IntermediateUpdated Dec 18, 2025
Text-to-Image Masterclass
Master the art of prompting with techniques for consistent, high-quality image generation including negative prompts and style control.
SC
Sarah Chen
Developer Advocate
12 min read
The Anatomy of a Great Prompt
A well-structured prompt typically includes:
- Subject - What you want to generate
- Style - Artistic style or medium
- Details - Specific attributes and features
- Lighting - Mood and atmosphere
- Quality modifiers - Technical quality terms
Basic to Advanced Prompts
Basic Prompt
text
A cat sitting on a chair
Improved Prompt
text
A fluffy orange tabby cat sitting on a vintage wooden chair, soft natural lighting from a nearby window, cozy home interior, photorealistic, high detail, 8k resolution
Style Modifiers That Work
Photography Styles
professional photographycinematic lightingstudio portraitgolden hour photographymacro photography
Art Styles
oil painting stylewatercolor illustrationdigital artanime styleconcept art
Quality Modifiers
highly detailed8k resolutionmasterpieceaward-winningprofessional
Using Negative Prompts
Negative prompts tell the model what to avoid:
python
result = client.run("fal-ai/flux/dev", {
"input": {
"prompt": "A professional headshot of a business woman",
"negative_prompt": "blurry, low quality, distorted, ugly, deformed hands",
"num_inference_steps": 28
}
})Common Negative Prompts
text
blurry, low quality, low resolution, ugly, distorted, deformed, extra limbs, bad anatomy, watermark, text, signature, cropped, out of frame
Controlling Image Composition
Aspect Ratios
python
# Portrait (social media stories)
{"width": 768, "height": 1344}
# Landscape (cinematic)
{"width": 1344, "height": 768}
# Square (Instagram)
{"width": 1024, "height": 1024}
# Ultrawide (banners)
{"width": 1536, "height": 640}Composition Keywords
centered compositionrule of thirdsclose-up shotwide angle shotbird's eye viewlow angle shot
Advanced Techniques
Weighted Prompts
Some models support emphasis with parentheses:
text
(beautiful sunset:1.3), mountains, lake, (vibrant colors:1.2)
Style Mixing
Combine multiple styles:
text
A portrait in the style of (Rembrandt:0.7) and (anime:0.3)
Model Selection Guide
| Model | Best For | Speed |
|---|---|---|
| FLUX Schnell | Quick iterations | ~1s |
| FLUX Dev | High quality | ~5s |
| FLUX Pro | Production | ~3s |
| SDXL | Versatile | ~4s |
Complete Example
python
from abstrakt import AbstraktClient
client = AbstraktClient()
result = client.run("fal-ai/flux/dev", {
"input": {
"prompt": """
A majestic white wolf standing on a snowy mountain peak,
aurora borealis dancing in the night sky behind,
cinematic lighting, photorealistic, highly detailed fur,
8k resolution, national geographic style photography
""",
"negative_prompt": "cartoon, illustration, blurry, low quality",
"image_size": {"width": 1344, "height": 768},
"num_inference_steps": 28,
"guidance_scale": 7.5
}
})
print(result.images[0].url)Next Steps
- Try image-to-image for variations
- Learn LoRA fine-tuning for custom styles
- Explore batch processing for efficiency
#prompting#flux#image-generation#techniques