Flux Best Practices

FLUX Best Practices

Use this skill when generating prompts for any BFL FLUX model to ensure optimal image quality and accurate prompt interpretation.

When to Use

• Creating prompts for FLUX.2 or FLUX.1 models
• Text-to-image (T2I) generation
• Image-to-image (I2I) editing with FLUX.2 models
• Structured scene generation with JSON
• Typography and text rendering
• Multi-reference style transfer
• Color-accurate brand generations

Quick Reference

Prompt Structure Formula

[Subject] + [Action/Pose] + [Style/Medium] + [Context/Setting] + [Lighting] + [Camera/Technical]

Model Selection

Critical Rules

1. NO negative prompts - FLUX does not support negative prompts; describe what you want 2. Be specific - Vague prompts produce mediocre results 3. Use natural language - Prose/narrative style works best 4. Specify lighting - Lighting has the biggest impact on quality 5. Quote text - Use "quoted text" for typography rendering 6. Hex colors - Use #RRGGBB format with color description

Related

For API integration (endpoints, polling, webhooks), see the bfl-api skill.

Rules Reference

Read individual rule files for detailed guidance:

• rules/core-principles.md - Universal FLUX prompting principles
• rules/flux2-models.md - FLUX.2 family: klein, max, pro, flex, dev
• rules/flux1-models.md - FLUX.1 family: older generation of FLUX.2 models - pro, Kontext, Fill
• rules/t2i-prompting.md - Text-to-image prompting patterns
• rules/i2i-prompting.md - Image-to-image editing with FLUX.2
• rules/json-structured-prompting.md - Complex scene composition
• rules/hex-color-prompting.md - Precise color specification
• rules/typography-text.md - Text rendering and typography
• rules/multi-reference-editing.md - Multi-image references
• rules/negative-prompt-alternatives.md - Positive alternatives
• rules/model-selection-guide.md - Choosing the right model

Example Prompt

A weathered fisherman in his 70s with deep wrinkles and a salt-and-pepper beard,
wearing a navy cable-knit sweater, standing at the helm of his wooden boat.
Golden hour sunlight from the left creates dramatic rim lighting on his profile.
Shot on Hasselblad with 85mm lens at f/2.8, shallow depth of field with harbor
lights creating soft bokeh in the background. Kodak Portra 400 color science.

FLUX Best Practices

Version 1.0.0 Black Forest Labs January 2026

Note:

This document is for AI agents and LLMs to follow when working with

FLUX image generation prompting and workflows. Humans may also find it useful,

but guidance here is optimized for automation and consistency.

---

Abstract

Comprehensive prompting and workflow guide for BFL FLUX image generation models. Covers all FLUX.2 and FLUX.1 models including text-to-image, image-to-image editing, JSON structured prompting, color specification, typography, and multi-reference editing. Each rule includes detailed explanations, examples of effective vs ineffective approaches, and model-specific guidance.

---

Table of Contents

1. Core Principles - CRITICAL

• 1.1 Core FLUX Prompting Principles

2. Model Selection - HIGH

• 2.1 FLUX Model Selection Guide
• 2.2 FLUX.1 Model Family
• 2.3 FLUX.2 Model Family

3. Text-to-Image Prompting - HIGH

• 3.1 Text-to-Image (T2I) Prompting

4. Image-to-Image Editing - HIGH

• 4.1 Image-to-Image (I2I) Prompting

5. JSON Structured Prompting - MEDIUM-HIGH

• 5.1 JSON Structured Prompting

6. Color Specification - MEDIUM

• 6.1 Hex Color Prompting

7. Typography and Text - MEDIUM

• 7.1 Typography and Text Prompting

8. Multi-Reference Editing - MEDIUM

• 8.1 Multi-Reference Image Editing

9. Positive Prompt Alternatives - MEDIUM

• 9.1 Negative Prompt Alternatives

---

1. Core Principles

Impact: CRITICAL

Universal prompting principles that apply to all FLUX models. Master these before diving into specific techniques.

1.1 Core FLUX Prompting Principles

Impact: MEDIUM

These principles apply to all FLUX models and form the foundation of effective prompting.

FLUX does NOT support negative prompts. Always describe what you WANT, not what you don't want.

Build prompts using this structure for consistent results:

Wrong Approach:

a portrait of a woman, no glasses, no hat, no makeup

Correct Approach:

a portrait of a woman with natural skin, clear face, bare head, visible eyes

Example:

A young woman with flowing auburn hair (subject)
dancing gracefully in mid-leap (action)
in the style of classical oil painting (style)
in a moonlit garden with roses (context)
soft diffused moonlight with subtle rim lighting (lighting)
medium shot, shallow depth of field (technical)

More specific prompts yield dramatically better results.

Vague: Poor Results

a cat sitting

Specific: Excellent Results

A fluffy orange tabby cat with bright green eyes sitting regally on a vintage
velvet armchair, afternoon sunlight streaming through lace curtains, warm
golden hour lighting, shallow depth of field, shot on medium format film

Write prompts as descriptive prose rather than keyword lists.

Keyword Style: Less Effective

woman, portrait, beautiful, blonde, studio, professional, 8k, detailed

Prose Style: More Effective

A professional studio portrait of a beautiful blonde woman in her thirties,
captured with soft studio lighting that accentuates her features, rendered
in stunning detail with natural skin texture and subtle catchlights in her eyes

Always specify lighting - it has the single greatest impact on image quality.

Natural Lighting:

• Golden hour - warm, soft, directional

• Overcast - soft, diffused, even

• Harsh midday - high contrast, strong shadows

• Dappled forest light - specular, organic patterns

Studio Lighting:

• Softbox - even, professional

• Rim light - edge definition, separation

• Butterfly lighting - beauty, glamour

• Rembrandt lighting - dramatic, classic portraits

Atmospheric Lighting:

• Volumetric fog - depth, mystery

• God rays - dramatic, spiritual

• Neon glow - urban, cyberpunk

• Candlelight - warm, intimate

Mood-Based Lighting:

• Dramatic shadows - tension, noir

• High key - bright, airy, clean

• Low key - moody, mysterious

• Chiaroscuro - strong contrast, painterly

FLUX prioritizes elements that appear earlier in the prompt. Front-load important elements.

Less Effective:

A forest background with soft lighting where a knight in shining armor stands

More Effective:

A knight in shining armor stands in a forest, soft dappled lighting filtering
through the canopy

Optimal prompt length is typically 30-80 words (FLUX can handle up to 512 tokens).

• Too short: Lacks direction, generic results

• Too long: Can become unfocused

• Sweet spot: Enough detail to guide, not so much it confuses

Build prompts iteratively:

1. Start with core subject and action

2. Add style and medium

3. Specify lighting and atmosphere

4. Include technical details

5. Refine based on results

Change one element at a time to understand what affects your output.

Reference: negative-prompt-alternatives.md

---

2. Model Selection

Impact: HIGH

Choosing the right FLUX model for your use case. Covers both FLUX.2 (latest) and FLUX.1 (legacy) model families.

2.1 FLUX Model Selection Guide

Impact: MEDIUM

Decision guide for selecting the optimal FLUX model based on your requirements.

| Priority | Recommended Model |

| ------------- | ------------------------------- |

| Speed | FLUX.2 [klein] |

| Quality | FLUX.2 [max] |

| Balance | FLUX.2 [pro] |

| Typography | FLUX.2 [flex] |

| Image Editing | FLUX.2 [klein], [pro], or [max] |

| Local/Free | FLUX.2 [dev] |

| Inpainting | FLUX.1 Fill |

Note: All FLUX.2 models natively support image-to-image editing via reference images. Simply provide your source image(s) as references and describe the desired changes.

By Speed:

| Model | Relative Speed | Best For |

| ----------------- | -------------- | ----------------------- |

| FLUX.2 [klein] 4B | Fastest | Rapid prototyping |

| FLUX.2 [klein] 9B | Very Fast | Better quality previews |

| FLUX.2 [pro] | Medium | Production workflows |

| FLUX.2 [flex] | Medium | Typography tasks |

| FLUX.2 [max] | Slower | Final hero images |

By Quality:

| Model | Quality Level | Trade-off |

| ----------------- | ------------- | -------------------------- |

| FLUX.2 [max] | Highest | Slowest, most expensive |

| FLUX.2 [pro] | High | Good balance |

| FLUX.2 [flex] | High (text) | Specialized for typography |

| FLUX.2 [klein] 9B | Good | Fast, slightly less detail |

| FLUX.2 [klein] 4B | Moderate | Fastest, preview quality |

By Cost:

Credit pricing: 1 credit = $0.01 USD. FLUX.2 uses megapixel-based pricing.

| Model | 1st MP | +MP | 1MP T2I | 1MP I2I | Volume Recommendation |

| ----------------- | ------ | ---- | ------- | ------- | --------------------------- |

| FLUX.2 [klein] 4B | 1.4c | 0.1c | $0.014 | $0.015 | High volume, previews |

| FLUX.2 [klein] 9B | 1.5c | 0.2c | $0.015 | $0.017 | High volume, better quality |

| FLUX.2 [pro] | 3c | 1.5c | $0.03 | $0.045 | Production workloads |

| FLUX.2 [max] | 7c | 3c | $0.07 | $0.10 | Hero images, premium |

| FLUX.2 [flex] | 6c | 6c | $0.06 | $0.12 | Typography |

| FLUX.2 [dev] | - | - | Free | Free | Local dev (non-commercial) |

Pricing formula: (firstMP + (outputMP-1) * mpPrice) + (inputMP * mpPrice) in cents

| Model | Price/Image | Use Case |

| -------------------- | ----------- | ----------------------- |

| FLUX.1 Kontext [pro] | $0.04 | Context-aware editing |

| FLUX.1 Kontext [max] | $0.08 | Max quality editing |

| FLUX1.1 [pro] | $0.04 | Standard T2I |

| FLUX1.1 [pro] Ultra | $0.06 | Ultra high-resolution |

| FLUX1.1 [pro] Raw | $0.06 | Candid photography feel |

| FLUX.1 Fill [pro] | $0.05 | Inpainting |

| FLUX.1 [pro] | $0.05 | Original pro model |

Use bfl.ai/pricing calculator for exact costs at different resolutions.

Creative Exploration / Ideation:

Recommended: FLUX.2 [klein]

• Fast iterations

• Quick concept testing

• Mood board generation

• Exploring prompt variations

Production Marketing Assets:

Recommended: FLUX.2 [pro]

• Consistent quality

• Reasonable speed

• Cost-effective at scale

• Reliable for automation

Hero Images / Premium Content:

Recommended: FLUX.2 [max]

• Maximum detail

• Best coherence

• Supports grounding search

• Worth the premium for key visuals

Typography / Signage / Posters:

Recommended: FLUX.2 [flex]

• Superior text rendering

• Adjustable quality settings

• Best for readable text

• UI mockups and infographics

Character Consistency:

Recommended: FLUX.2 [max] or [pro]

• Multi-reference support (up to 8-10 images)

• Best editing consistency

• Maintains identity across scenes

• Superior quality over FLUX.1 Kontext

Photo Editing / Retouching:

Recommended: FLUX.2 [klein], [pro], or [max]

• Native image-to-image support via references

• Style transfer

• Object modification

• Attribute changes

• Better results than FLUX.1 Kontext

Real-Time Information:

Recommended: FLUX.2 [max]

• Grounding search feature

• Current events

• Recent news visualization

• Weather/location data

Local Development / Testing:

Recommended: FLUX.2 [dev]

• No API costs

• Full control

• Fine-tuning experiments

• Offline capability

Editorial with Typography:

1. FLUX.2 [max] - Generate base image (highest quality)
2. FLUX.2 [flex] - Add text overlay pass

Character-Consistent Series:

1. FLUX.2 [max] - Create character reference
2. FLUX.2 [max]/[pro] - Generate consistent variations using reference images
3. FLUX.2 [klein] - Quick iteration on variations if needed

E-commerce Product Pipeline:

1. FLUX.2 [pro] - Bulk product generations
2. FLUX.2 [pro]/[klein] - Product variations (colors, angles) using references
3. FLUX.2 [flex] - Add promotional text/pricing

Limited Budget:

• High volume: FLUX.2 [klein] 4B

• Quality needed: FLUX.2 [pro] (best value)

Tight Deadline:

• Any task: FLUX.2 [klein]

• Quality matters: FLUX.2 [pro]

Maximum Quality Required:

• Always: FLUX.2 [max]

Text Must Be Readable:

• Always: FLUX.2 [flex]

Editing Existing Images:

• Fast edits: FLUX.2 [klein] with reference images

• Quality edits: FLUX.2 [max] or [pro] with reference images

• Alternative: FLUX.1 Kontext (FLUX.2 preferred)

Rate Limit Sensitivity:

Speed?      → FLUX.2 [klein]
Quality?    → FLUX.2 [max]
Balance?    → FLUX.2 [pro]
Text?       → FLUX.2 [flex]
Edit?       → FLUX.2 [klein/pro/max] with reference images
Free?       → FLUX.2 [dev]

• Prefer: FLUX.2 models (24 concurrent limit)

• Avoid: FLUX.1 Kontext Max (6 concurrent limit)

Key insight: All FLUX.2 models support image editing natively via reference images. FLUX.2 is recommended over FLUX.1 Kontext for editing tasks.

