Image Generation API¶

Learn how to generate AI-powered images using PrimeThink's Image Generation API. This guide focuses on best practices and effective usage patterns to help you get the most out of image generation capabilities.

Overview¶

The Image Generation API allows you to: - Generate images from text descriptions (text-to-image) - Use reference images to guide style and composition - Retrieve previously generated images - Control image dimensions, style, and quality

Authentication¶

All API requests require authentication. See API Authentication for details on obtaining and using your API key.

API Endpoints Quick Reference¶

For complete API specifications including all parameters, request/response formats, and error codes, see the Interactive API Documentation.

GET /images/{image_id}¶

Retrieves a previously generated image. Optionally resize by specifying a width parameter.

POST /images/tti¶

Generates images from text descriptions (text-to-image). Key parameters include: - prompt - Text description of the image (required) - provider - Provider: auto, openai, or google (default: user setting or auto) - size - Dimensions like 1024x1024, 1536x1024, 1024x1536 (default: 1024x1024) - style - Image style (default: realistic) - reference_images - Array of reference image URLs - reference_weight - Influence of reference images (0.0-1.0, default: 0.5) - count - Number of variations to generate (1-4, default: 1) - negative_prompt - What to exclude from the image - name - Custom filename for the generated image (without extension) - folder - Destination folder for saving (default: images)

Best Practices¶

Writing Effective Prompts¶

The quality of generated images depends heavily on how you craft your prompts. Here are proven strategies:

Be Specific and Detailed¶

Include concrete details about what you want to see. Vague prompts produce unpredictable results.

Good Examples: - "A professional headshot of a woman in her 30s wearing business attire, soft natural lighting from the side, neutral gray background, sharp focus" - "A cozy coffee shop interior with wooden furniture, warm Edison bulb lighting, plants on shelves, morning sunlight through large windows" - "A minimalist product photo of a blue ceramic mug on a white marble surface, soft shadows, overhead view"

Poor Examples: - "A woman" (too vague) - "Coffee shop" (lacks visual details) - "Product photo" (no specifics)

Structure Your Prompts¶

Organize prompts in this order for best results:

1. Subject - What is the main focus?
2. Details - Colors, materials, features
3. Environment - Setting, background, context
4. Lighting - Natural, artificial, mood
5. Style - Artistic approach or photography type
6. Quality markers - Technical descriptors

Example: "A red sports car [subject], gleaming metallic paint [details], on a mountain road at sunset [environment], golden hour lighting [lighting], cinematic photography style [style], high resolution, sharp focus [quality]"

Use Style Keywords Effectively¶

Add style descriptors to control the artistic approach:

• Photography: "professional photography", "DSLR", "bokeh", "macro lens", "aerial view"
• Art Styles: "oil painting", "watercolor", "digital art", "pencil sketch", "ink drawing"
• Visual Quality: "photorealistic", "highly detailed", "8k resolution", "crisp", "vibrant colors"
• Mood/Tone: "cinematic", "dramatic", "serene", "moody", "bright and airy"

Master Negative Prompts¶

Negative prompts are crucial for avoiding unwanted elements. Always include common issues:

Essential Negatives: - Quality issues: "blurry", "low quality", "pixelated", "distorted", "artifacts" - Unwanted elements: "text", "watermark", "logo", "signature" - Composition issues: "cropped", "out of frame", "extra limbs" (for people)

Use Case Specific: - Product photos: "cluttered background", "shadows", "reflections" - Portraits: "extra fingers", "deformed", "bad anatomy" - Landscapes: "people", "buildings" (if you want nature only)

Working with Reference Images¶

Reference images are powerful tools for maintaining consistent style and composition across multiple generations.

