AI Image Prompt Builder

This builder takes one line for the subject and four choices (style, lighting, mood and detail) and joins them with commas into a single English prompt. It is meant as a starting point: when you open an image generator and do not know what to type, or when you want to try the same subject in several styles and compare. It does not promise any particular image, and the sentence it produces is a draft that you are expected to edit.

The five parts of the prompt

Whatever service you use, a prompt tends to communicate best when it answers a few questions in order: what is in the picture, in what medium, under what light, with what feeling and with what finishing touches. The builder uses five slots that map onto those questions.

Every option and the text it inserts

This table is generated from the same lists that fill the drop-downs above, so it always matches what you see on screen.

The assembly rule is simple: subject, style, lighting, mood mood, detail, separated by a comma and a space. Only the mood slot gets an extra word, "mood", appended, so that "serene" reads as "serene mood" and cannot be mistaken for the subject. Whatever you type in the subject box is placed first, exactly as typed apart from collapsed spaces; it is not translated.

Three assembled examples

These sentences were produced by the same rule the page script uses. The counts split on spaces and commas.

Example 1. A cozy landscape for a greeting card
a log cabin in a snowy mountain forest, oil painting, golden hour lighting, cozy mood, pastel colors
17 words, 100 characters

Example 2. A product-style photo for a mock-up
a glass bottle of lemonade on a wooden table, photorealistic, studio lighting, serene mood, highly detailed, 4k
17 words, 111 characters

Example 3. A concept sketch for a game scene
a spaceship above a ruined city, cyberpunk style, neon lighting, mysterious mood, cinematic composition
14 words, 103 characters

All three have about the same length but ask for different media and light, so you can expect quite different pictures. To learn what a word actually does in the generator you use, keep the subject fixed and change a single slot at a time.

Making the subject specific

The subject is the one slot you write yourself, and it is the most common weak point. A bare noun leaves the model to invent everything else. Adding a look, a pose or a setting gives it something to hold on to without making the sentence long.

What changes from service to service

Image generators differ in how they read sentences, which options they support, how long a prompt may be and how well they handle languages other than English, and these details can change between model versions of the same service. Some respond well to flowing descriptions, some to a list of keywords, and some take aspect ratio, a random seed or things to exclude as separate settings or as extra parameters at the end of the line. Because that syntax changes often, this tool does not create parameters for any one service. It outputs plain text that can be pasted anywhere, and you add whatever options the service you use documents.

Refining the result

• Change one thing at a time. If you swap style, lighting and mood together you will not know which change helped.
• Be concrete about the subject. Say how many, where and doing what. Very long prompts can have parts ignored, so prefer a few clear details.
• Avoid contradictions. "Monochrome" next to "vibrant colors", or "minimalist" next to "highly detailed", gives the model opposing instructions and the output can become unstable.
• Handle exclusions the service's way. There is no universal way to say "without", so check the help page of the tool you use.
• Keep the prompts that worked. Many services produce a different picture each time from the same text, so save a good sentence together with any settings.

Mistakes and limits

• Expecting the prompt to guarantee the picture. A prompt is a request. Hands, fine text and small repeated objects are known weak points for many models even when the wording is good.
• Assuming other languages work equally well. If a non-English subject produces odd results, try the subject in English. The builder never translates what you enter.
• Naming living artists, trademarks or characters. Services may restrict this in their terms. Read the section below before publishing such images.
• Looking for saved history. The page does not send or store what you type. It assembles the text in the browser and everything disappears when you reload.

Copyright, likeness and terms of use

Where generated images may be used, whether a generated image is protected by copyright, and how training material is treated vary by country, by service and by subscription plan, and the rules are still developing. This page does not answer those questions. Before using output commercially, or publishing an image that resembles a real person or contains a brand or character, read the terms and content policy of the service you use and the law that applies to you. For anything important, ask a qualified professional.

Related tools

If you meet unfamiliar terms such as "seed" or "diffusion", the AI Terms Glossary explains them. To choose a canvas shape, use the Aspect Ratio Calculator; for a colour scheme, try theColor Palette Generator; and theImage Compressor reduces file size before you share a finished picture.

Frequently Asked Questions