The Prompt Generator helps turn a rough visual idea into a structured direction. It is useful when you know the subject or job but do not yet know how to organize the scene, style, camera, motion, or details to preserve. The structured builder is free to use for prompt planning without spending a render credit. It prepares text; it does not make the image or video until you choose a separate generation workflow.
These prompt examples are editable starting points, not tested outputs. Treat the structure as a way to make decisions visible, then adjust the words for the actual source image, audience, and intended use.
1. Build the image direction in layers
Start with the subject and the job: “a ceramic mug for a quiet product listing” or “a fictional adult character in a three-quarter editorial portrait.” Add the setting, framing, light, palette, and one or two important details. A short, ordered prompt is easier to refine than a paragraph that mixes every possible style.
For an image, a practical order is subject, action or pose, setting, composition, light, materials, and constraints. Say what should remain recognizable when you are working from a reference. The AI Photo Editor is useful when the next step is a background, style, color, or product-image edit. The AI Photo Enhancer fits a narrower request around exposure, color, noise, and clarity for one input image.
2. Describe motion only when you need video
For a video direction, describe the prepared frame, the subject’s motion, the camera movement, environmental motion, timing, and audio expectation when the workflow supports it. “A product remains centered while the camera makes a slow half-orbit; soft daylight shifts across the table; keep the label facing the camera” is more useful than a list of cinematic adjectives. Keep the action physically simple enough to review.
When you are ready to render an image, open Text to Image, where the current text-to-image workflow uses Nano Banana. For a video render, open Text to Video, where the current workflow uses Kling 3.0. The model, settings, available controls, and live credit quote belong to those render pages. The prompt planner does not make the render cost disappear: the actual image or video generation uses the current quoted credits beside Generate.
3. Use AI Refine only as an optional step
After signing in, an authenticated user may choose the existing AI Refine feature when it is offered in the builder. It is optional. You can always edit the structured fields yourself, and the refined wording still needs a human review. AI Refine does not replace the render page’s model choice, credit quote, or moderation state.
Keep the planning step and the render decision separate. The text you build may be reused, but each image or video request is evaluated with the current input, output, account, and policy checks. Read the quote before submission and inspect the result before sharing it.
4. Iterate with one meaningful change
If the result misses the point, change one field: the crop, the background, the color, the camera motion, or the subject detail. The Image to Image page helps when you have a reference-led transformation in mind. Browse Prompts for directions to adapt and Tools for other workflows. Keep the original prompt beside each render so you can tell which change improved the next version.