When to Use References¶

• Maintaining brand visual consistency
• Matching a specific artistic style
• Creating variations of an existing image
• Guiding color palettes and composition

Reference Weight Guidelines¶

The reference_weight parameter controls how much influence reference images have:

• 0.3-0.4: Light inspiration, mostly follow the text prompt
• 0.5-0.6: Balanced approach (recommended starting point)
• 0.7-0.8: Strong style matching, text prompt adds details
• 0.9-1.0: Very close to reference style (use sparingly)

Start with 0.5 and adjust based on results. Higher weights may override important aspects of your text prompt.

Using Multiple References¶

When providing multiple reference images: - They blend together to influence the final style - Limit to 2-3 images for best results - Use references with similar styles for coherent output - Different references create interesting hybrid styles

Choosing the Right Image Size¶

Select dimensions based on your specific use case:

Tip: Use auto to let the system choose optimal dimensions based on your prompt content.

Generating Multiple Variations¶

Use the count parameter (1-4) to generate multiple variations in a single request:

Benefits: - More cost-effective than separate requests - All variations processed simultaneously - Easier to compare options - Respects rate limits efficiently

When to Use: - Exploring different interpretations of a concept - A/B testing visual options - Creating a set of related images - Increasing chances of getting a perfect result

Iterative Refinement Strategy¶

Getting the perfect image often requires iteration:

1. Start broad - Generate with a basic prompt to see what works
2. Analyze results - Identify what's good and what needs improvement
3. Refine prompt - Add specific details or negative prompts based on observations
4. Use references - If you got close, use that image as a reference for the next generation
5. Adjust parameters - Try different sizes or styles if needed

Rate Limiting and Cost Management¶

Be strategic with your API usage:

• Batch requests: Use count parameter instead of multiple separate calls
• Cache results: Store and reuse generated images when appropriate
• Iterate efficiently: Review results before generating more variations
• Right-size images: Don't generate 1792x1024 if you only need 512x512
• Test prompts: Use smaller sizes during testing, then generate final images at full resolution

Common Mistakes to Avoid¶

1. Overly complex prompts - Keep prompts focused; too many details can confuse the model
2. Ignoring negative prompts - Always specify what you don't want
3. Wrong aspect ratio - Choose dimensions that match your use case
4. Inconsistent requests - Save successful prompts for reuse and consistency
5. Not using references - Reference images dramatically improve consistency
6. Skipping quality keywords - Always include "high quality", "detailed", or similar terms

Quick Start Examples¶

For complete code examples in multiple languages, see the Interactive API Documentation.

Basic Image Generation¶

curl -X POST "https://api.primethink.ai/images/tti" \
  -H "Authorization: Token YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A serene mountain landscape at sunset, golden hour lighting, highly detailed",
    "size": "1536x1024",
    "negative_prompt": "people, buildings, text, blurry"
  }' \
  --output image.png

With Reference Images¶

curl -X POST "https://api.primethink.ai/images/tti" \
  -H "Authorization: Token YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Modern office interior in the same style",
    "reference_images": ["https://example.com/style-ref.jpg"],
    "reference_weight": 0.6,
    "size": "1024x1024"
  }' \
  --output image.png

Generate Multiple Variations¶

curl -X POST "https://api.primethink.ai/images/tti" \
  -H "Authorization: Token YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Abstract geometric patterns, vibrant colors, modern art style",
    "size": "1024x1024",
    "count": 4
  }' \
  --output images.zip

With Custom Filename¶

curl -X POST "https://api.primethink.ai/images/tti" \
  -H "Authorization: Token YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Product hero image for marketing campaign",
    "size": "1536x1024",
    "name": "hero-banner",
    "folder": "marketing"
  }' \
  --output image.png

Specify Provider¶

curl -X POST "https://api.primethink.ai/images/tti" \
  -H "Authorization: Token YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Photorealistic mountain landscape at golden hour",
    "provider": "openai",
    "size": "1536x1024"
  }' \
  --output image.png

Need Help?¶

If you encounter any issues or have questions about the Image Generation API, please contact our support team at support@primethink.ai or visit our community forum.