AI image generation best practices

Be specific in your prompts

The more detail you include, the better the output.

Try to describe:

• The subject (what is happening)
• The setting (where it takes place)
• The style (photo, illustration, etc.)
• The mood or tone

Example:

Instead of: “a team meeting”

Try: “a photo like image of a diverse team in a modern office having a collaborative meeting, natural lighting, professional setting”

Specify image style

The AI currently generates a mix of styles.

If you want a realistic result, make sure to say:

• “photo like”
• “realistic”

If you want something more visual or stylised, you can ask for:

• Illustration
• Cartoon
• Minimal style

Control image orientation

By default:

• Lessons → 16:9
• Guides → 9:16

If you need something different, include it in your prompt:

• “portrait image”
• “9:16 format”
• “landscape image”

Use seed images effectively (Lessons only)

Seed images help guide the output.

To use them well:

• Upload a clear JPG using Add file
• Reference it in your prompt → “based on the image I attached…”
• Be explicit about what should change → e.g. “based on the image I attached, change the background to an office setting”

Iterate to improve results

You rarely get the perfect image on the first try.

Refine your result by:

• Adding more detail
• Adjusting style or tone
• Being specific about what to change

Example:

“Make the image brighter with more natural lighting and a more modern office background”

Practical tips

To get consistent results:

• Don’t switch from a scratch generated image to a seed image mid flow
• Treat these as separate workflows:
  • From scratch → iterate
  • With seed image → iterate
• Seed images are not supported in Guides

Quick checklist

Before generating an image, ask yourself:

• Have I described the scene clearly?
• Did I specify photo like if needed?
• Did I include the right orientation?
• Am I referencing a seed image correctly (if using one)?