Generate Image

​

Overview

• Text-to-image generation - Create any visual from natural language descriptions
• Multiple aspect ratios - From square social media posts to ultrawide banners
• Style control - Specify artistic styles, moods, and aesthetics
• Reproducibility - Use seeds to regenerate the same image consistently
• Format flexibility - Output as JPEG or PNG based on your needs
• Content safety - Built-in moderation with configurable tolerance levels

• Marketing and advertising - Generate unique visuals for campaigns
• Content creation - Illustrations for blogs, articles, and social media
• Product design - Rapid prototyping and concept visualization
• E-commerce - Product mockups and lifestyle images
• Gaming and entertainment - Concept art and asset creation
• Education - Visual aids and instructional materials
• Personalization - Custom visuals for user-generated content

​

How Image Generation Works

1. Prompt Analysis - Your text description is analyzed to understand subjects, styles, composition, and mood
2. Latent Generation - The model generates the image in a compressed latent space
3. Refinement - Multiple denoising steps progressively improve image quality
4. Upscaling - The final image is upscaled to high resolution
5. Safety Check - Content moderation filters are applied
6. Delivery - The image is stored temporarily and a URL is returned

​

Writing Effective Prompts

​

Prompt Structure

[Subject] [doing/being what] [setting/environment] [style/mood] [technical details]

• “A golden retriever puppy playing in a meadow, soft lighting, photorealistic, 8k”
• “Futuristic cityscape at night, neon lights, cyberpunk style, cinematic composition”
• “Watercolor painting of a lighthouse on rocky coast, pastel colors, impressionist”

​

Key Elements

• Be specific: “tabby cat” vs “cat”
• Include details: “elderly man with white beard” vs “man”
• Specify quantity: “three colorful hot air balloons”

• “running through forest”
• “floating in space”
• “sitting peacefully”

• “in a modern kitchen”
• “on a tropical beach at sunset”
• “deep underwater coral reef”

• Art styles: “oil painting,” “digital art,” “pencil sketch,” “3D render”
• Mood: “dramatic,” “peaceful,” “energetic,” “mysterious”
• Photography: “portrait mode,” “wide angle,” “macro photography”

• “highly detailed,” “8k resolution,” “sharp focus”
• “cinematic lighting,” “volumetric lighting,” “studio lighting”
• “shallow depth of field,” “bokeh background”

​

Prompt Tips

• ✅ Be specific and descriptive
• ✅ Use style references (“like Pixar animation”)
• ✅ Specify lighting and atmosphere
• ✅ Include composition details (“centered,” “close-up,” “aerial view”)
• ✅ Mention quality (“high quality,” “detailed,” “professional”)

• ❌ Use vague descriptions (“nice image,” “good looking”)
• ❌ Include multiple conflicting styles
• ❌ Over-complicate with too many subjects
• ❌ Use negative instructions (model interprets positively)

​

Examples

from incredible_python import Incredible

client = Incredible(api_key="YOUR_API_KEY")

response = client.generate_image(
    prompt="A red circle",
    aspect_ratio="1:1"
)

print("Image URL:", response.image_url)
print("Seed:", response.seed)

​

Request Parameters

​

prompt (required)

​

aspect_ratio (optional)

• 1:1 (Square) - 1024x1024px
  • Use for: Social media posts, profile pictures, icons
  • Platforms: Instagram posts, Twitter, LinkedIn
• 16:9 (Landscape) - 1920x1080px
  • Use for: YouTube thumbnails, presentations, website headers
  • Platforms: YouTube, blog headers, slides
• 9:16 (Portrait) - 1080x1920px
  • Use for: Mobile content, Stories, vertical video thumbnails
  • Platforms: Instagram Stories, TikTok, Snapchat
• 4:3 (Traditional) - 1024x768px
  • Use for: Traditional displays, presentations
  • Classic format for standard content
• 3:4 (Tall portrait) - 768x1024px
  • Use for: Magazine covers, book covers, posters
  • Traditional portrait orientation
• 21:9 (Ultrawide) - 2560x1080px
  • Use for: Banners, cinematic content, wide displays
  • Dramatic panoramic scenes
• 9:21 (Tall vertical) - 1080x2560px
  • Use for: Full-screen mobile experiences, tall banners
  • Maximum vertical canvas

​

output_format (optional)

• Smaller file size
• Faster generation
• No transparency
• Best for: Photographs, realistic scenes, web content

• Larger file size
• Lossless compression
• Supports transparency
• Best for: Graphics with transparency, logos, when quality is critical

• You need transparent backgrounds
• Image will be further edited
• Absolute quality preservation is required
• File size is not a concern

​

seed (optional)

• Consistency - Generate matching images across a series
• Iteration - Make minor prompt adjustments while maintaining core image
• A/B testing - Test variations systematically
• Version control - Recreate exact images later

# Generate initial image
response1 = client.generate_image(prompt="sunset over ocean")
seed = response1.seed  # Save the seed

# Generate similar image with same seed
response2 = client.generate_image(
    prompt="sunset over ocean with sailboat",  # Slightly modified
    seed=seed  # Same seed for consistency
)

​

prompt_upsampling (optional)

• Adds quality modifiers (“highly detailed,” “professional quality”)
• Improves style consistency
• Adds technical terms for better rendering

• Input: "mountain landscape"
• Upsampled: "majestic mountain landscape with snow-capped peaks, golden hour lighting, highly detailed, professional photography, 8k resolution"

• ✅ Short or basic prompts
• ✅ When you want maximum quality automatically
• ✅ For users who aren’t prompt engineering experts

• ❌ You’ve already written a detailed prompt
• ❌ You want precise control over every detail
• ❌ Testing specific prompt formulations

​

safety_tolerance (optional)

• 0 - Most strict: Blocks anything remotely questionable
• 2 - Moderate (default): Blocks clearly inappropriate content
• 6 - Most permissive: Allows artistic and mature themes

• Violence and gore
• Adult content
• Hate symbols
• Illegal activities
• Harmful or dangerous content

• 0-1: Children’s content, educational materials, public-facing apps
• 2-3: General purpose applications, business content
• 4-6: Artistic projects, mature audiences, where context matters

​

Best Practices

• Iterate on prompts - first generation rarely perfect
• Use style references from known artists or movements
• Include lighting details for photorealistic images
• Specify camera angles and composition
• Add quality modifiers (“highly detailed,” “sharp focus”)

• Cache frequently used images
• Generate at lower resolutions for prototypes
• Use seeds to avoid regenerating the same image
• Batch similar requests together

• Be specific about style and mood
• Include technical photography/art terms
• Experiment with different aspect ratios
• Try prompt_upsampling for automatic enhancement

• Dynamic content - Generate unique images for each user/session
• Asset libraries - Bulk generate variations for design systems
• Personalization - Create customized visuals based on user data
• Rapid prototyping - Quickly visualize concepts for stakeholder review

​

Important Notes

import requests
from pathlib import Path

# Generate image
response = client.generate_image(prompt="...")

# Download immediately
image_data = requests.get(response.image_url).content
Path("generated_image.jpg").write_bytes(image_data)

​

Response

{
  "success": true,
  "image_url": "https://storage.googleapis.com/...",
  "request_id": "abc123",
  "seed": 42,
  "revised_prompt": "Enhanced version of your prompt"
}