Prompting Guide - FLUX.2

Master FLUX.2 prompting with structured JSON, hex colors, and multi-reference techniques

FLUX.2 delivers exceptional prompt following and supports advanced techniques like structured JSON prompting, precise hex color control, and multi-reference image editing. This guide covers everything you need to create stunning results.

No negative prompts: FLUX.2 does not support negative prompts. Focus on describing what you want, not what you don’t want.

JSON Structured Prompting

FLUX.2 excels at interpreting structured JSON prompts, giving you precise control over every aspect of your image. This is particularly powerful for production workflows and automation.

The Base Schema

{
  "scene": "overall scene description",
  "subjects": [
    {
      "description": "detailed subject description",
      "position": "where in frame",
      "action": "what they're doing"
    }
  ],
  "style": "artistic style",
  "color_palette": ["#hex1", "#hex2", "#hex3"],
  "lighting": "lighting description",
  "mood": "emotional tone",
  "background": "background details",
  "composition": "framing and layout",
  "camera": {
    "angle": "camera angle",
    "lens": "lens type",
    "depth_of_field": "focus behavior"
  }
}

Building a Prompt Step by Step

Let’s build a product shot incrementally to see how each element contributes.

Step 1: Generating a coffee mug

{
  "scene": "Professional studio product photography setup with polished concrete surface",
  "subjects": [
    {
      "description": "Minimalist ceramic coffee mug with steam rising from hot coffee inside",
      "pose": "Stationary on surface",
      "position": "Center foreground on polished concrete surface",
      "color_palette": ["matte black ceramic"]
    }
  ],
  "style": "Ultra-realistic product photography with commercial quality",
  "color_palette": ["matte black", "concrete gray", "soft white highlights"],
  "lighting": "Three-point softbox setup creating soft, diffused highlights with no harsh shadows",
  "mood": "Clean, professional, minimalist",
  "background": "Polished concrete surface with studio backdrop",
  "composition": "rule of thirds",
  "camera": {
    "angle": "high angle",
    "distance": "medium shot",
    "focus": "Sharp focus on steam rising from coffee and mug details",
    "lens-mm": 85,
    "f-number": "f/5.6",
    "ISO": 200
  }
}

Step 1: Single matte black mug with steam

Step 2: Adding a second mug in a different color

{
  "scene": "Professional studio product photography setup with polished concrete surface",
  "subjects": [
    {
      "description": "Minimalist ceramic coffee mug with steam rising from hot coffee inside",
      "pose": "Stationary on surface",
      "position": "Center foreground on polished concrete surface",
      "color_palette": ["matte black ceramic"]
    },
    {
      "description": "Minimalist ceramic coffee mug, matching design to the black mug",
      "pose": "Stationary on surface",
      "position": "Right side of the black mug on polished concrete surface",
      "color_palette": ["matte yellow ceramic"]
    }
  ],
  "style": "Ultra-realistic product photography with commercial quality",
  "color_palette": ["matte black", "matte yellow", "concrete gray", "soft white highlights"],
  "lighting": "Three-point softbox setup creating soft, diffused highlights with no harsh shadows",
  "mood": "Clean, professional, minimalist",
  "background": "Polished concrete surface with studio backdrop",
  "composition": "rule of thirds",
  "camera": {
    "angle": "high angle",
    "distance": "medium shot",
    "focus": "Sharp focus on steam rising from coffee and both mugs in frame",
    "lens-mm": 85,
    "f-number": "f/5.6",
    "ISO": 200
  }
}

