Guide
Build a detailed, reusable image prompt with composition, lighting and negative terms. Independent tool, not affiliated with NovelAI.
Describe the image in a useful order
Begin with who or what is in the image and what is happening. Then add the setting, medium, framing and light. This keeps the central idea clear before secondary details accumulate. A prompt about a botanist inspecting a flower gives a model an action and a focal point; a list of quality adjectives does neither. Use the must-have field for two or three details that matter to your story. The composer orders your choices into plain language you can reuse in an image system of your choice.
Make composition and lighting agree
Choose a frame that gives your subject enough space. Portrait framing suits a single figure; a wide frame can establish a landscape or several characters. Specify the camera distance and the placement of the main subject before adding decorative details. Light should reinforce that choice: a small warm source can draw attention to a face or object, while broad ambient light describes the setting. Avoid mutually incompatible directions such as an extreme close-up and a full-body view in the same prompt. Change one major variable between trials so you can understand the result.
Use negative prompts as model-dependent guidance
The avoid field is exported separately because image systems handle negative prompts differently. Some accept a dedicated negative-prompt input; others only accept natural-language instructions. A negative term does not guarantee that an unwanted feature will disappear. Keep your strongest constraints in the positive prompt when possible, such as a plain background or one visible hand holding a notebook. This workspace does not add proprietary NovelAI tags or weighting syntax. Optional hosted generation depends on the service configured by the site operator and reports an error when it is unavailable.
How the result is made
A transparent template combines the supplied subject, scene, medium, composition, light and palette. It deduplicates comma-separated negative terms. Optional image generation is a separate explicit server request, never a relabeled prompt.