2.2 FLUX.1 Model Family

Impact: MEDIUM

Tip: FLUX.2 models are the latest generation and recommended for most use cases. FLUX.1 models are still available for specific needs.

Guide to FLUX.1 models and their specialized capabilities.

| Model | Purpose | Notes |

|-------|---------|-------|

| FLUX1.1 [pro] | Text-to-image | FLUX.2 [pro] offers improved results |

| FLUX.1 Kontext | Image-to-image | FLUX.2 with references recommended |

| FLUX.1 Kontext Max | Image-to-image | FLUX.2 [max] with references recommended |

| FLUX.1 Fill | Inpainting | Useful for specific inpainting tasks |

Fast and reliable text-to-image generation.

Characteristics:

• Strong prompt adherence

• Production-grade architecture

• Consistent, reliable results

• Scalable for high-volume

• Pricing: $0.04 per image

Prompting Style:

Standard descriptive prompts with clear subject and style specification.

Example Prompt:

A golden retriever puppy playing in autumn leaves, warm afternoon sunlight,
shallow depth of field with bokeh background, joyful expression, professional
pet photography style

Recommendation: FLUX.2 models with reference images provide improved editing results.

Context-aware image-to-image editing model for transformations and modifications.

Characteristics:

• Understands image context

• Preserves unedited regions

• Style transfer capabilities

• Object modification

• Basic to complex transformations

Prompting Strategies:

Add the text "OPEN" as a neon sign in the window, red glowing letters
with slight reflection on the glass, matching the nighttime atmosphere

Simple, direct instructions work best:

Be explicit about what to preserve:

For dramatic changes, be specific about preservation:

Reference specific artistic movements:

Describe text placement and integration:

Tips for Kontext:

• Be explicit about what should NOT change

• Start with simpler edits, build complexity

• Specify style preservation when needed

• Use for incremental refinement

Advanced multi-reference editing for complex compositions.

Characteristics:

• Handles up to 10 reference images

• Best editing consistency across references

• Complex scene composition

• Character consistency maintenance

• Rate limit: 6 concurrent requests

Multi-Reference Prompting:

Replace the top half of the person in image 1 with the clothing
from image 2, maintaining the pose and background

Describe relationships between images naturally:

Reference images by number for precision:

Tips for Kontext Max:

• Plan your reference images carefully

• Use natural language for relationships

• Specify which elements come from which image

• Leverage for character consistency across scenes

Specialized tool for object removal and area completion.

Characteristics:

• Clean object removal

• Intelligent background completion

• Texture-aware filling

• Seamless blending

Use Cases:

• Remove unwanted objects from photos

• Complete partial images

• Replace specific regions

• Clean up image artifacts

Prompting for Fill:

Complete with matching ocean waves and sandy beach

Describe what should fill the masked area:

Tips for Fill:

• Provide context about surrounding areas

• Specify texture and pattern continuation

• Describe lighting consistency

• Use for cleanup and removal tasks

2.3 FLUX.2 Model Family

Impact: MEDIUM

Complete guide to FLUX.2 variants and their optimal prompting strategies.

Key Feature: All FLUX.2 models natively support both text-to-image generation AND image-to-image editing via reference images. There's no need to use legacy FLUX.1 Kontext models for editing tasks.

| Model | Parameters | Best For | Speed | Reference Images |

|-------|------------|----------|-------|------------------|

| [klein] | 4B/9B | Fast iterations, previews, quick edits | Fastest | Up to 4 |

| [max] | - | Highest quality generation & editing | Slowest | Up to 8-10 |

| [pro] | - | Production balanced | Medium | Up to 8 |

| [flex] | - | Typography, text rendering | Medium | Up to 8 |

| [dev] | - | Local development | Varies | Varies |

Best for rapid prototyping, previews, and high-volume generation.

Characteristics:

• 4B or 9B parameter versions available

• Sub-second generation times

• Optimized for speed over maximum detail

• No prompt upsampling - be descriptive yourself

• Supports up to 4 reference images

Prompting Style: Narrative Prose:

Klein responds best to descriptive, narrative-style prompts with emphasis on lighting and atmosphere.

Example Prompt:

A cozy coffee shop interior bathed in warm afternoon light, steam rising lazily
from ceramic cups, worn leather armchairs arranged around small wooden tables,
bookshelves lining exposed brick walls, the soft atmosphere of a quiet afternoon
with dust motes floating in sunbeams through tall windows

Tips for [klein]:

• Write like a novelist describing a scene

• Front-load your subject (word order critical)

• Emphasize lighting descriptions heavily

• Keep prompts moderately detailed (40-70 words)

Premium model for final production assets and maximum detail.

Characteristics:

• Highest detail and coherence

• Best editing consistency

• Vast world knowledge

• Includes grounding search (real-time web data)

• Strongest prompt following

• Supports up to 8 reference images (API), 10 (playground)

Prompting Style: Technical + Descriptive:

[max] excels with detailed technical specifications combined with descriptive prose.

Example Prompt:

Portrait of a weathered fisherman, age 70, deep wrinkles telling stories of
decades at sea, salt-and-pepper beard with streaks of white, wearing a navy
cable-knit sweater with visible wool texture. Shot on Hasselblad X2D with
90mm f/2.8 lens at f/4, golden hour natural light from the left creating
strong rim lighting, shallow depth of field with soft bokeh from harbor
lights behind, Kodak Portra 400 color science with natural grain

Tips for [max]:

• Include camera and lens specifications for photorealism

• Specify film stock or digital sensor characteristics

• Use technical photography terms (aperture, focal length)

• Leverage grounding search for current events: "news photo of [recent event]"

Optimal balance of quality and speed for production workflows.

Characteristics:

• Good quality-to-speed ratio

• Reliable, consistent output

• Suitable for batch processing

• Supports prompt upsampling

• Supports up to 8 reference images

Prompting Style: Balanced Detail:

Standard detailed prompts work well without excessive technical specification.

Example Prompt:

A modern minimalist living room with floor-to-ceiling windows overlooking
a city skyline at dusk, clean white furniture with subtle textures, a single
statement plant in the corner, warm ambient lighting from hidden sources,
architectural photography style with clean lines and balanced composition

Tips for [pro]:

• Balance specificity with generation speed

• Good for template-based prompt systems

• Enable prompt upsampling for enhanced results

• Consistent quality for production pipelines

Optimized for text rendering and typographic content.

Characteristics:

• Superior text rendering quality

• Handles multiple text elements

• Adjustable steps (1-50) and guidance (1.5-10)

• Best for signage, posters, UI mockups

• Supports up to 8 reference images

Prompting Style: Typography-Focused:

Always quote text and specify font characteristics explicitly.

Example Prompt:

A modern minimalist poster design with the headline "DESIGN SUMMIT 2025"
in bold condensed sans-serif typography centered in the upper third,
subtitle "Innovation Meets Creativity" in lighter weight below,
date "MARCH 15-17" in small caps at the bottom, all text in white
on a gradient background transitioning from deep purple #4A0080 to
coral #FF6B6B, clean geometric accent lines, professional print quality

Tips for [flex]:

• Always quote exact text: "Your Text Here"

• Specify font style: serif, sans-serif, script, display, monospace

• Describe text hierarchy: headline, subhead, body

• Include placement: centered, left-aligned, upper third

• Adjust steps (higher = better quality) and guidance (higher = stricter)

For local development, testing, and non-commercial use.

Characteristics:

• Open weights on Hugging Face

• Runs locally (~13GB VRAM recommended)

• Full customization available

• Free for non-commercial use

• Base variants available (undistilled) for fine-tuning

Prompting Style: Standard:

Same prompting patterns as [pro] work well.

Tips for [dev]:

• Use for development and testing before production

• Experiment with prompt variations

• Good for fine-tuning experiments

• Check license for commercial use restrictions

All FLUX.2 models support image editing via reference images. This replaces the need for legacy FLUX.1 Kontext models.

How It Works:

1. Provide your source image(s) as reference images

2. Describe the desired changes in your prompt

3. The model preserves context while applying edits

Example: Style Transfer:

Reference: [your source image]
Prompt: Transform this image into a watercolor painting style,
maintaining the exact composition and subject positioning

Example: Object Modification:

Reference: [your source image]
Prompt: Change the car color to red while keeping everything else identical

Example: Character Consistency:

Reference: [character reference image]
Prompt: The same person from the reference image walking through
a busy Tokyo street at night, neon lights reflecting on wet pavement

Model Selection for Editing:

| Use Case | Recommended Model |

|----------|-------------------|

| Quick iterations/previews | FLUX.2 [klein] |

| Production quality edits | FLUX.2 [pro] |

| Maximum quality/complex edits | FLUX.2 [max] |

| Text/typography edits | FLUX.2 [flex] |

---

3. Text-to-Image Prompting

Impact: HIGH

Crafting effective prompts for generating images from text descriptions.

3.1 Text-to-Image (T2I) Prompting

Impact: MEDIUM

Comprehensive guide to crafting effective text-to-image prompts for FLUX models.

Basic Formula:

[Subject] + [Action] + [Style] + [Context] + [Lighting] + [Technical]

Expanded Framework:

[Main Subject] - who/what is the focus
[Attributes] - characteristics, details, clothing
[Action/Pose] - what they're doing
[Environment] - where, setting, background
[Style/Medium] - artistic approach
[Lighting] - light source, quality, mood
[Composition] - framing, camera angle
[Technical] - camera, lens, film stock

People/Portraits:

A distinguished professor in his 60s with silver hair and round spectacles,
wearing a tweed jacket with leather elbow patches, deep-set thoughtful eyes,
slight smile suggesting hidden wisdom

Animals:

A majestic snow leopard with piercing blue-grey eyes, thick spotted fur
dusted with snowflakes, powerful muscular build, alert posture on a
rocky outcrop

Objects/Products:

A vintage Leica M3 camera with worn brass edges showing decades of use,
black leather covering with patina, sitting on weathered wooden table

Landscapes:

A dramatic fjord at dawn, steep granite cliffs rising from mirror-still
water, wisps of morning mist, distant snow-capped peaks catching first
golden light

Architecture:

A brutalist concrete apartment building in late afternoon light, geometric
shadows creating abstract patterns, warm sunlight contrasting with cool
grey concrete

Photorealistic:

80s film photography, film grain, warm color cast, soft focus, nostalgic

Artistic Styles:

anime style, large expressive eyes, clean linework, cel shading, vibrant palette

Portrait Lighting:

Rembrandt lighting - 45 degree key light creating triangle shadow on cheek
Butterfly lighting - overhead key creating shadow under nose
Split lighting - 90 degree side light, half face in shadow
Loop lighting - slight angle creating small nose shadow

Natural Lighting:

Golden hour - warm, soft, directional light 1 hour before sunset
Blue hour - cool, ambient light just after sunset
Overcast - soft, even, diffused lighting
Harsh midday - strong contrast, defined shadows

Atmospheric:

Volumetric light - visible light rays through fog/dust
Rim lighting - backlight creating edge glow
Practical lighting - visible light sources in scene
Neon glow - colorful artificial urban lighting

Camera Bodies:

Shot on Hasselblad X2D - medium format, exceptional detail
Shot on Canon 5D Mark IV - professional DSLR quality
Shot on Leica M10 - rangefinder character, smooth tonality
Shot on iPhone 15 Pro - computational photography look

Lens Characteristics:

85mm f/1.4 - classic portrait, creamy bokeh
24mm f/2.8 - wide angle, environmental
50mm f/1.2 - natural perspective, shallow DOF
135mm f/2 - compressed perspective, smooth background
Macro lens - extreme close-up detail
Tilt-shift lens - miniature effect or architectural correction

Technical Settings:

f/1.4 - extremely shallow depth of field
f/2.8 - moderate background blur
f/8 - sharp throughout, landscape
f/16 - maximum sharpness, long exposure
ISO 100 - clean, no noise
ISO 3200 - visible grain, low light

Framing:

extreme close-up - filling frame with detail
close-up - head and shoulders
medium shot - waist up
full shot - entire body
wide shot - subject in environment
establishing shot - location focus

Angles:

eye level - natural, relatable
low angle - powerful, imposing
high angle - diminished, overview
Dutch angle - tension, unease
bird's eye - pattern, layout
worm's eye - dramatic upward view

Composition Rules:

rule of thirds - subject at intersection points
centered composition - symmetry, stability
leading lines - guiding eye to subject
frame within frame - natural framing elements
negative space - minimalist, breathing room

Editorial Portrait:

A fashion editorial portrait of a young woman with striking features and
high cheekbones, wearing an avant-garde geometric collar in silver, dramatic
side lighting creating strong shadows, shot on Hasselblad with 100mm lens
at f/2.8, studio background with subtle gradient, high fashion magazine style

Product Photography:

A premium wireless headphone product shot, matte black finish with rose gold
accents, floating at slight angle against pure white background, soft even
lighting eliminating harsh shadows, reflection visible on glossy surface below,
commercial catalog style, ultra sharp focus throughout

Landscape:

A misty morning in ancient redwood forest, towering trees disappearing into
fog above, ferns covering forest floor in layers of green, single shaft of
golden sunlight breaking through canopy, shot on large format camera, rich
detail in bark textures, Ansel Adams inspired black and white with deep tones

Architectural:

Modern minimalist beach house at golden hour, floor-to-ceiling glass walls
reflecting sunset colors, clean white concrete and natural wood, infinity
pool merging with ocean horizon, architectural photography style, wide angle
showing full structure, warm evening light

---

4. Image-to-Image Editing

Impact: HIGH

Techniques for editing and transforming existing images using FLUX.2 models.