Step 2: Added a yellow mug to the composition

Step 3: Change the color of the steam

{
  "scene": "Professional studio product photography setup with polished concrete surface",
  "subjects": [
    {
      "description": "Minimalist ceramic coffee mug with bright red steam rising from hot coffee inside",
      "pose": "Stationary on surface",
      "position": "Center foreground on polished concrete surface",
      "color_palette": ["matte black ceramic", "bright red steam"]
    },
    {
      "description": "Minimalist ceramic coffee mug, matching design to the black mug",
      "pose": "Stationary on surface",
      "position": "Right side of the black mug on polished concrete surface",
      "color_palette": ["matte yellow ceramic"]
    }
  ],
  "style": "Ultra-realistic product photography with commercial quality",
  "color_palette": ["matte black", "matte yellow", "bright red", "concrete gray", "soft white highlights"],
  "lighting": "Three-point softbox setup creating soft, diffused highlights with no harsh shadows",
  "mood": "Clean, professional, minimalist",
  "background": "Polished concrete surface with studio backdrop",
  "composition": "rule of thirds",
  "camera": {
    "angle": "high angle",
    "distance": "medium shot",
    "focus": "Sharp focus on steam rising from coffee and both mugs in frame",
    "lens-mm": 85,
    "f-number": "f/5.6",
    "ISO": 200
  }
}

Step 3: Changed the steam color to bright red

You can include the JSON directly in your prompt, or flatten it into natural language. FLUX.2 understands both formats.

HEX Color Code Prompting

FLUX.2 supports precise color matching using hex codes. This is essential for brand consistency and design work.

Basic Syntax

Signal hex colors with keywords like “color” or “hex” followed by the code:

The vase has color #02eb3c
The background is hex #1a1a2e

Gradient Colors

Apply gradients by specifying start and end colors:

Prompt: “A vase on a table in living room, the color of the vase is a gradient, starting with color #02eb3c and finishing with color #edfa3c. The flowers inside the vase have the color #ff0088”

Hex colors applied to vase and flowers

Brand color matching for product design

Color in JSON Prompts

Combine hex colors with structured prompts for maximum control:

{
  "scene": "Makeup flat lay on marble surface",
  "subjects": [
    {
      "description": "eyeshadow palette",
      "colors": ["#E91E63", "#9C27B0", "#673AB7", "#3F51B5"]
    }
  ],
  "style": "beauty product photography",
  "lighting": "soft diffused overhead lighting"
}

Hex codes work best when clearly associated with specific objects. Vague references like “use #FF0000 somewhere” may produce inconsistent results.

Infographics and Data Visualization

FLUX.2 can generate infographics with clean typography and structured layouts.

Infographic Template

{
  "type": "infographic",
  "title": "Your Main Title",
  "subtitle": "Supporting context",
  "sections": [
    {
      "heading": "Section 1",
      "content": "Key information",
      "visual": "icon or chart type"
    }
  ],
  "color_scheme": ["#primary", "#secondary", "#accent"],
  "style": "modern, clean, corporate"
}

Example Prompt:

“Create a vertical infographic about coffee consumption worldwide. Title: ‘Global Coffee Culture’. Include 3 sections with statistics, use icons for each country, color scheme #4A2C2A (brown) and #F5E6D3 (cream). Modern minimalist style with clean typography.”

Data visualization with clean typography

Typography and Design

FLUX.2 excels at generating clean typography, product marketing materials, and magazine layouts.

Product advertisement with clean typography

Magazine cover layout

Product Ad: “Samsung Galaxy S25 Ultra product advertisement, ‘Ultra-strong titanium’ headline, ‘Shielded in a strong titanium frame, your Galaxy S25 Ultra always stays protected’ subtext, close-up of phone edge showing titanium frame, dark gradient background, clean minimalist tech aesthetic, professional product photography”

Magazine Cover: “Women’s Health magazine cover, April 2025 issue, ‘Spring forward’ headline, woman in green outfit sitting on orange blocks, white sneakers, ‘Covid: five years on’ feature text, ‘15 skincare habits’ callout, professional editorial photography, magazine layout with multiple text elements”

Multi-Language Prompting

FLUX.2 has excellent multi-language understanding. You can prompt in your native language for more culturally authentic results.

