Best Practices
Prompts
Raspberry AI
Best Practices

Prompts

A prompt consists of words, phrases or sentences that describe the desired appearance of an image. Raspberry AI analyzes these words to guide the creation process and generate the image you envision. Prompts are the way you communicate your vision to Raspberry, so that an image can be generated. Even the simplest prompts will create an image, and it takes practice to get the images you want.

Be Precise!
When prompting for apparel or accessories, it is important to include:
  1. Color
  2. Material
  3. Product - what the garment or accessory is.
If your image has an AI model, you can be just as descriptive about them. Describe ethnicity, hairstyle and color, pose, age and more.
Be Descriptive!

Prompt boxes respond to both descriptive language and clear command-style instructions. Words like “change,” “add,” “remove,” “replace,” or “place” can be useful — especially when you are directing a specific update to an image.

That said, commands work best when they are paired with visual detail. The platform still needs to understand what you want the final image to look like, not just what action to take.

For example, instead of only writing:
“Change jacket to brown.”

Include descriptive detail so the change is unambiguous:
“Change jacket to a dark brown brown corduroy jacket with visible texture.”

When using a reference image, avoid vague instructions that don’t clearly identify what should change. The system doesn’t inherently know which part of the image you’re referring to unless you describe it.

Rather than:
“Change jacket to brown.”

Describe the element you want updated:
“Dark brown corduroy jacket.”

If you’re unsure how to describe what’s already in the image, you can use Describe Image as a shortcut, then edit only the part of the description you want to change.

Being descriptive can also include details beyond the product itself, such as:

  • The model (pose, age, hair, expression)
  • Lighting and mood
  • Camera angle or framing
  • How reference images should be interpreted

Clear, visual descriptions help the platform produce more predictable, intentional results.

Keep Going!
Prompting is an iterative process. Even skilled prompt engineers have to build upon their first prompts to get closer to their desired outcome. If you get close to your vision, but not quite what you wanted, you have a couple of options.
  1. Use the generated image that is closest to your desired outcome as a reference image. Describe the image and tweak the prompt to your liking.
  2. If only small changes are needed, and you want most of the image to stay as-is, bring it into the Edit module to make more precise changes.