4.1 Image-to-Image (I2I) Prompting

Impact: MEDIUM

Guide to effective image-to-image editing with FLUX models.

All FLUX.2 models support image-to-image editing via reference images:

• FLUX.2 [klein]: Up to 4 reference images - fast editing

• FLUX.2 [pro]: Up to 8 reference images - balanced quality/speed

• FLUX.2 [max]: Up to 8-10 reference images - highest quality editing

• FLUX.2 [flex]: Up to 8 reference images - best for typography edits

Simply provide your source image as a reference and describe the desired changes. The model understands image context and can modify specific elements while preserving others.

Note: FLUX.2 models are recommended for image editing. They provide better results than the older FLUX.1 Kontext models.

Preferred: Use URLs directly - simpler and more convenient than base64.

When you have an image URL, pass it directly to input_image:

The API fetches URLs automatically. Both URL and base64 work, but URLs are recommended when available.

Simple Modifications:

Add snow to the ground

Direct, single-change instructions:

Attribute Changes:

Age the person to appear 20 years older

Modifying specific characteristics:

Explicit Preservation:

Transform the daytime photo to nighttime, maintaining the exact
composition, colors of the subject's outfit, and lighting direction

When you need to keep specific elements unchanged:

Style Preservation:

Add rain effects to the scene while preserving the painting's
impressionist brushwork and color palette

Preventing unwanted style shifts:

Environmental Changes:

Change to spring with cherry blossoms, fresh green leaves,
soft warm lighting

Style Transfer:

Transform into a watercolor painting with soft edges,
transparent washes, and paper texture visible

Subject Modifications:

Make the subject appear younger, around 25 years old

Object Editing:

Swap the coffee mug for an ornate teacup with floral pattern

Adding Text:

Include a wooden sign with hand-painted text "Welcome Home"
mounted above the door

Modifying Text:

Update the poster text to "SUMMER SALE 2025" maintaining the design

For dramatic transformations, consider breaking into steps:

Step-by-Step Approach:

Transform this modern office into a Victorian library with completely
different furniture, add a fireplace, change the lighting to candlelit,
and age the photograph

Instead of:

Try sequential edits:

1. Change the furniture style to Victorian antique pieces

2. Add a stone fireplace on the right wall

3. Transform lighting to warm candlelit atmosphere

4. Apply vintage photograph aesthetic with sepia tones

Avoid Vague Instructions:

Bad: Make it look better
Good: Increase contrast, add warm color grading, sharpen details

Be Specific About Scope:

Bad: Change the background
Good: Replace the office background with a tropical beach at sunset,
      maintaining the subject's exact position and lighting direction

Explicit Style Preservation:

Bad: Make it nighttime
Good: Transform to nighttime while maintaining the photorealistic style,
      add appropriate artificial lighting sources

1. Start Simple - Begin with single-element changes

2. Be Explicit - State what should change AND what should stay

3. Reference Context - Mention existing elements when relevant

4. Iterate - Refine through multiple small edits rather than one large one

5. Preserve Deliberately - Always specify style/composition preservation needs

---

5. JSON Structured Prompting

Impact: MEDIUM-HIGH

Using structured JSON for complex multi-element scene composition.

5.1 JSON Structured Prompting

Impact: MEDIUM

For complex scenes with multiple elements, spatial relationships, or production automation, use JSON-structured prompts.

• Multiple characters with distinct attributes

• Precise spatial positioning

• Complex scene composition

• Reproducible, template-based prompts

• Programmatic prompt generation

• Production workflows with variable substitution

Flatten your JSON into flowing prose for the actual prompt:

From JSON:

{
  "subjects": [
    {
      "type": "person",
      "description": "elderly craftsman with weathered hands",
      "position": "seated at workbench",
      "action": "carefully carving wood"
    }
  ],
  "scene": { "setting": "traditional workshop", "time": "morning" },
  "technical": { "lighting": "natural window light from right" }
}

To Prompt:

{
  "composition": {
    "layout": "triangular",
    "focal_point": "center-left intersection",
    "depth_layers": [
      {
        "layer": "foreground",
        "elements": ["flowers in vase"],
        "focus": "soft blur"
      },
      {
        "layer": "midground",
        "elements": ["main subject"],
        "focus": "sharp"
      },
      {
        "layer": "background",
        "elements": ["window", "garden view"],
        "focus": "soft blur"
      }
    ]
  }
}

Use JSON structure for template-based generation:

Define explicit spatial relationships:

1. Use IDs for References - Give subjects IDs when they interact

2. Separate Concerns - Keep scene, subjects, style, and technical distinct

3. Be Consistent - Use the same terminology throughout

4. Include All Details - Don't assume, specify everything

5. Flatten for Execution - Convert to natural language before sending to model

6. Version Templates - Track template versions for reproducibility

---

6. Color Specification

Impact: MEDIUM

Precise color control using hex codes for brand-accurate generations.

6.1 Hex Color Prompting

Impact: MEDIUM

FLUX supports hex color codes (#RRGGBB) for precise color specification, essential for brand consistency and exact color matching.

Include hex codes directly in your prompt with descriptive names:

Use these keywords to indicate color specification:

1. Always Pair with Description:

Good: #FF6B6B (coral pink)
Bad: #FF6B6B

Never use hex codes alone - include the color name:

2. Associate with Specific Objects:

A product shot featuring a smartphone with a #1DA1F2 (Twitter blue) case,
resting on a #14171A (near black) matte surface

Clearly connect colors to their targets:

3. Limit Color Palette:

Color palette for the scene: #2ECC71 (emerald green), #3498DB (sky blue),
#F1C40F (sunflower yellow), #FFFFFF (pure white)

3-5 colors typically work best. Too many can confuse the model:

Brand Colors:

Corporate office reception with brand colors prominently featured:
walls in #0066CC (company blue), accent furniture in #FF6600 (company orange),
logo displayed in #FFFFFF (white) against the blue backdrop

Interior Design:

Scandinavian minimalist bedroom with #F5F5F5 (warm white) walls,
#8B4513 (saddle brown) wooden headboard and nightstands,
#708090 (slate gray) linen bedding, and #DAA520 (goldenrod) accent lamp

Fashion:

Editorial fashion photo: model wearing #000000 (black) cashmere turtleneck,
#FF4500 (orange-red) wide-leg wool pants, #C0C0C0 (silver) geometric earrings,
against a #F0F0F0 (light gray) studio backdrop

Product Design:

Premium headphones product shot: #1C1C1E (space gray) aluminum body,
#F5F5F7 (silver) mesh ear cups, #FF9500 (iOS orange) accent ring around controls

Digital Art:

Synthwave cityscape: #FF00FF (magenta) and #00FFFF (cyan) neon signs,
#1A1A2E (deep navy) night sky, #E94560 (hot pink) setting sun on horizon,
#16213E (dark blue) building silhouettes, rain-slicked streets reflecting lights

Data Visualization:

Sunset sky gradient from #FF6B6B (coral) at horizon through
#FFA07A (light salmon) to #87CEEB (sky blue) at top

Specify gradients with start and end colors:

Complementary: Opposite on color wheel

Scene using complementary colors: #3498DB (blue) dominant with
#E67E22 (orange) accents for visual pop

Analogous: Adjacent colors

Harmonious palette using analogous colors: #9B59B6 (purple),
#8E44AD (deep purple), #3498DB (blue) - flowing naturally together

Triadic: Evenly spaced

Vibrant triadic scheme: #E74C3C (red), #F1C40F (yellow),
#3498DB (blue) - balanced and dynamic

Monochromatic: Single hue variations

# Social Media
Twitter/X Blue: #1DA1F2
Facebook Blue: #1877F2
Instagram Gradient: #833AB4 to #FD1D1D
LinkedIn Blue: #0A66C2

# Tech
Apple Gray: #1C1C1E
Google Blue: #4285F4
Microsoft Blue: #00A4EF
Amazon Orange: #FF9900

# Design
Figma Purple: #A259FF
Dribbble Pink: #EA4C89
Behance Blue: #1769FF

For reference only - always verify current brand guidelines:

Color Not Accurate:

• Add the color name alongside hex

• Specify the exact object the color applies to

• Use fewer total colors in the prompt

Color Bleeding:

• Clearly delineate which objects get which colors

• Use spatial descriptions: "the LEFT chair in #color"

Muddy Colors:

• Check hex code accuracy

• Specify lighting that won't shift colors

• Use "maintaining exact color #XXXXXX" for emphasis

---

7. Typography and Text

Impact: MEDIUM

Rendering text and typography within generated images.

7.1 Typography and Text Prompting

Impact: MEDIUM

Guide to rendering text in FLUX images. Use FLUX.2 [flex] for best typography results.

Always quote the exact text you want rendered:

1. Use Quotation Marks:

Correct: A poster with "HELLO WORLD" in bold letters
Wrong: A poster with HELLO WORLD in bold letters

2. Specify Font Style:

"ADVENTURE" in bold sans-serif typography
"Welcome" in elegant cursive script
"CHAPTER ONE" in classic serif typeface
"CODE" in monospace terminal font
"SALE!" in decorative display lettering

3. Describe Size Hierarchy:

Large headline "BREAKING NEWS" above smaller subtext "Details inside"

4. Indicate Placement:

"OPEN" sign centered in storefront window
"EXIT" text positioned above doorway
"Page 1" in bottom right corner

5. Front-Load Text:

Good: A sign reading "FRESH BREAD" in a bakery window...
Less Good: A bakery window with a sign that says "FRESH BREAD"...

Place text descriptions early in the prompt for better accuracy:

Sans-Serif (Modern/Clean):

"MINIMAL" in clean geometric sans-serif, Swiss modernist style
"TECH SUMMIT" in bold condensed grotesque typeface
"future" in thin uppercase sans-serif, contemporary design

Serif: Classic/Elegant

"The New Yorker" in traditional serif typeface, editorial masthead
"LUXURY" in high-contrast Didone serif with thin/thick strokes
"Wisdom" in old-style serif with subtle bracketed serifs

Script/Cursive (Decorative):

"With Love" in flowing calligraphic script with flourishes
"Signature" in connected brush script, casual elegance
"Romance" in formal copperplate script, wedding invitation style

Display/Decorative:

"ROCK CONCERT" in distressed vintage concert poster lettering
"CIRCUS" in ornate Victorian display type with decorative elements
"RETRO" in 1970s rounded bubble letters

Handwritten:

"Note to self" in casual handwritten style, slightly imperfect
"Thanks!" in quick marker pen handwriting
"ideas" in sketchy pencil handwriting

Monospace:

"CODE_COMPLETE" in terminal monospace, developer aesthetic
"SYSTEM" in typewriter monospace, vintage tech
"DEBUG" in LCD-style digital monospace

Neon Signs:

Glowing neon sign spelling "OPEN 24/7" in pink neon tubes with
blue outline, slight glow and reflection, night scene

Metallic/3D:

"GOLD" in three-dimensional metallic gold letters with realistic
reflections and subtle shadows, luxury aesthetic

Embossed/Debossed:

"PREMIUM" embossed into leather surface, subtle shadows showing
the raised letterforms

Outlined:

"MODERN" in outline-only letters, no fill, thin white stroke
on dark background

Gradient Text:

"SUMMER" with gradient fill from #FF6B6B (coral) at top to
#4ECDC4 (teal) at bottom

Poster Design:

Event poster with "SUMMER FEST 2025" as large headline in bold
condensed sans-serif at top, "JULY 15-17" as medium subheading
in regular weight, "Central Park, NYC" as small body text at
bottom, all in white text on #FF6B35 (sunset orange) background