French: Un marché alimentaire dans la campagne normande

Thai: ตลาดอาหารเช้าในชนบทใกล้กรุงเทพฯ

Korean: 서울 도심의 옥상 정원

French: “Un marché alimentaire dans la campagne normande, des marchands vendent divers légumes, fruits. Lever de soleil, temps un peu brumeux”

Thai: “ตลาดอาหารเช้าในชนบทใกล้กรุงเทพฯ พ่อค้าแม่ค้ากำลังขายผักและผลไม้นานาชนิด บรรยากาศยามพระอาทิตย์ขึ้น มีหมอกจาง ๆ ปกคลุม สงบและอบอุ่น”

Korean: “서울 도심의 옥상 정원, 저녁 노을이 지는 하늘 아래에서 사람들이 작은 등불을 켜고 있다. 화려한 네온사인이 멀리 반짝이고, 정원에는 다양한 꽃들이 피어 있다. 분위기는 따뜻하고 낭만적이다”

Prompting in the native language of the content you’re creating often produces more culturally authentic results—local markets, architecture, and atmosphere are rendered with greater accuracy.

Comic Strips and Sequential Art

Create consistent comic panels with character continuity. The key is to define your character in detail and maintain that description across panels.

The Diffusion Man Story

Generate each panel separately while keeping character descriptions consistent:

Panel 1: The Crisis

Panel 2: The Transformation

Panel 3: The Battle

Panel 4: Victory

Panel 1 Prompt: The Crisis

“Style: Classic superhero comic Character: Worried scientist frantically typing on glowing holographic keyboard, face illuminated by blue light showing deep concern Setting: Massive computer server room with sparking circuits and red warning lights flashing on monitors Text: ‘The AI models are corrupting! We need Diffusion Man!’ Mood: Tense, urgent + dramatic blue and red tones”

Panel 2 Prompt: The Transformation

“Style: Classic superhero comic with dynamic action lines and electric energy effects Character: Diffusion Man/Mild-mannered programmer (30 years old, brown skin tone, short natural fade haircut with black hair, black-framed glasses, light blue button-up shirt, athletic build, strong jawline) body begins to glow with swirling gradients of deep purple, electric blue, and hot pink energy, mathematical equations and neural network patterns flowing around him in glowing lines Setting: Small office with computer monitors displaying code and error messages Text: ‘When noise becomes signal, I am… DIFFUSION MAN!’ Mood: Powerful, transformative + dramatic backlighting and energy radiating outward in waves”

Panel 3 Prompt: The Battle

“Style: Classic superhero comic with explosive action and dynamic composition Character: Diffusion Man (athletic 30-year-old with brown skin tone and short natural fade haircut with black hair, wearing sleek bodysuit with gradient patterns from deep purple to electric blue to hot pink, glowing neural network emblem on chest with interconnected nodes, short gradient cape, purple half-mask showing strong jawline and confident expression) extends both hands forward in powerful stance, shooting beams of structured noise and latent space energy at corrupted digital monsters made of glitching pixels and broken code Setting: Digital cyberspace environment with floating data cubes and cascading binary code Text: ‘Time to DENOISE this chaos!’ Mood: Intense, action-packed + bright energy flashes and electric effects”

Panel 4 Prompt: Victory

“Style: Classic superhero comic with warm, triumphant colors and clean composition Character: Diffusion Man (athletic 30-year-old with brown skin tone and short natural fade haircut with black hair, wearing sleek gradient bodysuit from deep purple to electric blue to hot pink, glowing neural network emblem on chest, short gradient cape flowing behind him, purple half-mask, strong jawline, confident heroic smile) stands heroically giving thumbs up gesture to grateful scientist beside him, her computer screens now showing stable green indicators and success messages Setting: Calm server room with soft blue ambient lighting and orderly data streams flowing smoothly in organized patterns Text: ‘You saved us, Diffusion Man! The models are generating perfectly again!’ Mood: Victorious, hopeful + golden sunset-like tones streaming through windows”

Character Consistency: Notice how Diffusion Man’s description stays detailed and consistent across panels—brown skin tone, short natural fade haircut, gradient bodysuit from purple to blue to pink, neural network emblem, purple half-mask. Repeat these details in every panel prompt.

Photorealistic Styles

FLUX.2 excels at photorealistic generation. Reference specific eras and techniques for distinctive looks.

