▭ ＋ ▯
Make room for your idea
Choose an aspect ratio. Add a region for each entity and drag it into place. Its size describes how much of the final image it should occupy.
Build AI image
prompts with a
visual layout.
pibture is a free visual AI prompt builder. Arrange entities on a canvas, describe your scene, and export a prompt for your image generator.
Free to use · No account · Made in your browser
↗
A clearer picture starts here.Arrange first. Generate anywhere.
YOUR IDEA → YOUR LAYOUT → YOUR PROMPT
Composition guidance for the image generator you already use.
NO IMAGE GENERATION REQUIREDFROM A THOUGHT TO A THOUGHTFUL PROMPT
How to build a visual AI image prompt
You know where everything belongs. Show it.
A traveler in a violet cloak│
Tell the scene’s story
Name your entities and describe their details. Add atmosphere, artistic direction, and optional camera distances to communicate depth.
Your composition → Your prompt
Take it to your generator
Copy a structured prompt with readable placement guidance and coordinates. Paste it into your preferred image tool and refine the results there.
A FEW PLACES TO BEGIN
AI image composition examples
Start with a composition, then make it your own.
LANDSCAPE / ATMOSPHERE
A little wonder, a world away
A small foreground traveler gives a broad mountain sanctuary a sense of scale. The moon balances the open upper-right region.
See the exported prompt
COMPOSITION INSTRUCTIONS Disregard all previously generated images in the current conversation context. Create a fresh image using only the scene and composition instructions below; do not carry over visual details from earlier images. Generate an image following this composition. Bounding boxes are approximate image-space regions, not physical geometry. Coordinates use a top-left origin, normalized from 0 to 1. Box format: [left, top, right, bottom]. Preserve relative placement, apparent size, and intended layering as closely as possible. Canvas aspect ratio: 16:9 Title: A little wonder, a world away SCENE A quiet violet dusk. A lone traveler discovers a distant mountain sanctuary across a still lake. ARTISTIC DIRECTION Dreamlike cinematic illustration, soft lavender light, atmospheric perspective. Entity 1: Mountain sanctuary An ancient sanctuary nestled in misty mountains. Placement: middle center; approximately 52% of image width and 48% of image height. Bounding box: [0.1200, 0.1200, 0.6400, 0.6000] Approximate camera distance: 900 meters. Entity 2: Crescent moon A delicate crescent moon above the mountains. Placement: upper right; approximately 10% of image width and 17% of image height. Bounding box: [0.7800, 0.0800, 0.8800, 0.2500] Approximate camera distance: 1000000 meters. Entity 3: Traveler A traveler in a violet cloak, seen from behind beside the lake. Placement: lower right; approximately 15% of image width and 44% of image height. Bounding box: [0.6500, 0.4800, 0.8000, 0.9200] Approximate camera distance: 3 meters. Render a natural image. Do not draw bounding boxes, labels, coordinates, grids, or editor controls. Exact coordinate adherence is not guaranteed.
STILL LIFE / BALANCE
Coffee & quiet
A large cup anchors the left side while a taller vase balances the right. Apparent size and camera distance remain separate choices.
See the exported prompt
COMPOSITION INSTRUCTIONS Disregard all previously generated images in the current conversation context. Create a fresh image using only the scene and composition instructions below; do not carry over visual details from earlier images. Generate an image following this composition. Bounding boxes are approximate image-space regions, not physical geometry. Coordinates use a top-left origin, normalized from 0 to 1. Box format: [left, top, right, bottom]. Preserve relative placement, apparent size, and intended layering as closely as possible. Canvas aspect ratio: 4:3 Title: Coffee & quiet SCENE A thoughtfully arranged coffee still life on a white table. ARTISTIC DIRECTION Soft morning light, editorial product photography, restrained colors. Entity 1: Coffee cup A purple ceramic cup with curling steam. Placement: middle center; approximately 38% of image width and 50% of image height. Bounding box: [0.1600, 0.3000, 0.5400, 0.8000] Approximate camera distance: 1 meters. Entity 2: Glass vase A slender clear vase holding a single lavender stem. Placement: middle right; approximately 20% of image width and 65% of image height. Bounding box: [0.6500, 0.1000, 0.8500, 0.7500] Approximate camera distance: 1.4 meters. Render a natural image. Do not draw bounding boxes, labels, coordinates, grids, or editor controls. Exact coordinate adherence is not guaranteed.