Book Cover:

Book cover design: "THE GREAT GATSBY" in elegant art deco gold
lettering centered in upper third, author name "F. SCOTT FITZGERALD"
in smaller gold caps below, #1A1A2E (midnight blue) background
with geometric gold accents

Magazine Cover:

Fashion magazine cover with "VOGUE" in classic serif masthead at top,
cover line "SPRING COLLECTION" in bold sans-serif, "The New Rules of Style"
in lighter weight italic, all in white against dramatic portrait

Signage:

Vintage diner sign: "MEL'S DINER" in red neon script lettering,
"OPEN" below in separate green neon block letters, chrome border,
1950s Americana aesthetic

Business Card:

Minimalist business card with "JOHN SMITH" in medium weight sans-serif,
"Creative Director" in lighter weight below, contact details in small
type at bottom, #2C3E50 (dark blue) text on white background

Centered Composition:

Centered text layout: "WELCOME" in large caps at center,
perfectly balanced with equal margins

Left-Aligned:

Left-aligned text block: "Company Name" as header,
"Tagline goes here" below, flush left alignment

Text on Path:

"GOING IN CIRCLES" text following a circular path around
the center of the design

Text Overlay:

"ADVENTURE AWAITS" in bold white text overlaid on landscape
photograph, positioned in lower third with slight shadow for readability

Steps Parameter:

• Higher steps (30-50) = better text quality

• Lower steps (10-20) = faster, lower quality

Guidance Parameter:

• Higher guidance (6-10) = stricter prompt following

• Lower guidance (1.5-4) = more creative interpretation

Recommended Settings:

For clean typography: steps=50, guidance=7
For artistic text: steps=30, guidance=4

Misspelled Words:

• Keep text short (1-4 words work best)

• Use common words when possible

• Repeat the exact text in the prompt

Illegible Text:

• Specify larger text size

• Use simpler fonts (sans-serif)

• Ensure high contrast with background

• Use [flex] model

Wrong Font Style:

Instead of: "text in a nice font"
Use: "text in bold geometric sans-serif similar to Futura"

Be more specific:

Text Not Appearing:

• Front-load text description in prompt

• Put text in quotes

• Specify exact placement

• Reduce other prompt complexity

---

8. Multi-Reference Editing

Impact: MEDIUM

Combining multiple reference images for style transfer and composition.

8.1 Multi-Reference Image Editing

Impact: MEDIUM

Guide to using multiple reference images for character consistency, style transfer, and complex compositions.

FLUX.2 models support multiple reference images for advanced editing:

• FLUX.2 [klein]: Up to 4 reference images - fast editing

• FLUX.2 [pro]: Up to 8 via API - balanced quality/speed

• FLUX.2 [max]: Up to 8 via API, 10 in playground - highest quality

• FLUX.2 [flex]: Up to 8 via API - best for typography

Note: FLUX.2 models are recommended over FLUX.1 Kontext Max for better results.

Preferred: Use URLs directly - simpler and more convenient than base64.

Pass image URLs directly to input_image, input_image_2, etc.:

The API fetches URLs automatically. Both URL and base64 work, but URLs are recommended when available.

Natural Language Description:

The person from image 1 is sitting at the cafe table from image 2,
wearing the outfit from image 3, with the warm lighting style of image 4

Describe relationships between images naturally:

Explicit Indexing:

Combine the face from image 1 with the hairstyle from image 2
on the body pose from image 3

Reference images by number for precision:

Character Consistency:

The same person from image 1, now seated at a desk in a modern office,
same clothing and hairstyle, different environment

Maintain the same character across multiple scenes:

For sequential consistency:

Style Transfer:

Apply the color grading and mood from image 2 to the scene in image 1

Apply the style of one image to another:

Pose Guidance:

The person from image 1 in the exact pose shown in image 2,
placed in the environment from image 3

Use a reference for body positioning:

Object Composition:

Place the product from image 1 on the table setting from image 2,
using the lighting style from image 3

Combine elements from multiple images:

Background Replacement:

Keep the subject from image 1 exactly as shown, replace the background
with the beach scene from image 2, match the lighting naturally

Two Characters:

Image 1 (person A) and image 2 (person B) having a conversation
at a coffee shop table, person A on the left gesturing, person B
on the right listening intently

Group Composition:

The three people from images 1, 2, and 3 standing together for a
group photo, arranged left to right in that order, friendly poses,
outdoor park setting

Selective Attribute Transfer:

The face and expression from image 1, the hairstyle from image 2,
wearing the outfit from image 3, in the pose from image 4

Partial Transfer:

Arrange the scene using the layout shown in the collage input:
- Person from image 1 in the left position
- Object from image 2 in the center position
- Background element from image 3 filling the right side

Use a collage input for layout guidance:

1. Clear Image Roles:

Image 1: face/identity reference
Image 2: pose/body reference
Image 3: style/aesthetic reference
Image 4: environment/background reference

Specify what each reference provides:

2. Quality References:

• Use high-quality, clear reference images

• Ensure good lighting in references

• Avoid heavily processed or filtered images

3. Consistent Lighting:

...ensure the lighting direction matches across all elements,
with main light source from the upper left

When combining elements:

4. Resolution Awareness:

For [pro] API with 9MP total limit:

• At 1MP output: up to 8 reference images comfortably

• Calculate: input images + output = total MP

5. Explicit Relationships:

Create a scene combining:
- The woman from image 1 (keep exact face, expression, hair)
- Wearing the vintage dress from image 2 (exact pattern and cut)
- In the pose from image 3 (seated position, arm placement)
- Set in the library from image 4 (bookshelves, furniture)
- Using the warm lighting style from image 5 (golden hour quality)

Position her in the center of frame, medium shot, looking slightly
to the right with a thoughtful expression.

Don't assume - specify exactly how elements relate:

Elements Not Transferring:

• Be more specific about which element from which image

• Use explicit indexing ("from image 1")

• Reduce the number of references and complexity

Inconsistent Blending:

• Specify lighting consistency

• Describe how elements should interact

• Use style references to unify the composition

Identity Drift:

• Emphasize key identifying features

• Use phrases like "maintaining exact likeness"

• Provide multiple angles of the same subject if available

---

9. Positive Prompt Alternatives

Impact: MEDIUM

Strategies for achieving results without negative prompts, which FLUX does not support.

9.1 Negative Prompt Alternatives

Impact: MEDIUM

FLUX does not support negative prompts. This guide provides positive alternatives for common negative prompt patterns.

Negative prompts can actually make models focus MORE on unwanted elements. Instead, describe exactly what you DO want - this gives clearer direction and better results.

For any unwanted element:

1. Identify what you don't want

2. Ask: "What would be there instead?"

3. Describe the positive alternative

People/Crowds:

| Instead of | Use |

|-----------|-----|

| "no people" | "empty", "deserted", "solitary", "abandoned" |

| "no crowds" | "quiet", "peaceful", "secluded", "private" |

| "without background people" | "isolated subject", "clean background", "solo figure" |

Example:

Bad: A beach scene, no people
Good: A deserted beach at dawn, pristine untouched sand, solitary seagull

Skin/Appearance:

| Instead of | Use |

|-----------|-----|

| "no makeup" | "natural skin", "bare face", "fresh-faced" |

| "no blemishes" | "clear skin", "smooth complexion", "healthy glow" |

| "no wrinkles" | "youthful skin", "smooth features" |

Example:

Bad: Portrait of woman, no makeup, no blemishes
Good: Portrait of a woman with natural clear skin, fresh-faced with a healthy glow

Accessories:

| Instead of | Use |

|-----------|-----|

| "no glasses" | "visible eyes", "unobstructed gaze", "clear eye contact" |

| "no hat" | "bare head", "visible hair", "uncovered head" |

| "no jewelry" | "minimal accessories", "understated", "unadorned" |

Example:

Bad: Man portrait, no glasses, no hat
Good: Portrait of a man with clear direct gaze, wind-swept visible hair

Colors:

| Instead of | Use |

|-----------|-----|

| "no color" | "monochrome", "black and white", "grayscale" |

| "not colorful" | "muted tones", "subdued palette", "desaturated" |

| "no bright colors" | "neutral tones", "earth tones", "soft pastels" |

Example:

Bad: Landscape photo, no bright colors
Good: Landscape in muted earth tones, soft morning light, desaturated palette

Text/Watermarks:

| Instead of | Use |

|-----------|-----|

| "no text" | "clean surfaces", "unmarked", "text-free" |

| "no watermark" | "pristine image", "clean composition" |

| "no logos" | "unbranded", "plain", "logo-free surface" |

Example:

Bad: Product photo, no watermark, no text
Good: Clean product photography with pristine unmarked surfaces, minimal unbranded design

Style/Era:

| Instead of | Use |

|-----------|-----|

| "not modern" | "traditional", "classical", "vintage", "historical" |

| "no CGI look" | "photorealistic", "authentic", "natural", "organic" |

| "not cartoonish" | "realistic", "lifelike", "naturalistic" |

Example:

Bad: Building design, not modern, no futuristic elements
Good: Traditional Victorian architecture with classical ornate details and period-accurate features

Quality/Artifacts:

| Instead of | Use |

|-----------|-----|

| "no blur" | "sharp focus", "crisp details", "tack-sharp" |

| "no noise" | "clean image", "smooth gradients", "low ISO" |

| "no artifacts" | "pristine quality", "clean render", "flawless" |

Example:

Bad: Portrait, no blur, no noise
Good: Tack-sharp portrait with pristine image quality, smooth skin tones, crisp details

Objects:

| Instead of | Use |

|-----------|-----|

| "no cars" | "pedestrian area", "car-free zone", "walking street" |

| "no buildings" | "open landscape", "natural scenery", "wilderness" |

| "no furniture" | "empty room", "bare space", "minimalist interior" |

Example:

Bad: Street scene, no cars, no modern buildings
Good: Historic cobblestone walking street lined with traditional stone buildings from the 1800s

Weather/Environment:

| Instead of | Use |

|-----------|-----|

| "no rain" | "clear sky", "dry weather", "sunny day" |

| "no clouds" | "clear blue sky", "cloudless", "perfect visibility" |

| "not dark" | "well-lit", "bright", "daylight", "illuminated" |

Example:

Bad: Outdoor portrait, no rain, no clouds, not dark
Good: Outdoor portrait under clear blue sky on a bright sunny day, perfect natural lighting

Composition:

| Instead of | Use |

|-----------|-----|

| "no distractions" | "clean composition", "focused framing", "minimal elements" |

| "nothing in background" | "solid background", "isolated subject", "clean backdrop" |

| "no clutter" | "organized", "tidy", "minimal", "streamlined" |

Example:

Bad: Product shot, no distractions, nothing in background
Good: Product on clean white seamless backdrop, isolated subject, minimal focused composition

Original Negative-Heavy Prompt:

Portrait of a woman, no glasses, no makeup, no wrinkles, no blemishes,
no bright colors, no distracting background, no harsh lighting

Positive Rewrite:

Portrait of a youthful woman with clear natural skin and visible bright eyes,
fresh-faced with a healthy glow, wearing muted earth tones against a soft
blurred neutral background, gentle diffused lighting creating soft shadows

Original Negative-Heavy Prompt:

Landscape photo, no people, no buildings, no power lines, no modern elements,
no overcast sky, no dead trees

Positive Rewrite:

Pristine wilderness landscape with lush green living forest, clear blue sky,
untouched natural scenery stretching to the horizon, peaceful solitude with
only birdsong and wind, golden hour sunlight filtering through healthy foliage

| Unwanted | Positive Alternative |

|----------|---------------------|

| No people | Empty, solitary, deserted |

| No makeup | Natural, fresh-faced, bare |

| No text | Clean, unmarked, pristine |

| No blur | Sharp, crisp, tack-sharp |

| No modern | Traditional, vintage, classical |

| No dark | Bright, well-lit, luminous |

| No busy | Minimal, clean, focused |

| No artificial | Natural, organic, authentic |

---

References

1. https://docs.bfl.ai 2. https://bfl.ai

Sections

This file defines all sections, their ordering, impact levels, and descriptions. The section ID (in parentheses) is the filename prefix used to group rules.

---

1. Core Principles (core)

Impact: CRITICAL Description: Universal prompting principles that apply to all FLUX models. Master these before diving into specific techniques.

2. Model Selection (model, flux2, flux1)

Impact: HIGH Description: Choosing the right FLUX model for your use case. Covers both FLUX.2 (latest) and FLUX.1 (legacy) model families.

3. Text-to-Image Prompting (t2i)

Impact: HIGH Description: Crafting effective prompts for generating images from text descriptions.

4. Image-to-Image Editing (i2i)

Impact: HIGH Description: Techniques for editing and transforming existing images using FLUX.2 models.

