Text-to-Image Generation

Transform your creative ideas into stunning images using cutting-edge AI image generation models.

Overview

This skill combines the power of Stability AI's Stable Diffusion and Black Forest Labs' FLUX models to generate high-quality images from text descriptions. Choose the right model based on your needs - photorealism, artistic styles, or text rendering.

Available Tools

Stability AI

•{{txt2img}}: Generate images using Stable Diffusion Ultra - great for artistic and creative outputs

FLUX Models (Black Forest Labs)

•{{flux_pro}}: FLUX Pro 1.1 - highest quality photorealistic generation with excellent text rendering
•{{flux_flex}}: Flexible generation with adjustable guidance scale for creative control

When to Use Each Tool

Tool	Best For	Quality	Speed
{{flux_pro}}	Photorealistic images, product photos, text in images	Highest	~30s
{{flux_flex}}	Creative freedom, artistic styles, experimentation	High	~20s
{{txt2img}}	General purpose, diverse styles, quick iterations	High	~10s

Usage Examples

Example 1: Photorealistic Product Photography

Use {{flux_pro}} for the best photorealistic results:

json

{
  "prompt": "Professional product photo of sleek wireless headphones on a marble surface, studio lighting, 4K, highly detailed",
  "width": 1024,
  "height": 1024,
  "prompt_upsampling": true,
  "output_format": "png"
}

Example 2: Artistic Illustration

Use {{txt2img}} for artistic and creative styles:

json

{
  "prompt": "Watercolor painting of a cozy coffee shop in autumn, warm colors, artistic style",
  "negative_prompt": "blurry, low quality, distorted",
  "output_format": "png"
}

Example 3: Text in Images

Use {{flux_pro}} for accurate text rendering:

json

{
  "prompt": "Motivational poster with the text 'Dream Big' in elegant typography, inspirational design",
  "width": 768,
  "height": 1024,
  "prompt_upsampling": true
}

Example 4: Creative Experimentation

Use {{flux_flex}} with adjustable guidance for creative freedom:

json

{
  "prompt": "Surreal landscape with floating islands, dreamlike atmosphere, fantasy art",
  "guidance": 2.5,
  "width": 1440,
  "height": 768
}

Prompt Engineering Tips

Be Descriptive

More details lead to better results:

•❌ "a dog"
•✅ "a golden retriever puppy playing in autumn leaves, soft afternoon sunlight, shallow depth of field"

Specify Style

Include artistic style keywords:

•"cinematic lighting", "studio photography", "oil painting", "watercolor", "digital art"
•"4K", "highly detailed", "professional", "award-winning"

Use Negative Prompts (Stability AI)

With {{txt2img}}, exclude unwanted elements:

•"blurry, low quality, distorted, disfigured"

Aspect Ratios

Choose dimensions that match your content:

•1024x1024: Square format, balanced compositions
•1024x768: Landscape, scenic views
•768x1024: Portrait, vertical subjects
•1440x1024: Wide cinematic shots

Parameters

Common Parameters

Parameter	Description	Default
prompt	Text description of the image	Required
width	Image width (256-1440)	1024
height	Image height (256-1440)	1024
seed	Random seed for reproducibility	Random
output_format	png, jpeg, or webp	png

FLUX-Specific ({{flux_pro}}, {{flux_flex}})

Parameter	Description	Default
prompt_upsampling	Auto-enhance prompts	false
safety_tolerance	Content filter (0-6)	2
guidance	Guidance scale ({{flux_flex}} only)	3.5

Stability-Specific ({{txt2img}})

Parameter	Description	Default
negative_prompt	What to avoid	None

Best Practices

•Start with {{flux_pro}} for most use cases - it provides the best quality
•Use prompt_upsampling to automatically enhance your prompts with more details
•Save your seeds to reproduce or iterate on good results
•Lower guidance (2-3) in {{flux_flex}} for more creative freedom
•Higher guidance (4-5) in {{flux_flex}} for strict prompt adherence

Important: Language Requirements

⚠️ All prompts MUST be in English!

The Stability AI and FLUX APIs only support English prompts. If the user provides a prompt in another language (e.g., Chinese, Japanese, Spanish), you MUST:

•Translate the user's prompt to English before calling the tool
•Enhance the translated prompt with descriptive English keywords
•Keep the original intent and style of the user's request

Translation Example

User request: "画一个可爱的男孩牵着小狗在公园散步"

Translated and enhanced prompt for the tool:

json

{
  "prompt": "A cute young boy walking a friendly dog in a sunny park, warm lighting, heartwarming scene, high quality, detailed illustration, soft colors"
}

Always translate non-English prompts to detailed English descriptions before calling any image generation tool.

Limitations

•Maximum resolution: 1440x1440 (FLUX), 2048x2048 (Stability)
•Processing time: 10-60 seconds depending on model
•Some content may be filtered by safety systems
•API rate limits apply
•Prompts must be in English only