Style Reference Guide

Style	Key Descriptors
Modern Digital	”shot on Sony A7IV, clean sharp, high dynamic range”
2000s Digicam	”early digital camera, slight noise, flash photography, candid, 2000s digicam style”
80s Vintage	”film grain, warm color cast, soft focus, 80s vintage photo”
Analog Film	”shot on Kodak Portra 400, natural grain, organic colors”

Modern Photorealism

2000s Digicam

80s Vintage

Modern Photorealism: “Soaking wet tiger cub taking shelter under a banana leaf in the rainy jungle, close up photo”

2000s Digicam: “Sloth out drinking in Bangkok at night in a street full of party folks, 2000s digicam style, people in the background fading”

80s Vintage: “A group of baby penguins in a trampoline park, having the time of their lives, 80s vintage photo”

Camera and Lens Simulation

Be specific about camera settings for authentic results:

Shot on Hasselblad X2D, 80mm lens, f/2.8, natural lighting

Canon 5D Mark IV, 24-70mm at 35mm, golden hour, shallow depth of field

Multi-Reference Image Editing

*[pro] API has a 9MP total limit for input+output. At 1MP output you can use up to 8 reference images, at 2MP output up to 7, and so on.

Multi-reference is powerful for:

• Fashion shoots: Combine clothing items into styled outfits
• Interior design: Place furniture and decor in rooms
• Product composites: Combine multiple products in scenes
• Character consistency: Maintain identity across variations

Fashion Editorial Example (8 references)

Prompt: “A spiritual architectural photograph captured on expired Kodak Ektachrome 64 slide film cross-processed from 1987 with a 35mm spherical lens at f/5.6, featuring model standing before small forest chapel in clearing. The model wears the outfit, positioned on stone steps leading to wooden chapel, red creating stark contrast against weathered brown timber. Background shows traditional Schwarzwald chapel - dark wood construction with small bell tower, carved wooden door, religious paintings under eaves, surrounding clearing with wild flowers, tall firs creating natural cathedral, small cemetery with wooden crosses. Dappled forest light at 1/125. Cross-processed Ektachrome showing extreme color shifts - cyan-magenta split, warm wood tones pushed to orange-brown, oversaturated red, crushed black shadows, blown highlights, heavy grain creating mysterious atmosphere. Composition emphasizes sacred spaces and pilgrimage. Thomas Struth church interiors, Candida Höfer architectural documentation, religious tourism meets fashion editorial, spiritual Schwarzwald mysticism.”

8 input references (clothing, accessories, style elements) → combined output

For multi-reference editing, describe how each input should be used. The model combines clothing items, accessories, and style references into a cohesive scene based on your prompt.

Prompt Upsampling

FLUX.2 offers a prompt_upsampling parameter that automatically enhances your prompt for better results. This is useful for:

• Quick iterations without crafting detailed prompts
• Exploring creative variations
• When you have a basic concept but want richer output

Prompt upsampling adds detail and context to your prompt automatically. Your original intent is preserved while the model expands on visual elements.

Best Practices Summary

Structure for Control

Use JSON structured prompts when you need precise control over multiple elements. Start simple and add complexity as needed.

Be Specific with Colors

Always associate hex codes with specific objects. “The car is #FF0000” works better than “use red #FF0000 in the image.”

Describe What You Want

FLUX.2 has no negative prompts. Instead of “no blur,” say “sharp focus throughout.” Instead of “no people,” describe an “empty scene.”

Reference Camera and Style

For photorealism, specify camera models, lenses, and film stocks. “Shot on Fujifilm X-T5, 35mm f/1.4” produces more authentic results than “professional photo.”

Use Native Languages

Prompt in the language that best describes your desired cultural context. French for Parisian scenes, Japanese for anime styles.

Layer Multi-Reference Carefully

When using multiple input images, clearly describe the role of each: subject from image 1, style from image 2, background from image 3.

Quick Reference

Technique	When to Use	Key Syntax
JSON Prompts	Complex scenes, automation	{"scene": "...", "style": "..."}
Hex Colors	Brand work, precise matching	color #FF5733 or hex #FF5733
Camera References	Photorealism	shot on [camera], [lens], [settings]
Style Eras	Period-specific looks	80s vintage, 2000s digicam
Multi-Reference	Composite images	[pro]: 8, [flex]: 10, [dev]: ~6

Try FLUX.2

Test these prompting techniques in the playground before integrating.