5. JSON Structured Prompting (json)

Impact: MEDIUM-HIGH Description: Using structured JSON for complex multi-element scene composition.

6. Color Specification (hex)

Impact: MEDIUM Description: Precise color control using hex codes for brand-accurate generations.

7. Typography and Text (typography)

Impact: MEDIUM Description: Rendering text and typography within generated images.

8. Multi-Reference Editing (multi)

Impact: MEDIUM Description: Combining multiple reference images for style transfer and composition.

9. Positive Prompt Alternatives (negative)

Impact: MEDIUM Description: Strategies for achieving results without negative prompts, which FLUX does not support.

Core FLUX Prompting Principles

These principles apply to all FLUX models and form the foundation of effective prompting.

1. Positive Descriptions Only

FLUX does NOT support negative prompts. Always describe what you WANT, not what you don't want.

Wrong Approach

a portrait of a woman, no glasses, no hat, no makeup

Correct Approach

a portrait of a woman with natural skin, clear face, bare head, visible eyes

See negative-prompt-alternatives.md for comprehensive replacement strategies.

2. Prompt Structure Formula

Build prompts using this structure for consistent results:

[Subject] + [Action/Pose] + [Style/Medium] + [Context/Setting] + [Lighting] + [Technical Details]

Example

A young woman with flowing auburn hair (subject)
dancing gracefully in mid-leap (action)
in the style of classical oil painting (style)
in a moonlit garden with roses (context)
soft diffused moonlight with subtle rim lighting (lighting)
medium shot, shallow depth of field (technical)

3. Specificity Matters

More specific prompts yield dramatically better results.

Vague (Poor Results)

a cat sitting

Specific (Excellent Results)

A fluffy orange tabby cat with bright green eyes sitting regally on a vintage
velvet armchair, afternoon sunlight streaming through lace curtains, warm
golden hour lighting, shallow depth of field, shot on medium format film

4. Natural Language Works Best

Write prompts as descriptive prose rather than keyword lists.

Keyword Style (Less Effective)

woman, portrait, beautiful, blonde, studio, professional, 8k, detailed

Prose Style (More Effective)

A professional studio portrait of a beautiful blonde woman in her thirties,
captured with soft studio lighting that accentuates her features, rendered
in stunning detail with natural skin texture and subtle catchlights in her eyes

5. Lighting is Critical

Always specify lighting - it has the single greatest impact on image quality.

Natural Lighting

• Golden hour - warm, soft, directional
• Overcast - soft, diffused, even
• Harsh midday - high contrast, strong shadows
• Dappled forest light - specular, organic patterns

Studio Lighting

• Softbox - even, professional
• Rim light - edge definition, separation
• Butterfly lighting - beauty, glamour
• Rembrandt lighting - dramatic, classic portraits

Atmospheric Lighting

• Volumetric fog - depth, mystery
• God rays - dramatic, spiritual
• Neon glow - urban, cyberpunk
• Candlelight - warm, intimate

Mood-Based Lighting

• Dramatic shadows - tension, noir
• High key - bright, airy, clean
• Low key - moody, mysterious
• Chiaroscuro - strong contrast, painterly

6. Word Order Matters

FLUX prioritizes elements that appear earlier in the prompt. Front-load important elements.

Less Effective

A forest background with soft lighting where a knight in shining armor stands

More Effective

A knight in shining armor stands in a forest, soft dappled lighting filtering
through the canopy

7. Medium Prompt Length

Optimal prompt length is typically 30-80 words (FLUX can handle up to 512 tokens).

• Too short: Lacks direction, generic results
• Too long: Can become unfocused
• Sweet spot: Enough detail to guide, not so much it confuses

8. Iterative Refinement

Build prompts iteratively:

1. Start with core subject and action 2. Add style and medium 3. Specify lighting and atmosphere 4. Include technical details 5. Refine based on results

Change one element at a time to understand what affects your output.

FLUX.1 Model Family

Tip: FLUX.2 models are the latest generation and recommended for most use cases. FLUX.1 models are still available for specific needs.

Guide to FLUX.1 models and their specialized capabilities.

Model Overview

FLUX1.1 [pro]

Fast and reliable text-to-image generation.

Characteristics

• Strong prompt adherence
• Production-grade architecture
• Consistent, reliable results
• Scalable for high-volume
• Pricing: $0.04 per image

Prompting Style

Standard descriptive prompts with clear subject and style specification.

Example Prompt

A golden retriever puppy playing in autumn leaves, warm afternoon sunlight,
shallow depth of field with bokeh background, joyful expression, professional
pet photography style

FLUX.1 Kontext - Image Editing

Recommendation: FLUX.2 models with reference images provide improved editing results.

Context-aware image-to-image editing model for transformations and modifications.

Characteristics

• Understands image context
• Preserves unedited regions
• Style transfer capabilities
• Object modification
• Basic to complex transformations

Prompting Strategies

Basic Edits

Simple, direct instructions work best:

Change the color of the car to red

Make the sky sunset orange and pink

Controlled Edits

Be explicit about what to preserve:

Change the setting to nighttime while maintaining the exact same
composition and the painting's artistic style

Complex Transformations

For dramatic changes, be specific about preservation:

Transform the modern office into a Victorian library with the same
furniture arrangement. Keep the window positions and overall room
proportions identical.

Style Transfer

Reference specific artistic movements:

Transform this photograph into a Bauhaus art style with geometric
shapes and primary colors, maintaining the original composition
and subject positioning

Text Editing

Describe text placement and integration:

Add the text "OPEN" as a neon sign in the window, red glowing letters
with slight reflection on the glass, matching the nighttime atmosphere

Tips for Kontext

• Be explicit about what should NOT change
• Start with simpler edits, build complexity
• Specify style preservation when needed
• Use for incremental refinement

FLUX.1 Kontext Max

Advanced multi-reference editing for complex compositions.

Characteristics

• Handles up to 10 reference images
• Best editing consistency across references
• Complex scene composition
• Character consistency maintenance
• Rate limit: 6 concurrent requests

Multi-Reference Prompting

Natural Language References

Describe relationships between images naturally:

The person from image 1 is sitting in the cafe from image 2,
wearing the outfit from image 3, with the lighting style of image 4

Explicit Indexing

Reference images by number for precision:

Replace the top half of the person in image 1 with the clothing
from image 2, maintaining the pose and background

Tips for Kontext Max

• Plan your reference images carefully
• Use natural language for relationships
• Specify which elements come from which image
• Leverage for character consistency across scenes

FLUX.1 Fill - Inpainting

Specialized tool for object removal and area completion.

Characteristics

• Clean object removal
• Intelligent background completion
• Texture-aware filling
• Seamless blending

Use Cases

• Remove unwanted objects from photos
• Complete partial images
• Replace specific regions
• Clean up image artifacts

Prompting for Fill

Describe what should fill the masked area:

Fill with continuation of the brick wall texture and ivy

Complete with matching ocean waves and sandy beach

Tips for Fill

• Provide context about surrounding areas
• Specify texture and pattern continuation
• Describe lighting consistency
• Use for cleanup and removal tasks

FLUX.2 Model Family

Complete guide to FLUX.2 variants and their optimal prompting strategies.

Key Feature: All FLUX.2 models natively support both text-to-image generation AND image-to-image editing via reference images. There's no need to use legacy FLUX.1 Kontext models for editing tasks.

Model Overview

FLUX.2 [klein] - Fast Generation

Best for rapid prototyping, previews, and high-volume generation.

Characteristics

• 4B or 9B parameter versions available
• Sub-second generation times
• Optimized for speed over maximum detail
• No prompt upsampling - be descriptive yourself
• Supports up to 4 reference images

Prompting Style: Narrative Prose

Klein responds best to descriptive, narrative-style prompts with emphasis on lighting and atmosphere.

Example Prompt

A cozy coffee shop interior bathed in warm afternoon light, steam rising lazily
from ceramic cups, worn leather armchairs arranged around small wooden tables,
bookshelves lining exposed brick walls, the soft atmosphere of a quiet afternoon
with dust motes floating in sunbeams through tall windows

Tips for [klein]

• Write like a novelist describing a scene
• Front-load your subject (word order critical)
• Emphasize lighting descriptions heavily
• Keep prompts moderately detailed (40-70 words)

FLUX.2 [max] - Highest Quality

Premium model for final production assets and maximum detail.

Characteristics

• Highest detail and coherence
• Best editing consistency
• Vast world knowledge
• Includes grounding search (real-time web data)
• Strongest prompt following
• Supports up to 8 reference images (API), 10 (playground)

Prompting Style: Technical + Descriptive

[max] excels with detailed technical specifications combined with descriptive prose.

Example Prompt

Portrait of a weathered fisherman, age 70, deep wrinkles telling stories of
decades at sea, salt-and-pepper beard with streaks of white, wearing a navy
cable-knit sweater with visible wool texture. Shot on Hasselblad X2D with
90mm f/2.8 lens at f/4, golden hour natural light from the left creating
strong rim lighting, shallow depth of field with soft bokeh from harbor
lights behind, Kodak Portra 400 color science with natural grain

Tips for [max]

• Include camera and lens specifications for photorealism
• Specify film stock or digital sensor characteristics
• Use technical photography terms (aperture, focal length)
• Leverage grounding search for current events: "news photo of [recent event]"

FLUX.2 [pro] - Production Balanced

Optimal balance of quality and speed for production workflows.

Characteristics

• Good quality-to-speed ratio
• Reliable, consistent output
• Suitable for batch processing
• Supports prompt upsampling
• Supports up to 8 reference images

Prompting Style: Balanced Detail

Standard detailed prompts work well without excessive technical specification.

Example Prompt

A modern minimalist living room with floor-to-ceiling windows overlooking
a city skyline at dusk, clean white furniture with subtle textures, a single
statement plant in the corner, warm ambient lighting from hidden sources,
architectural photography style with clean lines and balanced composition

Tips for [pro]

• Balance specificity with generation speed
• Good for template-based prompt systems
• Enable prompt upsampling for enhanced results
• Consistent quality for production pipelines

FLUX.2 [flex] - Typography Specialist

Optimized for text rendering and typographic content.

Characteristics

• Superior text rendering quality
• Handles multiple text elements
• Adjustable steps (1-50) and guidance (1.5-10)
• Best for signage, posters, UI mockups
• Supports up to 8 reference images

Prompting Style: Typography-Focused

Always quote text and specify font characteristics explicitly.

Example Prompt

A modern minimalist poster design with the headline "DESIGN SUMMIT 2025"
in bold condensed sans-serif typography centered in the upper third,
subtitle "Innovation Meets Creativity" in lighter weight below,
date "MARCH 15-17" in small caps at the bottom, all text in white
on a gradient background transitioning from deep purple #4A0080 to
coral #FF6B6B, clean geometric accent lines, professional print quality

Tips for [flex]

• Always quote exact text: "Your Text Here"
• Specify font style: serif, sans-serif, script, display, monospace
• Describe text hierarchy: headline, subhead, body
• Include placement: centered, left-aligned, upper third
• Adjust steps (higher = better quality) and guidance (higher = stricter)

FLUX.2 [dev] - Local Development

For local development, testing, and non-commercial use.

Characteristics

• Open weights on Hugging Face
• Runs locally (~13GB VRAM recommended)
• Full customization available
• Free for non-commercial use
• Base variants available (undistilled) for fine-tuning

Prompting Style: Standard

Same prompting patterns as [pro] work well.

Tips for [dev]

• Use for development and testing before production
• Experiment with prompt variations
• Good for fine-tuning experiments
• Check license for commercial use restrictions

Image-to-Image Editing with FLUX.2

All FLUX.2 models support image editing via reference images. This replaces the need for legacy FLUX.1 Kontext models.

How It Works

1. Provide your source image(s) as reference images 2. Describe the desired changes in your prompt 3. The model preserves context while applying edits

Example: Style Transfer

Reference: [your source image]
Prompt: Transform this image into a watercolor painting style,
maintaining the exact composition and subject positioning

Example: Object Modification

Reference: [your source image]
Prompt: Change the car color to red while keeping everything else identical

Example: Character Consistency

Reference: [character reference image]
Prompt: The same person from the reference image walking through
a busy Tokyo street at night, neon lights reflecting on wet pavement

Model Selection for Editing

Hex Color Prompting

FLUX supports hex color codes (#RRGGBB) for precise color specification, essential for brand consistency and exact color matching.

