Skip to main content

Generate Your First Image

Let's create your first AI-generated image in Photoshop! This guide will walk you through the complete process.

Prerequisites​

Before you begin, make sure you have:

  • ✅ Installed Cubby Image (Installation Guide)
  • ✅ Set up at least one API key (API Keys Setup)
  • ✅ Accepted the Terms of Use (on first launch)
  • ✅ An open Photoshop document

Step-by-Step Guide​

1. Open Cubby Image​

  1. In Photoshop, go to Plugins → Cubby Image
  2. The plugin panel appears on the right side of your screen

2. Choose Your Model​

Click the Model dropdown to see all available models.

For beginners, we recommend:

  • Nano Banana (Gemini 2.5 Flash Image)
    • Fastest generation times
    • Great for testing and iterations
    • Good quality for most use cases
    • Available on all three providers

For high-quality final work:

  • Seedream 4.0
    • Excellent quality and detail
    • Takes longer to generate
    • Best for final artwork

3. Select Your API Provider​

Choose which API provider to use from the dropdown:

  • Replicate - Widest model selection
  • fal.ai - Fast inference
  • Gemini - Only for Nano Banana

Make sure you have an API key set up for the provider you choose.

4. Write Your Prompt​

In the prompt field, describe the image you want to create.

Example prompts:

A serene mountain landscape at sunset, crystal clear lake, purple and orange sky, photorealistic
A cute robot character illustration, friendly face, colorful design, digital art, high detail
Modern minimalist living room, scandinavian design, natural light, professional interior photography

Prompting tips:

  • Be specific about the subject, style, and mood
  • Include details about lighting and composition
  • Add quality descriptors (photorealistic, detailed, professional, etc.)
  • Different models respond to different prompting styles - experiment!

5. Set Number of Variants (Optional)​

Choose how many variations to generate at once (1-4):

  • 1 variant: Fastest, lowest cost
  • 2-4 variants: Explore different interpretations, costs more

Start with 1 variant to test your prompt, then increase for exploration.

6. Add Context Images (Optional)​

For your first image, you can skip this. But context images let you:

  • Use reference images to guide style
  • Mix and match different images
  • Edit existing images with AI

We'll cover this in the Generating Images guide.

7. Generate!​

Click the Generate button.

The plugin will:

  1. Show a progress bar
  2. Send your request to the API
  3. Display the progress status
  4. Add the generated image(s) as new layer(s) in your document

Typical generation times:

  • Nano Banana: 10-30 seconds
  • Seedream 4.0: 30-90 seconds

8. Review Your Results​

Once complete:

  • The generated image appears as a new layer in your Layers panel
  • The layer is automatically named with the model used
  • Your original layers remain unchanged

Working with Generated Images​

Now that you have your first AI-generated image, you can:

Edit and Refine​

  • Move it with the Move tool (V)
  • Transform it (Cmd/Ctrl + T)
  • Apply filters (Filter menu)
  • Add adjustments (Layer → New Adjustment Layer)
  • Mask parts (Layer Mask button)
  • Blend with other layers (blend modes)

Generate Variations​

  1. Select the generated layer
  2. Click Reuse Prompt in Cubby Image
  3. Try a different model or provider
  4. Click Generate again
Session-Based

The "Reuse Prompt" feature only works for layers generated in the current session. Close and reopen the document, and this data is no longer available.

Organize Your Work​

  • Rename layers with descriptive names
  • Group related generations (Cmd/Ctrl + G)
  • Delete unwanted results to keep your file clean

Tips for Better Results​

Start Simple​

For your first few attempts:

  1. Use simple, clear prompts
  2. Choose Nano Banana for speed
  3. Generate just 1 variant
  4. Learn what works before getting complex

Iterate and Refine​

  1. Generate an initial image
  2. Analyze what you like and don't like
  3. Adjust your prompt accordingly
  4. Generate again
  5. Repeat until satisfied

Experiment with Different Models​

Each model has strengths:

  • Nano Banana: Fast iterations, good for testing concepts
  • Seedream 4.0: Photorealistic, detailed, artistic
  • Flux Kontext Pro/Max: Advanced context understanding
  • QWEN Image Edit Plus: Best for editing existing images

Try the same prompt with different models to see which produces your preferred results.

Be Specific in Prompts​

❌ Vague: "a dog" ✅ Specific: "a golden retriever puppy sitting in a sunny garden, professional photography, shallow depth of field, warm natural lighting"

❌ Vague: "city at night" ✅ Specific: "futuristic cyberpunk city at night, neon lights reflecting on wet streets, rain, cinematic atmosphere, highly detailed"

Troubleshooting​

"API Key not found" or generation fails​

Solution:

  1. Go to Settings (gear icon)
  2. Check that you've entered and saved an API key
  3. Verify you selected a provider that matches your API key
  4. Try clicking Delete and re-entering the key

Generation is very slow​

This is normal for some models:

  • Seedream 4.0 takes longer than Nano Banana
  • Higher quality = longer wait times
  • Be patient, the results are worth it!

If unusually slow:

  • Check your internet connection
  • Try a different API provider
  • Check the provider's status page

Image doesn't match my prompt​

Try these solutions:

  1. Make your prompt more specific and detailed
  2. Try rephrasing with different words
  3. Add style keywords (photorealistic, illustration, etc.)
  4. Try a different model - each interprets prompts differently
  5. Read best practices for your specific model

Image quality is poor​

Solutions:

  1. Use more detailed and specific prompts
  2. Try Seedream 4.0 or Flux Kontext Max for higher quality
  3. Add quality descriptors: "highly detailed", "professional", "4k", "sharp"
  4. Check that you're using a good API key with proper billing
  5. Try generating multiple variants to get better options

What's Next?​

Now that you've generated your first image, explore more features:

Learn Advanced Features​

Explore Models​

Check the APIs & Models reference to learn about:

  • Model capabilities and differences
  • Pricing information
  • Best use cases for each model

Get Inspiration​

Quick Reference​

Basic workflow:

  1. Open Cubby Image panel
  2. Choose model and provider
  3. Write detailed prompt
  4. Click Generate
  5. Wait for result
  6. Edit as needed in Photoshop

Remember:

  • Generated images = new layers (non-destructive)
  • More detail in prompts = better results
  • Different models = different strengths
  • Experiment and iterate!

Happy creating! 🎨