JSON Prompt Guide

Ideogram 4 JSON Prompt Guide

A focused guide to structured JSON prompts for Ideogram 4: field roles, bbox layout, text elements, style blocks, and practical prompt planning.

What this page covers

Focused topics, kept narrow so the page stays useful.

ideogram 4 json promptideogram json prompt

Start with the image goal

A strong JSON prompt begins with a short high-level description. This is the brief that tells the model what the image should become before you split the scene into style and elements.

Keep it to one or two sentences.

Name the format: poster, product shot, logo concept, avatar, or illustration.

Mention readable text only when the image needs typography.

Separate style from structure

Use style_description for the overall visual treatment, then compositional_deconstruction for background and elements. This makes the prompt easier to edit than a single long paragraph.

Use photo for photographic scenes.

Use art_style for graphic design, illustration, painting, or 3D outputs.

Use color_palette for dominant colors instead of vague color words only.

Use bbox only where placement matters

Bounding boxes are most useful for dense posters, product ads, UI-style compositions, and multi-text images. They are optional for simple scenes, but valuable when the layout must be planned.

Ideogram documentation uses [y_min, x_min, y_max, x_max].

Use broad zones first; do not overfit tiny boxes.

Keep each exact word in a separate text element.