Syntax

Include hex codes directly in your prompt with descriptive names:

A modern living room with walls painted in #2C3E50 (dark blue-gray),
accent pillows in #E74C3C (vibrant red), and a #F39C12 (warm amber)
throw blanket on a #ECF0F1 (off-white) sofa

Signal Keywords

Use these keywords to indicate color specification:

color #02eb3c
hex #edfa3c
in #FF5733
using color code #3498DB

Best Practices

1. Always Pair with Description

Never use hex codes alone - include the color name:

Good: #FF6B6B (coral pink)
Bad: #FF6B6B

2. Associate with Specific Objects

Clearly connect colors to their targets:

A product shot featuring a smartphone with a #1DA1F2 (Twitter blue) case,
resting on a #14171A (near black) matte surface

3. Limit Color Palette

3-5 colors typically work best. Too many can confuse the model:

Color palette for the scene: #2ECC71 (emerald green), #3498DB (sky blue),
#F1C40F (sunflower yellow), #FFFFFF (pure white)

Use Cases

Brand Colors

Corporate office reception with brand colors prominently featured:
walls in #0066CC (company blue), accent furniture in #FF6600 (company orange),
logo displayed in #FFFFFF (white) against the blue backdrop

Interior Design

Scandinavian minimalist bedroom with #F5F5F5 (warm white) walls,
#8B4513 (saddle brown) wooden headboard and nightstands,
#708090 (slate gray) linen bedding, and #DAA520 (goldenrod) accent lamp

Fashion

Editorial fashion photo: model wearing #000000 (black) cashmere turtleneck,
#FF4500 (orange-red) wide-leg wool pants, #C0C0C0 (silver) geometric earrings,
against a #F0F0F0 (light gray) studio backdrop

Product Design

Premium headphones product shot: #1C1C1E (space gray) aluminum body,
#F5F5F7 (silver) mesh ear cups, #FF9500 (iOS orange) accent ring around controls

Digital Art

Synthwave cityscape: #FF00FF (magenta) and #00FFFF (cyan) neon signs,
#1A1A2E (deep navy) night sky, #E94560 (hot pink) setting sun on horizon,
#16213E (dark blue) building silhouettes, rain-slicked streets reflecting lights

Data Visualization

Infographic showing market share: segments in #2ECC71 (green) for growth,
#E74C3C (red) for decline, #3498DB (blue) for stable, #95A5A6 (gray) for other,
clean #FFFFFF (white) background

Gradient Colors

Specify gradients with start and end colors:

Abstract background starting with color #02eb3c (bright green) and
finishing with color #edfa3c (lime yellow), smooth horizontal gradient

Sunset sky gradient from #FF6B6B (coral) at horizon through
#FFA07A (light salmon) to #87CEEB (sky blue) at top

Color Harmony Patterns

Complementary (Opposite on color wheel)

Scene using complementary colors: #3498DB (blue) dominant with
#E67E22 (orange) accents for visual pop

Analogous (Adjacent colors)

Harmonious palette using analogous colors: #9B59B6 (purple),
#8E44AD (deep purple), #3498DB (blue) - flowing naturally together

Triadic (Evenly spaced)

Vibrant triadic scheme: #E74C3C (red), #F1C40F (yellow),
#3498DB (blue) - balanced and dynamic

Monochromatic (Single hue variations)

Sophisticated monochromatic blue: #1A5276 (dark navy), #2980B9 (medium blue),
#85C1E9 (light blue), #D4E6F1 (pale blue) - elegant depth

Combining with JSON Structured Prompts

{
  "scene": {
    "setting": "modern tech startup office",
    "mood": "innovative, energetic"
  },
  "colors": {
    "primary": "#6C5CE7 (electric purple)",
    "secondary": "#00CEC9 (teal)",
    "accent": "#FD79A8 (pink)",
    "neutral": "#DFE6E9 (light gray)",
    "dark": "#2D3436 (charcoal)"
  },
  "application": {
    "walls": "neutral #DFE6E9",
    "furniture": "dark #2D3436",
    "accent_pieces": "primary #6C5CE7",
    "plants": "secondary #00CEC9 pots"
  }
}

Common Brand Color References

For reference only - always verify current brand guidelines:

# Social Media
Twitter/X Blue: #1DA1F2
Facebook Blue: #1877F2
Instagram Gradient: #833AB4 to #FD1D1D
LinkedIn Blue: #0A66C2

# Tech
Apple Gray: #1C1C1E
Google Blue: #4285F4
Microsoft Blue: #00A4EF
Amazon Orange: #FF9900

# Design
Figma Purple: #A259FF
Dribbble Pink: #EA4C89
Behance Blue: #1769FF

Troubleshooting

Color Not Accurate

• Add the color name alongside hex
• Specify the exact object the color applies to
• Use fewer total colors in the prompt

Color Bleeding

• Clearly delineate which objects get which colors
• Use spatial descriptions: "the LEFT chair in #color"

Muddy Colors

• Check hex code accuracy
• Specify lighting that won't shift colors
• Use "maintaining exact color #XXXXXX" for emphasis

Image-to-Image (I2I) Prompting

Guide to effective image-to-image editing with FLUX models.

Overview

All FLUX.2 models support image-to-image editing via reference images:

• FLUX.2 [klein]: Up to 4 reference images - fast editing
• FLUX.2 [pro]: Up to 8 reference images - balanced quality/speed
• FLUX.2 [max]: Up to 8-10 reference images - highest quality editing
• FLUX.2 [flex]: Up to 8 reference images - best for typography edits

Simply provide your source image as a reference and describe the desired changes. The model understands image context and can modify specific elements while preserving others.

Note: FLUX.2 models are recommended for image editing. They provide better results than the older FLUX.1 Kontext models.

Providing Images

Preferred: Use URLs directly - simpler and more convenient than base64.

When you have an image URL, pass it directly to input_image:

{
  "prompt": "Change the background to a beach sunset",
  "input_image": "https://example.com/photo.jpg"
}

The API fetches URLs automatically. Both URL and base64 work, but URLs are recommended when available.

Basic Edit Patterns

Simple Modifications

Direct, single-change instructions:

Change the car color to red

Make the sky a dramatic sunset

Add snow to the ground

Attribute Changes

Modifying specific characteristics:

Change her hair color to platinum blonde

Make the building taller

Age the person to appear 20 years older

Controlled Editing

Explicit Preservation

When you need to keep specific elements unchanged:

Change the background to a beach scene while keeping the subject's
pose, clothing, and expression exactly the same

Transform the daytime photo to nighttime, maintaining the exact
composition, colors of the subject's outfit, and lighting direction

Style Preservation

Preventing unwanted style shifts:

Change the season to autumn with falling leaves, but maintain
the photograph's realistic style and color grading

Add rain effects to the scene while preserving the painting's
impressionist brushwork and color palette

Transformation Types

Environmental Changes

Time of Day

Convert to golden hour lighting with warm tones and long shadows,
keeping all other elements identical

Transform to blue hour with city lights beginning to glow,
maintaining the exact composition

Weather

Add heavy rain with wet reflections on surfaces, dark overcast sky

Create a foggy atmosphere with reduced visibility, mysterious mood

Season

Transform to winter with snow covering surfaces, bare trees,
cold blue color cast

Change to spring with cherry blossoms, fresh green leaves,
soft warm lighting

Style Transfer

Artistic Movements

Transform into Art Nouveau style with flowing organic lines,
decorative patterns, and muted earth tones

Convert to Pop Art style with bold primary colors, halftone dots,
and high contrast graphic treatment

Artist References

Reimagine in the style of Monet with visible brushstrokes,
soft focus, and impressionist color harmony

Transform to match Edward Hopper's style with dramatic lighting,
urban isolation feeling, and muted palette

Medium Conversion

Convert this photograph to a detailed pencil sketch with
careful shading and visible line work

Transform into a watercolor painting with soft edges,
transparent washes, and paper texture visible

Subject Modifications

Clothing Changes

Change the outfit to a formal black suit with white shirt and red tie

Replace the casual clothes with traditional Japanese kimono in blue floral pattern

Expression Changes

Change the expression to a warm genuine smile

Make the expression more serious and contemplative

Age Modifications

Age the subject to appear as a wise elderly person with grey hair and wrinkles

Make the subject appear younger, around 25 years old

Object Editing

Addition

Add a vintage leather briefcase in the subject's left hand

Place a steaming cup of coffee on the table

Removal

Remove the background people, replace with empty street

Remove the text/logo from the shirt, replace with solid color

Replacement

Replace the modern car with a 1960s vintage Mustang in cherry red

Swap the coffee mug for an ornate teacup with floral pattern

Text Editing

Adding Text

Add a neon sign reading "OPEN 24 HOURS" in the window,
glowing red letters with blue outline

Include a wooden sign with hand-painted text "Welcome Home"
mounted above the door

Modifying Text

Change the store sign to read "BAKER'S DOZEN" in the same style

Update the poster text to "SUMMER SALE 2025" maintaining the design

Complex Multi-Step Edits

For dramatic transformations, consider breaking into steps:

Step-by-Step Approach

Instead of:

Transform this modern office into a Victorian library with completely
different furniture, add a fireplace, change the lighting to candlelit,
and age the photograph

Try sequential edits: 1. Change the furniture style to Victorian antique pieces 2. Add a stone fireplace on the right wall 3. Transform lighting to warm candlelit atmosphere 4. Apply vintage photograph aesthetic with sepia tones

Common Pitfalls

Avoid Vague Instructions

Bad: Make it look better
Good: Increase contrast, add warm color grading, sharpen details

Be Specific About Scope

Bad: Change the background
Good: Replace the office background with a tropical beach at sunset,
      maintaining the subject's exact position and lighting direction

Explicit Style Preservation

Bad: Make it nighttime
Good: Transform to nighttime while maintaining the photorealistic style,
      add appropriate artificial lighting sources

Best Practices

1. Start Simple - Begin with single-element changes 2. Be Explicit - State what should change AND what should stay 3. Reference Context - Mention existing elements when relevant 4. Iterate - Refine through multiple small edits rather than one large one 5. Preserve Deliberately - Always specify style/composition preservation needs

JSON Structured Prompting

For complex scenes with multiple elements, spatial relationships, or production automation, use JSON-structured prompts.

When to Use

• Multiple characters with distinct attributes
• Precise spatial positioning
• Complex scene composition
• Reproducible, template-based prompts
• Programmatic prompt generation
• Production workflows with variable substitution

Basic Structure

{
  "scene": {
    "setting": "description of environment",
    "time": "time of day/period",
    "mood": "atmospheric quality"
  },
  "subjects": [
    {
      "type": "person/object/animal",
      "description": "detailed description",
      "position": "location in frame",
      "action": "what they're doing"
    }
  ],
  "style": {
    "medium": "photography/painting/illustration",
    "technique": "specific style details",
    "reference": "artist or style reference"
  },
  "technical": {
    "camera": "camera and lens",
    "lighting": "lighting setup",
    "composition": "framing details"
  },
  "colors": ["#hex1", "#hex2"]
}

Single Subject Example

{
  "scene": {
    "setting": "cozy home office with bookshelves",
    "time": "late afternoon",
    "mood": "focused, peaceful"
  },
  "subjects": [
    {
      "type": "person",
      "description": "woman in her 30s, dark curly hair in loose bun, wearing casual cream sweater",
      "position": "seated at desk, center frame",
      "action": "typing on laptop, slight smile of concentration"
    }
  ],
  "style": {
    "medium": "photography",
    "technique": "lifestyle editorial",
    "reference": "kinfolk magazine aesthetic"
  },
  "technical": {
    "camera": "Sony A7III with 50mm f/1.8",
    "lighting": "soft natural window light from left",
    "composition": "medium shot, rule of thirds"
  }
}

Multi-Character Scene

{
  "scene": {
    "setting": "Victorian-era drawing room with ornate wallpaper and antique furniture",
    "time": "evening, candlelit",
    "mood": "tense, mysterious"
  },
  "subjects": [
    {
      "id": "detective",
      "type": "person",
      "description": "tall man in his 50s, sharp features, grey at temples, wearing brown tweed suit",
      "position": "standing center-left, facing right",
      "action": "examining a letter with magnifying glass, intense focus"
    },
    {
      "id": "lady",
      "type": "person",
      "description": "elegant woman in her 40s, auburn hair in Victorian updo, emerald green evening dress",
      "position": "seated on chaise lounge, right side",
      "action": "watching the detective with concealed anxiety, hands clasped"
    },
    {
      "id": "butler",
      "type": "person",
      "description": "elderly man in formal butler attire, stoic expression",
      "position": "background, near doorway",
      "action": "standing at attention, observing"
    }
  ],
  "style": {
    "medium": "oil painting",
    "technique": "classical realism with dramatic lighting",
    "reference": "Victorian narrative painting, John Singer Sargent"
  },
  "technical": {
    "lighting": "warm candlelight as key, cool moonlight through window as fill",
    "composition": "triangular arrangement of figures, detective at apex"
  }
}

Product Scene with Colors

{
  "scene": {
    "setting": "minimalist product photography studio",
    "mood": "clean, premium, aspirational"
  },
  "subjects": [
    {
      "type": "product",
      "description": "sleek wireless earbuds in charging case",
      "position": "center, slightly angled",
      "details": "matte finish, subtle branding"
    }
  ],
  "style": {
    "medium": "commercial photography",
    "technique": "high-end product shot",
    "reference": "Apple product photography"
  },
  "technical": {
    "camera": "Phase One with 120mm macro",
    "lighting": "large softbox overhead, subtle fill from below",
    "composition": "centered, hero product shot"
  },
  "colors": {
    "product": "#1A1A2E",
    "accent": "#E94560",
    "background": "#FFFFFF"
  }
}

Converting JSON to Natural Language

Flatten your JSON into flowing prose for the actual prompt:

From JSON

{
  "subjects": [
    {
      "type": "person",
      "description": "elderly craftsman with weathered hands",
      "position": "seated at workbench",
      "action": "carefully carving wood"
    }
  ],
  "scene": { "setting": "traditional workshop", "time": "morning" },
  "technical": { "lighting": "natural window light from right" }
}

To Prompt

An elderly craftsman with weathered hands seated at his workbench in a
traditional workshop, carefully carving wood with focused precision.
Morning natural light streams through the window from the right,
illuminating the wood shavings and tools scattered across the worn surface.

