The quality of the prompt directly impacts the quality of the result. Air's AI works best with clear, simple descriptions.
The Basics
Keep it short and natural
One or two sentences, like a photo caption:
"Put [water bottle] on a wooden table"
"Set [coffee], [pastries], [notebook] on a marble flatlay"
"Place [sneakers] on concrete floor in urban setting"
Modern AI models are trained on natural language descriptions, not technical specifications. Simple prompts give the AI room to apply its understanding of composition, lighting, and style.
Lead with subject and style
Start with what the image is and the overall aesthetic:
"[Editorial product photo of sneakers] on concrete"
"[Minimalist hero image with smartphone] remove background and replace with white background"
"[Cosmetic bottles] on pink silk fabric"
Subject first, setting second. This helps the AI prioritize what matters most.
Don't describe lighting, shadows, or camera settings
The AI handles these automatically. Adding technical photography terms just adds noise:
Don't say:
"[Laptop] with soft natural lighting, shallow depth of field and professional composition"
Do say:
"[Laptop] on a clean desk"
The AI already knows what good lighting and composition look like. Let it do its job.
Describe what the image IS, not what to avoid
The AI can't process negative instructions:
Don't say:
"No clutter," "Don't make it busy," "No harsh shadows"
Do say:
"Clean desk," "Minimal background," "Soft scene"
Frame everything positively — describe what should be there, not what shouldn't.
When Editing Existing Images
Be specific about what's in the image
If the user is placing an existing subject into a new scene, the prompt must describe the subject accurately so the AI preserves it:
Good:
"[Two red beer cans] on a sandy beach"
Bad:
"Beers on a beach" (too vague — the AI might substitute different products)
Vague descriptions cause the AI to reinterpret or replace the subject with something similar but wrong.
Don't guess what's in an image
For image references, Air will view the image first to confirm details. Guessing without specifics produces completely unrelated results.
Example:
User: "Put the products from [image] in a kitchen scene"
Construct an accurate prompt like "Take the two champagne flutes and dark champagne bottle from [image] and place them on a kitchen counter in a modern kitchen. Keep the flutes as clear glass stems"
Provide context for the transformation
When editing a scene, describe the desired result clearly:
Good:
"Change the background to a modern office"
"Make this look like it's on a beach at sunset"
"Put this product in a cozy living room"
"Change this to a front-facing angle at eye level"
"Regenerate this from a higher angle with a closer view"
You can also describe the viewpoint you want when editing an existing image. Direction, elevation, or distance can help Air regenerate the scene from the perspective you have in mind.
Vague:
"Change the background" (to what?)
"Make this look better" (how?)
The more specific you can provide about the desired outcome, the better the result.
Understanding AI Behavior
AI is creative, not deterministic
The same prompt can produce slightly different results each time. The AI interprets the description and generates based on patterns it's learned — there's natural variation.
What this means for users:
If the result isn't quite right, ask for adjustments or try again. Small tweaks to the prompt can shift the output significantly.
Generation vs. editing: an important distinction
If you provide any image — even for a completely new scene — Air uses AI editing, not generation.
Editing: The AI sees the source image and transforms it while preserving the subject
Generation: The AI creates from scratch with no reference image
Example:
Provide a product photo and says "put this in a forest scene" → Air uses editing (the AI sees the product and places it in a forest).
Advanced Tips
Iterate and refine
AI results can be adjusted. If the first result is close but not perfect, ask for tweaks:
"Make the background lighter"
"Move the product more to the left"
"Add more space around the subject"
Small adjustments often get to the exact result you want.
Combine AI with manual edits
You don't have to rely on AI alone. Air can combine AI operations with manual edits in one request:
"Remove the background and resize to 1080×1080"
"Change the background to white and add the logo in the top right"
Air chains the operations together seamlessly.
Summary: The Formula for a Good Prompt
[Subject] + [Setting] + [Optional: specific detail]
"Water bottle in a modern home on wooden table"
"Sneakers on concrete in urban setting"
"Two red cans on a sandy beach at sunset"
Keep it short. Keep it clear. Let the AI handle the rest.