictures, up to 14 reference images and precise edits. Describe the scene in full sentences and quote any words exactly.
GPT Image 2
Text to image and editing up to 4K. Plain, descriptive sentences with the most important detail first.
Flux 2 Pro
Photoreal detail, hands and faces, and multi-reference edits. Lens, light and texture words pay off; keep to one scene.
Seedream 5.0 Pro
Strong on style and concept art, up to 2K. Name one medium and let the model fill in the detail.
Ideogram V3
Made for posters, logos and packaging with readable lettering. Keep the words short and say where they go.
Midjourney V7
Its own distinctive look from short, comma-separated phrases, with the aspect ratio as --ar.
Prompt examples you can adapt
Product shot on a clean background
A product photo of a matte black water bottle, standing on a white seamless background with a subtle shadow, centered. Soft diffused studio light. Eye-level shot on a 35mm lens. Square 1:1 composition. Swap the product and keep the rest for a consistent catalogue. Draft on Nano Banana 2, then make the final set on Nano Banana Pro. Nano Banana 2, $0.03 an image.
Lifestyle scene
A woman reading on a sunlit balcony with a cup of coffee, morning, plants along the railing, lifestyle photo, golden hour sunlight, 85mm portrait lens with a shallow depth of field, warm amber and cream color palette, sharp focus, natural skin and material texture, vertical 4:5 frame Lifestyle shots gain the most from light and lens words. FLUX 2 Pro handles faces and hands well here. Flux 2 Pro, $0.03 an image.
Social graphic with a headline
A flat vector illustration of a rocket lifting off over a city skyline. A navy and orange color palette. Show the exact text "Launch Week" as a bold headline at the top, in clean, legible lettering with every letter spelled right. Tall vertical 9:16 composition, balanced graphic layout. Keep the headline under five words. Ideogram V3 and Nano Banana Pro give the cleanest lettering. Ideogram V3, $0.03 an image.
Editorial illustration
A watercolor illustration of a fox in a snowy forest at dusk, loose brush strokes, pine trees fading into mist. A muted blues and soft pink color palette. Landscape 3:2 composition. Illustrations need no camera words. Style and colour carry most of the result. Seedream 5.0 Pro, $0.04 an image.
What it costs to run your prompt
| モデル | 開発元 | 最低価格 | At |
|---|---|---|---|
| Nano Banana 2 | $0.03 | 1K | |
| GPT Image 2 | OpenAI | $0.03 | 1K |
| Flux 2 Pro | Black Forest Labs | $0.03 | 1MP |
| Ideogram V3 | Ideogram | $0.03 | turbo mode |
| Nano Banana Pro | $0.04 | 1K | |
| Seedream 5.0 Pro | ByteDance | $0.04 | 1K |
| Midjourney V7 | Midjourney | $0.06 | relax mode |
Prices on Revana as of November 2026; larger sizes cost more, and every size is on each model's page. The generator itself is free, and the price does not depend on how long the prompt is.
Testing one prompt on several models
Running one prompt through GPT Image 2, Flux 2 Pro and Nano Banana Pro with four images each is 12 images, which comes to $0.40 at the smallest sizes. That is a cheap way to find the model that suits your style; keep the winner for the full batch, since the prompt carries over unchanged.
Failed images cost nothing
If a route errors or is busy, Revana moves the job to another route for that model and charges once for the image that arrives. A failed image costs $0.00, and there is no retry code to write.
Prepaid credit from $10
Credit starts at $10, has no subscription and never expires, and one balance pays for image, video and language models. $10 covers 250 Nano Banana Pro images at 1K. The full rules are on the pricing page.
From prompt to finished image in three steps
- Build and copy the prompt. Fill in the fields, read the live prompt and adjust it until it describes the picture you want. Plain text works in most tools; JSON keeps each part separate for apps that build prompts in code.
- Quote and run it on Revana. Send the prompt to
POST /v1/generationswith the model and size, or check the price first with the free quote endpoint. Every image model takes the same request, so switching models is a one-word change, andmax_cost_usdcaps what a job may cost. - Or let your agent do it. Connect the Revana MCP server and Claude, ChatGPT, Cursor or Codex can write the prompt, quote it and make the image in one conversation, spending from your prepaid credit. Each image shows in your dashboard with its exact cost.
Prompt habits that waste images
Being too vague
a cool product photo leaves the model guessing. Add the subject detail, the setting and the light; specific prompts need fewer retries.
Stacking clashing styles
A photo, an oil painting and an anime frame at once gives a picture that is none of them. Choose one style a prompt and run separate prompts to compare.
Ignoring the aspect ratio
A square image cropped into a 9:16 story loses the subject. Set the ratio before you generate; the generator adds it to every prompt.
Changing everything at once
When a result is close, change one part at a time so you can tell which change helped, and note each change so the winning prompt is easy to reuse.
FAQ
What does an AI image prompt generator do?
It turns a short idea into a detailed prompt by adding the parts image models respond to, such as style, lighting, camera and composition. This one also adjusts the wording for the model you pick and gives you the Revana request that makes the picture.
Is this prompt tool free?
Yes. The prompt generator runs in your browser and costs nothing. Making the image needs a free Revana API key and prepaid credit, with images from $0.03 each on Nano Banana 2 and GPT Image 2 and Flux 2 Pro and Ideogram V3 (prices as of November 2026).
Do the prompts work in other tools?
Yes. The prompts are plain text, so you can paste them into any image tool that takes a written prompt. The model setting only changes the wording and length to suit that model best.
How long should an image prompt be?
Most good prompts run from one to three sentences, or about 20 to 60 words. Longer prompts can work, but extra detail often competes for attention, so keep the subject first and cut anything that does not change the picture.
Which model is best for text in images?
Nano Banana Pro and Ideogram V3 are the usual picks for readable text such as headlines, labels and posters. Both run on your Revana key, so you can test the same prompt on each and keep the result that reads best.
Does a longer prompt cost more?
No. Image models on Revana are priced per image at the size you choose, not by prompt length. A detailed prompt costs the same as a short one, and it usually saves money by needing fewer retries.
画像は商用利用できますか?
Commercial use depends on each model maker's terms, so check them for the model you use before publishing. Each model page on Revana names its maker.
Turn your prompt into an image
Create your free Revana key, paste the prompt and make your first image in minutes. You pay only for images that are delivered, and the Revana team can set up volume terms on the same key.