Template Variables

Use JSON structure for template-based generation:

{
  "template": "product_hero",
  "variables": {
    "product_name": "{{PRODUCT_NAME}}",
    "product_color": "{{PRODUCT_COLOR}}",
    "brand_primary": "{{BRAND_HEX_1}}",
    "brand_secondary": "{{BRAND_HEX_2}}",
    "background_style": "{{BG_STYLE}}"
  },
  "prompt_template": "Professional product photography of {{PRODUCT_NAME}} in {{PRODUCT_COLOR}}, brand colors {{BRAND_HEX_1}} and {{BRAND_HEX_2}} accents, {{BG_STYLE}} background, commercial quality"
}

Spatial Relationships

Define explicit spatial relationships:

{
  "composition": {
    "layout": "triangular",
    "focal_point": "center-left intersection",
    "depth_layers": [
      {
        "layer": "foreground",
        "elements": ["flowers in vase"],
        "focus": "soft blur"
      },
      {
        "layer": "midground",
        "elements": ["main subject"],
        "focus": "sharp"
      },
      {
        "layer": "background",
        "elements": ["window", "garden view"],
        "focus": "soft blur"
      }
    ]
  }
}

Best Practices

1. Use IDs for References - Give subjects IDs when they interact 2. Separate Concerns - Keep scene, subjects, style, and technical distinct 3. Be Consistent - Use the same terminology throughout 4. Include All Details - Don't assume, specify everything 5. Flatten for Execution - Convert to natural language before sending to model 6. Version Templates - Track template versions for reproducibility

FLUX Model Selection Guide

Decision guide for selecting the optimal FLUX model based on your requirements.

Quick Decision Matrix

Decision Tree

What's your primary need?

├─ Generate images (text-to-image OR image editing)
│   ├─ Need FASTEST possible generation?
│   │   └─ FLUX.2 [klein] (supports up to 4 reference images)
│   │
│   ├─ Need HIGHEST quality output?
│   │   └─ FLUX.2 [max] (supports up to 8-10 reference images)
│   │
│   ├─ Need TEXT/TYPOGRAPHY in image?
│   │   └─ FLUX.2 [flex] (supports up to 8 reference images)
│   │
│   ├─ Need BALANCED speed/quality?
│   │   └─ FLUX.2 [pro] (supports up to 8 reference images)
│   │
│   └─ Need LOCAL/FREE generation?
│       └─ FLUX.2 [dev]
│
├─ Need REAL-TIME web information?
│   └─ FLUX.2 [max] (grounding search)
│
└─ FLUX.1 family (only use when explicitly asked by user)
    ├─ FLUX.1 Kontext - context-aware editing
    └─ FLUX.1 Fill - inpainting/object removal

Note: All FLUX.2 models natively support image-to-image editing via reference images. Simply provide your source image(s) as references and describe the desired changes.

Detailed Model Comparisons

By Speed

By Quality

By Cost

Credit pricing: 1 credit = $0.01 USD. FLUX.2 uses megapixel-based pricing.

FLUX.2 Models

Pricing formula: (firstMP + (outputMP-1) * mpPrice) + (inputMP * mpPrice) in cents

FLUX.1 Models

Use bfl.ai/pricing calculator for exact costs at different resolutions.

Use Case Recommendations

Creative Exploration / Ideation

Recommended: FLUX.2 [klein]

• Fast iterations
• Quick concept testing
• Mood board generation
• Exploring prompt variations

Production Marketing Assets

Recommended: FLUX.2 [pro]

• Consistent quality
• Reasonable speed
• Cost-effective at scale
• Reliable for automation

Hero Images / Premium Content

Recommended: FLUX.2 [max]

• Maximum detail
• Best coherence
• Supports grounding search
• Worth the premium for key visuals

Typography / Signage / Posters

Recommended: FLUX.2 [flex]

• Superior text rendering
• Adjustable quality settings
• Best for readable text
• UI mockups and infographics

Character Consistency

Recommended: FLUX.2 [max] or [pro]

• Multi-reference support (up to 8-10 images)
• Best editing consistency
• Maintains identity across scenes
• Superior quality over FLUX.1 Kontext

Photo Editing / Retouching

Recommended: FLUX.2 [klein], [pro], or [max]

• Native image-to-image support via references
• Style transfer
• Object modification
• Attribute changes
• Better results than FLUX.1 Kontext

Real-Time Information

Recommended: FLUX.2 [max]

• Grounding search feature
• Current events
• Recent news visualization
• Weather/location data

Local Development / Testing

Recommended: FLUX.2 [dev]

• No API costs
• Full control
• Fine-tuning experiments
• Offline capability

Editorial with Typography

1. FLUX.2 [max] - Generate base image (highest quality)
2. FLUX.2 [flex] - Add text overlay pass

Character-Consistent Series

1. FLUX.2 [max] - Create character reference
2. FLUX.2 [max]/[pro] - Generate consistent variations using reference images
3. FLUX.2 [klein] - Quick iteration on variations if needed

E-commerce Product Pipeline

1. FLUX.2 [pro] - Bulk product generations
2. FLUX.2 [pro]/[klein] - Product variations (colors, angles) using references
3. FLUX.2 [flex] - Add promotional text/pricing

Constraint-Based Selection

Limited Budget

• High volume: FLUX.2 [klein] 4B
• Quality needed: FLUX.2 [pro] (best value)

Tight Deadline

• Any task: FLUX.2 [klein]
• Quality matters: FLUX.2 [pro]

Maximum Quality Required

• Always: FLUX.2 [max]

Text Must Be Readable

• Always: FLUX.2 [flex]

Editing Existing Images

• Fast edits: FLUX.2 [klein] with reference images
• Quality edits: FLUX.2 [max] or [pro] with reference images
• Alternative: FLUX.1 Kontext (FLUX.2 preferred)

Rate Limit Sensitivity

• Prefer: FLUX.2 models (24 concurrent limit)
• Avoid: FLUX.1 Kontext Max (6 concurrent limit)

Summary Cheat Sheet

Speed?      → FLUX.2 [klein]
Quality?    → FLUX.2 [max]
Balance?    → FLUX.2 [pro]
Text?       → FLUX.2 [flex]
Edit?       → FLUX.2 [klein/pro/max] with reference images
Free?       → FLUX.2 [dev]

Key insight: All FLUX.2 models support image editing natively via reference images. FLUX.2 is recommended over FLUX.1 Kontext for editing tasks.

Multi-Reference Image Editing

Guide to using multiple reference images for character consistency, style transfer, and complex compositions.

Overview

FLUX.2 models support multiple reference images for advanced editing:

• FLUX.2 [klein]: Up to 4 reference images - fast editing
• FLUX.2 [pro]: Up to 8 via API - balanced quality/speed
• FLUX.2 [max]: Up to 8 via API, 10 in playground - highest quality
• FLUX.2 [flex]: Up to 8 via API - best for typography

Note: FLUX.2 models are recommended over FLUX.1 Kontext Max for better results.

Providing Images

Preferred: Use URLs directly - simpler and more convenient than base64.

Pass image URLs directly to input_image, input_image_2, etc.:

{
  "prompt": "Person from image 1 wearing outfit from image 2",
  "input_image": "https://example.com/person.jpg",
  "input_image_2": "https://example.com/outfit.jpg"
}

The API fetches URLs automatically. Both URL and base64 work, but URLs are recommended when available.

Reference Methods

Natural Language Description

Describe relationships between images naturally:

The person from image 1 is sitting at the cafe table from image 2,
wearing the outfit from image 3, with the warm lighting style of image 4

Explicit Indexing

Reference images by number for precision:

Replace the background of image 1 with the landscape from image 2,
maintaining the subject's exact position and lighting

Combine the face from image 1 with the hairstyle from image 2
on the body pose from image 3

Use Cases

Character Consistency

Maintain the same character across multiple scenes:

Input: Reference image of character
Prompt: The character from image 1 walking through a busy Tokyo street
at night, neon lights reflecting on wet pavement

For sequential consistency:

The same person from image 1, now seated at a desk in a modern office,
same clothing and hairstyle, different environment

Style Transfer

Apply the style of one image to another:

Transform image 1 into the artistic style of image 2,
maintaining the original composition and subject

Apply the color grading and mood from image 2 to the scene in image 1

Pose Guidance

Use a reference for body positioning:

The person from image 1 in the exact pose shown in image 2,
placed in the environment from image 3

Object Composition

Combine elements from multiple images:

Place the product from image 1 on the table setting from image 2,
using the lighting style from image 3

Background Replacement

Keep the subject from image 1 exactly as shown, replace the background
with the beach scene from image 2, match the lighting naturally

Multi-Character Scenes

Two Characters

Image 1 (person A) and image 2 (person B) having a conversation
at a coffee shop table, person A on the left gesturing, person B
on the right listening intently

Group Composition

The three people from images 1, 2, and 3 standing together for a
group photo, arranged left to right in that order, friendly poses,
outdoor park setting

Attribute Mixing

Selective Attribute Transfer

The face and expression from image 1, the hairstyle from image 2,
wearing the outfit from image 3, in the pose from image 4

Partial Transfer

Apply only the color palette from image 2 to image 1,
keeping all other aspects (style, composition, lighting) unchanged

Collage Method

Use a collage input for layout guidance:

Arrange the scene using the layout shown in the collage input:
- Person from image 1 in the left position
- Object from image 2 in the center position
- Background element from image 3 filling the right side

Best Practices

1. Clear Image Roles

Specify what each reference provides:

Image 1: face/identity reference
Image 2: pose/body reference
Image 3: style/aesthetic reference
Image 4: environment/background reference

2. Quality References

• Use high-quality, clear reference images
• Ensure good lighting in references
• Avoid heavily processed or filtered images

3. Consistent Lighting

When combining elements:

...ensure the lighting direction matches across all elements,
with main light source from the upper left

4. Resolution Awareness

For [pro] API with 9MP total limit:

• At 1MP output: up to 8 reference images comfortably
• Calculate: input images + output = total MP

5. Explicit Relationships

Don't assume - specify exactly how elements relate:

Vague: The person and the background together
Better: The person from image 1 standing in the foreground,
the beach from image 2 visible behind them at a distance

Complex Composition Example

Create a scene combining:
- The woman from image 1 (keep exact face, expression, hair)
- Wearing the vintage dress from image 2 (exact pattern and cut)
- In the pose from image 3 (seated position, arm placement)
- Set in the library from image 4 (bookshelves, furniture)
- Using the warm lighting style from image 5 (golden hour quality)

Position her in the center of frame, medium shot, looking slightly
to the right with a thoughtful expression.

Troubleshooting

Elements Not Transferring

• Be more specific about which element from which image
• Use explicit indexing ("from image 1")
• Reduce the number of references and complexity

Inconsistent Blending

• Specify lighting consistency
• Describe how elements should interact
• Use style references to unify the composition

Identity Drift

• Emphasize key identifying features
• Use phrases like "maintaining exact likeness"
• Provide multiple angles of the same subject if available

Negative Prompt Alternatives

FLUX does not support negative prompts. This guide provides positive alternatives for common negative prompt patterns.

Why No Negative Prompts?

Negative prompts can actually make models focus MORE on unwanted elements. Instead, describe exactly what you DO want - this gives clearer direction and better results.

