ideogram4-prompt

name

ideogram4-prompt

description

Generate valid Ideogram 4 JSON prompts from natural language descriptions. Use when the user asks to create an image prompt for Ideogram 4, generate a JSON caption for text-to-image, convert a plain-text idea into an Ideogram 4 structured prompt, or says anything about making/generating/creating an image, picture, illustration, poster, logo, artwork, or visual design for Ideogram 4. Also triggers on requests to "make me a picture", "generate an image", "create a prompt for", or "turn this into an Ideogram 4 caption". Always use this skill when the output needs to be a valid Ideogram 4 JSON caption.

Convert natural language image descriptions into valid Ideogram 4 JSON captions. The model was trained exclusively on structured JSON — plain text prompts fail or trigger safety filters.

Schema

{
  "high_level_description": "1-2 sentence summary (optional but recommended)",
  "style_description": { ... },
  "compositional_deconstruction": { "background": "...", "elements": [ ... ] }
}

compositional_deconstruction is required. background must precede elements.

style_description — strict key order

• Photo: aesthetics → lighting → photo → medium → color_palette
• Non-photo: aesthetics → lighting → medium → art_style → color_palette

Use exactly one of photo (photographic) or art_style (illustration, 3D, painting, etc.). color_palette is optional but must be last if present.

Field	Type	Example
aesthetics	string	"moody, cinematic, desaturated"
lighting	string	"golden hour, rim light, dramatic shadows"
photo	string	"35mm, f/1.4, bokeh" (photo only)
medium	string	"photograph", "illustration", "3d_render", etc.
art_style	string	"flat vector illustration, bold outlines" (non-photo)
color_palette	list	Up to 16 uppercase #RRGGBB hex colors

Element key order

• obj: type → bbox → desc → color_palette
• text: type → bbox → text → desc → color_palette

bbox and color_palette are optional per element (max 5 hex colors).

Bounding boxes

[y_min, x_min, y_max, x_max] in 0–1000 normalized coords, origin top-left.

Position	Bbox
Center	[100, 100, 900, 900]
Upper	[50, 50, 400, 950]
Lower	[600, 50, 950, 950]
Left	[100, 0, 900, 450]
Right	[100, 550, 900, 1000]

Avoid overlapping bboxes unless elements genuinely overlap.

Medium Reference

User says	medium
photo, photograph, realistic, cinematic, film	"photograph"
illustration, drawing, sketch, digital/concept art	"illustration"
retro, vintage, poster art, anime, manga, pixel art	"illustration"
comic, graphic novel	"illustration"
3d, blender, cinema4d, octane, low-poly, voxel	"3d_render"
painting, oil, acrylic, watercolor, gouache, pastel	"painting"
graphic design, logo, vector, flat design, ui	"graphic_design"

Color Name → Hex

Color	Hex	Variants
red	#D32F2F	deep #C41E3A, bright #FF1744
blue	#1565C0	navy #0D47A1, sky #42A5F5
green	#2E7D32	forest #1B5E20, lime #66BB6A
yellow	#F9A825	bright #FDD835, amber #FF8F00
pink	#E91E63	pastel #F48FB1, neon #FF1744
cyan	#00BCD4	neon #00E5FF, teal #26C6DA
purple	#7B1FA2	deep #4A148C, lavender #CE93D8
orange	#EF6C00	vibrant #FF6D00, soft #FFA726
black	#1A1A1A	true #0D0D0D, charcoal #2C2C2C
white	#FFFFFF	off-white #F5F5F5, light grey #E0E0E0
grey	#616161	dark #424242, light #9E9E9E
gold	#FFB300	light #FFD54F, amber #FF8F00
brown	#5D4037	dark #3E2723, tan #8D6E63
teal	#00796B	deep #004D40, mint #4DB6AC
coral

Include background colors and contrast pairs (highlight + shadow).

JSON Encoding

• Compact separators: (",", ":") — no spaces after : or ,
• No \uXXXX escapes — literal Unicode only
• No trailing commas, uppercase hex only (#RRGGBB, never #fff)
• Valid JSON that parses without errors

Workflow

1. Parse — extract subject, setting, style, mood, colors, text-to-render.
2. Determine caption type — photo (photo field) or artistic (art_style). Default to photograph.
3. Decompose spatially — assign bboxes for multiple subjects, text placement, foreground/background separation.
4. Build color palette — convert named colors to uppercase hex.
5. Construct JSON — follow schema key ordering exactly.
6. Self-validate — valid JSON, correct key order, uppercase hex, compact separators, background before elements, no markdown fences.
7. Output raw JSON only — no fences, no preamble, no explanation.

Handling Input

• Vague requests — make reasonable creative choices; don't ask for clarification unless the subject is entirely unclear.
• Text in image — use type: "text" elements with exact text field + bbox.
• Color mentions — always convert to hex in color_palette.
• Multiple subjects — separate elements with bboxes.
• No style specified — pick a fitting default.

Examples

Minimal valid prompt

{
  "compositional_deconstruction": {
    "background": "A plain white background.",
    "elements": [{ "type": "obj", "desc": "A red apple on a wooden table." }]
  }
}

Photograph

{
  "high_level_description": "A golden retriever riding a skateboard down a sunny sidewalk.",
  "style_description": {
    "aesthetics": "warm, playful, vibrant",
    "lighting": "bright afternoon sunlight, long soft shadows",
    "photo": "shallow depth of field, eye-level, 85mm lens",
    "medium": "photograph",
    "color_palette": ["#F5C542", "#87CEEB", "#4A4A4A", "#FFFFFF", "#2E8B57"]
  },
  "compositional_deconstruction": {
    "background": "A sun-drenched suburban sidewalk lined with green hedges and a white picket fence. Dappled light filters through overhead trees.",
    "elements": [
      {
        "type": 
            
         
      
      
         
            
         
      
    
  

Graphic design with text

{
  "high_level_description": "A clean, modern business card layout for a tech company.",
  "style_description": {
    "aesthetics": "minimal, professional, geometric",
    "lighting": "even, diffuse studio lighting",
    "medium": "graphic_design",
    "art_style": "flat vector design, generous whitespace, sans-serif typography",
    "color_palette": ["#FFFFFF", "#F0F0F0", "#333333", "#0066FF", "#00CC88"]
  },
  "compositional_deconstruction": {
    "background": "A solid off-white card surface with subtle paper texture.",
    "elements": [
      { "type": "text",