Skip to main content
Prompting Best Practices for Images (Runable Recommendations) To get consistently strong results, focus on writing clear, specific prompts that describe exactly what you want to create. Start simple, then refine with intentional edits.

​
Use Canvas mode for visual adjustments

For post-generation changes like repositioning, resizing, cropping, or layout tweaks, use Canvas mode instead of regenerating. Canvas lets you:
  • Reposition images by dragging
  • Resize/scale images with the transformer
  • Crop to different aspect ratios
  • Mark-edit to add/remove elements with AI inpainting
  • Text edit to modify text in images
  • Combine multiple images into sequences
This helps you refine faster and reduce credit usage. Reserve regeneration for changes that require completely new imagery or fundamental style/lighting/composition changes.
Note: Canvas works on already-generated images. For changes before generation (like adjusting the prompt’s composition or lighting), edit your prompt and regenerate.
Be specific and direct Describe the subject, style, composition, and mood in clear terms. Avoid vague instructions that don’t guide the output.
  • ✅ “Product photo of wireless headphones on marble surface, soft lighting, minimalist”
  • ❌ “Make it look professional”
Refine with focused instructions When improving results, change one element at a time (like lighting, composition, or color palette). This helps the AI apply updates more accurately.
  • ✅ “Increase contrast and add warmer tones”
  • ❌ “Make everything better and more vibrant and add more details”
Choose the right tool for the job
  • Image generation: Create new visuals from scratch
  • Edit with AI: Refine existing images with style or mood changes
  • Background removal/replacement: Clean product shots or change scenes
  • Outpaint: Extend images to new aspect ratios
  • Upscale: Enhance resolution for web and export
  • AI Try-On: Preview clothing and styles
  • Inspirations: Fast design with instant image swaps
Only include actionable details Focus on elements the AI can control: subject, style, composition, lighting, color palette, mood, aspect ratio, and specific objects. Skip commentary that doesn’t change the output. Keep the structure simple Subject → Style → Context Examples:
  • “Mountain landscape → dramatic sunset, vibrant colors → for website hero”
  • “Product shot of coffee mug → clean white background, soft shadows → e-commerce listing”
  • “Portrait of business professional → natural lighting, neutral backdrop → LinkedIn profile”
Optimize for forward progress Use prompts that move toward your goal rather than asking for review.
  • ✅ “Add more negative space around the product”
  • ✅ “Reduce background clutter”
  • ✅ “Shift lighting to golden hour”
  • ❌ “Does this look good?”
  • ❌ “Can you improve this?”