Replacement Strategy

For any unwanted element: 1. Identify what you don't want 2. Ask: "What would be there instead?" 3. Describe the positive alternative

Common Replacements

People/Crowds

Example:

Bad: A beach scene, no people
Good: A deserted beach at dawn, pristine untouched sand, solitary seagull

Skin/Appearance

Example:

Bad: Portrait of woman, no makeup, no blemishes
Good: Portrait of a woman with natural clear skin, fresh-faced with a healthy glow

Accessories

Example:

Bad: Man portrait, no glasses, no hat
Good: Portrait of a man with clear direct gaze, wind-swept visible hair

Colors

Example:

Bad: Landscape photo, no bright colors
Good: Landscape in muted earth tones, soft morning light, desaturated palette

Text/Watermarks

Example:

Bad: Product photo, no watermark, no text
Good: Clean product photography with pristine unmarked surfaces, minimal unbranded design

Style/Era

Example:

Bad: Building design, not modern, no futuristic elements
Good: Traditional Victorian architecture with classical ornate details and period-accurate features

Quality/Artifacts

Example:

Bad: Portrait, no blur, no noise
Good: Tack-sharp portrait with pristine image quality, smooth skin tones, crisp details

Objects

Example:

Bad: Street scene, no cars, no modern buildings
Good: Historic cobblestone walking street lined with traditional stone buildings from the 1800s

Weather/Environment

Example:

Bad: Outdoor portrait, no rain, no clouds, not dark
Good: Outdoor portrait under clear blue sky on a bright sunny day, perfect natural lighting

Composition

Example:

Bad: Product shot, no distractions, nothing in background
Good: Product on clean white seamless backdrop, isolated subject, minimal focused composition

Complex Replacement Examples

Original Negative-Heavy Prompt

Portrait of a woman, no glasses, no makeup, no wrinkles, no blemishes,
no bright colors, no distracting background, no harsh lighting

Positive Rewrite

Portrait of a youthful woman with clear natural skin and visible bright eyes,
fresh-faced with a healthy glow, wearing muted earth tones against a soft
blurred neutral background, gentle diffused lighting creating soft shadows

Original Negative-Heavy Prompt

Landscape photo, no people, no buildings, no power lines, no modern elements,
no overcast sky, no dead trees

Positive Rewrite

Pristine wilderness landscape with lush green living forest, clear blue sky,
untouched natural scenery stretching to the horizon, peaceful solitude with
only birdsong and wind, golden hour sunlight filtering through healthy foliage

Quick Reference Card

Text-to-Image (T2I) Prompting

Comprehensive guide to crafting effective text-to-image prompts for FLUX models.

Prompt Structure Framework

Basic Formula

[Subject] + [Action] + [Style] + [Context] + [Lighting] + [Technical]

Expanded Framework

[Main Subject] - who/what is the focus
[Attributes] - characteristics, details, clothing
[Action/Pose] - what they're doing
[Environment] - where, setting, background
[Style/Medium] - artistic approach
[Lighting] - light source, quality, mood
[Composition] - framing, camera angle
[Technical] - camera, lens, film stock

Subject Types

People/Portraits

A distinguished professor in his 60s with silver hair and round spectacles,
wearing a tweed jacket with leather elbow patches, deep-set thoughtful eyes,
slight smile suggesting hidden wisdom

Animals

A majestic snow leopard with piercing blue-grey eyes, thick spotted fur
dusted with snowflakes, powerful muscular build, alert posture on a
rocky outcrop

Objects/Products

A vintage Leica M3 camera with worn brass edges showing decades of use,
black leather covering with patina, sitting on weathered wooden table

Landscapes

A dramatic fjord at dawn, steep granite cliffs rising from mirror-still
water, wisps of morning mist, distant snow-capped peaks catching first
golden light

Architecture

A brutalist concrete apartment building in late afternoon light, geometric
shadows creating abstract patterns, warm sunlight contrasting with cool
grey concrete

Style Categories

Photorealistic

Modern Digital

shot on Sony A7IV, clean and sharp, high dynamic range, professional color grading

Film Photography

shot on Kodak Portra 400, natural film grain, organic colors, slight warmth

Vintage Digital (2000s)

early digital camera aesthetic, slight noise, flash photography, candid feel

80s Film

80s film photography, film grain, warm color cast, soft focus, nostalgic

Artistic Styles

Oil Painting

classical oil painting style, visible brushstrokes, rich colors, dramatic lighting

Watercolor

delicate watercolor painting, soft edges, transparent washes, paper texture visible

Digital Art

polished digital illustration, clean lines, vibrant colors, professional concept art

Anime/Manga

anime style, large expressive eyes, clean linework, cel shading, vibrant palette

Lighting Patterns

Portrait Lighting

Rembrandt lighting - 45 degree key light creating triangle shadow on cheek
Butterfly lighting - overhead key creating shadow under nose
Split lighting - 90 degree side light, half face in shadow
Loop lighting - slight angle creating small nose shadow

Natural Lighting

Golden hour - warm, soft, directional light 1 hour before sunset
Blue hour - cool, ambient light just after sunset
Overcast - soft, even, diffused lighting
Harsh midday - strong contrast, defined shadows

Atmospheric

Volumetric light - visible light rays through fog/dust
Rim lighting - backlight creating edge glow
Practical lighting - visible light sources in scene
Neon glow - colorful artificial urban lighting

Camera and Lens Simulation

Camera Bodies

Shot on Hasselblad X2D - medium format, exceptional detail
Shot on Canon 5D Mark IV - professional DSLR quality
Shot on Leica M10 - rangefinder character, smooth tonality
Shot on iPhone 15 Pro - computational photography look

Lens Characteristics

85mm f/1.4 - classic portrait, creamy bokeh
24mm f/2.8 - wide angle, environmental
50mm f/1.2 - natural perspective, shallow DOF
135mm f/2 - compressed perspective, smooth background
Macro lens - extreme close-up detail
Tilt-shift lens - miniature effect or architectural correction

Technical Settings

f/1.4 - extremely shallow depth of field
f/2.8 - moderate background blur
f/8 - sharp throughout, landscape
f/16 - maximum sharpness, long exposure
ISO 100 - clean, no noise
ISO 3200 - visible grain, low light

Composition Techniques

Framing

extreme close-up - filling frame with detail
close-up - head and shoulders
medium shot - waist up
full shot - entire body
wide shot - subject in environment
establishing shot - location focus

Angles

eye level - natural, relatable
low angle - powerful, imposing
high angle - diminished, overview
Dutch angle - tension, unease
bird's eye - pattern, layout
worm's eye - dramatic upward view

Composition Rules

rule of thirds - subject at intersection points
centered composition - symmetry, stability
leading lines - guiding eye to subject
frame within frame - natural framing elements
negative space - minimalist, breathing room

Complete Example Prompts

Editorial Portrait

A fashion editorial portrait of a young woman with striking features and
high cheekbones, wearing an avant-garde geometric collar in silver, dramatic
side lighting creating strong shadows, shot on Hasselblad with 100mm lens
at f/2.8, studio background with subtle gradient, high fashion magazine style

Product Photography

A premium wireless headphone product shot, matte black finish with rose gold
accents, floating at slight angle against pure white background, soft even
lighting eliminating harsh shadows, reflection visible on glossy surface below,
commercial catalog style, ultra sharp focus throughout

Landscape

A misty morning in ancient redwood forest, towering trees disappearing into
fog above, ferns covering forest floor in layers of green, single shaft of
golden sunlight breaking through canopy, shot on large format camera, rich
detail in bark textures, Ansel Adams inspired black and white with deep tones

Architectural

Modern minimalist beach house at golden hour, floor-to-ceiling glass walls
reflecting sunset colors, clean white concrete and natural wood, infinity
pool merging with ocean horizon, architectural photography style, wide angle
showing full structure, warm evening light

Typography and Text Prompting

Guide to rendering text in FLUX images. Use FLUX.2 [flex] for best typography results.

Basic Syntax

Always quote the exact text you want rendered:

A coffee shop chalkboard sign displaying "TODAY'S SPECIAL" in decorative script

Core Rules

1. Use Quotation Marks

Correct: A poster with "HELLO WORLD" in bold letters
Wrong: A poster with HELLO WORLD in bold letters

2. Specify Font Style

"ADVENTURE" in bold sans-serif typography
"Welcome" in elegant cursive script
"CHAPTER ONE" in classic serif typeface
"CODE" in monospace terminal font
"SALE!" in decorative display lettering

3. Describe Size Hierarchy

Large headline "BREAKING NEWS" above smaller subtext "Details inside"

4. Indicate Placement

"OPEN" sign centered in storefront window
"EXIT" text positioned above doorway
"Page 1" in bottom right corner

5. Front-Load Text

Place text descriptions early in the prompt for better accuracy:

Good: A sign reading "FRESH BREAD" in a bakery window...
Less Good: A bakery window with a sign that says "FRESH BREAD"...

Font Style Categories

Sans-Serif (Modern/Clean)

"MINIMAL" in clean geometric sans-serif, Swiss modernist style
"TECH SUMMIT" in bold condensed grotesque typeface
"future" in thin uppercase sans-serif, contemporary design

Serif (Classic/Elegant)

"The New Yorker" in traditional serif typeface, editorial masthead
"LUXURY" in high-contrast Didone serif with thin/thick strokes
"Wisdom" in old-style serif with subtle bracketed serifs

Script/Cursive (Decorative)

"With Love" in flowing calligraphic script with flourishes
"Signature" in connected brush script, casual elegance
"Romance" in formal copperplate script, wedding invitation style

Display/Decorative

"ROCK CONCERT" in distressed vintage concert poster lettering
"CIRCUS" in ornate Victorian display type with decorative elements
"RETRO" in 1970s rounded bubble letters

Handwritten

"Note to self" in casual handwritten style, slightly imperfect
"Thanks!" in quick marker pen handwriting
"ideas" in sketchy pencil handwriting

Monospace

"CODE_COMPLETE" in terminal monospace, developer aesthetic
"SYSTEM" in typewriter monospace, vintage tech
"DEBUG" in LCD-style digital monospace

Text Effects

Neon Signs

Glowing neon sign spelling "OPEN 24/7" in pink neon tubes with
blue outline, slight glow and reflection, night scene

Metallic/3D

"GOLD" in three-dimensional metallic gold letters with realistic
reflections and subtle shadows, luxury aesthetic

Embossed/Debossed

"PREMIUM" embossed into leather surface, subtle shadows showing
the raised letterforms

Outlined

"MODERN" in outline-only letters, no fill, thin white stroke
on dark background

Gradient Text

"SUMMER" with gradient fill from #FF6B6B (coral) at top to
#4ECDC4 (teal) at bottom

Multi-Text Compositions

Poster Design

Event poster with "SUMMER FEST 2025" as large headline in bold
condensed sans-serif at top, "JULY 15-17" as medium subheading
in regular weight, "Central Park, NYC" as small body text at
bottom, all in white text on #FF6B35 (sunset orange) background

Book Cover

Book cover design: "THE GREAT GATSBY" in elegant art deco gold
lettering centered in upper third, author name "F. SCOTT FITZGERALD"
in smaller gold caps below, #1A1A2E (midnight blue) background
with geometric gold accents

Magazine Cover

Fashion magazine cover with "VOGUE" in classic serif masthead at top,
cover line "SPRING COLLECTION" in bold sans-serif, "The New Rules of Style"
in lighter weight italic, all in white against dramatic portrait

Signage

Vintage diner sign: "MEL'S DINER" in red neon script lettering,
"OPEN" below in separate green neon block letters, chrome border,
1950s Americana aesthetic

Business Card

Minimalist business card with "JOHN SMITH" in medium weight sans-serif,
"Creative Director" in lighter weight below, contact details in small
type at bottom, #2C3E50 (dark blue) text on white background

Text Placement Strategies

Centered Composition

Centered text layout: "WELCOME" in large caps at center,
perfectly balanced with equal margins

Left-Aligned

Left-aligned text block: "Company Name" as header,
"Tagline goes here" below, flush left alignment

Text on Path

"GOING IN CIRCLES" text following a circular path around
the center of the design

Text Overlay

"ADVENTURE AWAITS" in bold white text overlaid on landscape
photograph, positioned in lower third with slight shadow for readability

Technical Considerations for [flex]

Steps Parameter

• Higher steps (30-50) = better text quality
• Lower steps (10-20) = faster, lower quality

Guidance Parameter

• Higher guidance (6-10) = stricter prompt following
• Lower guidance (1.5-4) = more creative interpretation

Recommended Settings

For clean typography: steps=50, guidance=7
For artistic text: steps=30, guidance=4

Troubleshooting

Misspelled Words

• Keep text short (1-4 words work best)
• Use common words when possible
• Repeat the exact text in the prompt

Illegible Text

• Specify larger text size
• Use simpler fonts (sans-serif)
• Ensure high contrast with background
• Use [flex] model

Wrong Font Style

Be more specific:

Instead of: "text in a nice font"
Use: "text in bold geometric sans-serif similar to Futura"

Text Not Appearing

• Front-load text description in prompt
• Put text in quotes
• Specify exact placement
• Reduce other prompt complexity