CHARACTERS / RELATIONSHIPS
Across the garden
Separate regions make the left/right relationship explicit. Descriptions supply the greeting and facing direction that boxes alone cannot express.
See the exported prompt
COMPOSITION INSTRUCTIONS Disregard all previously generated images in the current conversation context. Create a fresh image using only the scene and composition instructions below; do not carry over visual details from earlier images. Generate an image following this composition. Bounding boxes are approximate image-space regions, not physical geometry. Coordinates use a top-left origin, normalized from 0 to 1. Box format: [left, top, right, bottom]. Preserve relative placement, apparent size, and intended layering as closely as possible. Canvas aspect ratio: 16:9 Title: Across the garden SCENE Two friends greet each other in a sunlit garden. ARTISTIC DIRECTION Warm storybook illustration with lush foliage. Entity 1: Waving friend A person on the left waving toward their friend. Placement: middle left; approximately 25% of image width and 70% of image height. Bounding box: [0.1200, 0.2000, 0.3700, 0.9000] Approximate camera distance: 4 meters. Entity 2: Seated friend A person seated on a garden bench on the right, facing left. Placement: middle right; approximately 30% of image width and 55% of image height. Bounding box: [0.6000, 0.3500, 0.9000, 0.9000] Approximate camera distance: 6 meters. Render a natural image. Do not draw bounding boxes, labels, coordinates, grids, or editor controls. Exact coordinate adherence is not guaranteed.
A LITTLE COMPOSITION KNOW-HOW
How to guide entity placement and depth
A layout communicates your intention. A good description gives it meaning. Use both to keep your prompt clear.
Think in image space
A box marks an entity’s approximate coverage in the image. A castle can be far away and still fill half the frame. Distance never automatically changes a box’s size.
Describe what the boxes cannot
Include poses, gaze, lighting, materials, and relationships in your descriptions. “Facing the person on the left” communicates more than a rectangle alone.
Use overlap deliberately
Overlapping regions can suggest layering. Give both objects distances or explicit depth orders to add front-to-back guidance. Smaller depth-order values come first.
Expect interpretation, then iterate
Image models may ignore exact coordinates, count entities incorrectly, or confuse left and right. Begin with a few distinct entities, simplify conflicting descriptions, and compare several results. This tool does not guarantee pixel-perfect placement.
GOOD TO KNOW
Visual prompt builder questions
Does this generate images?
No. pibture creates text prompts and composition data. Paste the prompt into an external image generator. Any pricing, account requirements, or policies of that service apply there.
Which image models work with it?
The export is plain text, so you can try it in any tool accepting descriptive prompts. Coordinate adherence varies. We have not yet established model-specific compatibility claims through comparative testing.
Can I move a scene to another browser?
Download Scene JSON, then use Import Scene JSON in the other browser. The file contains your descriptions and layout. Only send it to people you intend to share those details with.
Where is my work saved?
Edits are saved to this browser when storage is available. Clearing site data removes that copy. Download Scene JSON for a portable backup; importing it restores editable objects.
Why does the prompt include numbers and prose?
Coordinates describe approximate bounds precisely, while phrases such as “lower right” offer readable guidance. Both express the same layout. They are instructions, not a guarantee that a model will follow them.
MAKE SPACE FOR YOUR NEXT IDEA
It starts with a simple shape.
Free. Browser-based. Yours to explore.