✅ Best for

• Creating images from scratch — describe a scene and get a photorealistic image with no source photo required
• Concept art & mockups — "a person standing in front of a yellow wall, studio lighting, professional headshot style"
• Product imagery — generate stylised product shots from a text description
• Background plates — generate a background to composite into other work
• Inspiration & references — quickly visualise lighting setups or scene compositions before shooting
• Marketing & social content — produce on-brand images from a brief

💡 Tips

• Claude will expand your description into a detailed, model-optimised prompt before generating
• Include style keywords: "photorealistic", "shallow depth of field", "studio lighting", "cinematic"
• Mention the subject, setting, lighting, mood and camera angle for the best results

❌ Not ideal for

• Editing an existing photo — this generates a brand new image, not a modified version of your upload
• Preserving a specific person's likeness — no reference image is used
• Pixel-precise edits to an existing shot (use Nano Banana 2 Photo Editing instead)

✅ Best for

• Complex, multi-step instructions — "remove the glare, brighten the background, and sharpen the face" in a single prompt
• High-fidelity portrait work — same face-preserving capability as Nano Banana 2, with more faithful prompt interpretation
• Professional client deliverables — when output quality matters and cost is secondary
• Text within images — can accurately add or edit text rendered inside photos

💡 Tips

• Use Nano Banana 2 first — only upgrade to Pro when you need more precise instruction following
• Requires Gemini Pro Image preview access — if you see a quota error, your account doesn't have access yet

❌ Not ideal for

• Quick test edits — use Nano Banana 2 first at half the cost
• High-volume use — at £0.10/edit, 100 edits/month = £10
• Accounts without Pro Image API access (restricted preview)

✅ Best for

• Adding objects to a scene — paint a mask over an empty area and describe what to add ("a red parrot", "a coffee cup")
• Background replacement — mask the background, describe the new one
• Removing objects — mask an object and describe what should fill the space
• Creative compositing — adding elements that don't exist in the original photo at all
• Budget testing — the cheapest way to try the plugin workflow

💡 Tips

• Always paint a mask — without one, dall-e-2 makes very conservative changes
• Paint only the area you want changed — smaller masks = better results
• Output is always 1024×1024 square (center-cropped from your original)
• Does not preserve face identity — avoid masking over faces

❌ Not ideal for

• Glare or blemish removal on faces — replaces the entire masked area with a new face
• Edits that require keeping a specific person looking like themselves
• Non-square images — the center-crop may lose parts of portrait or landscape shots

✅ Best for

• Same use cases as dall-e-2 — adding objects, replacing backgrounds, removing items
• When you need a larger file — output is upscaled from 1024px to 2048px on the server using GD, giving a bigger canvas at no extra API cost
• Web banners & compositing — where you need a larger image to work with in a design tool

💡 Tips

• The upscale is done by PHP GD on your server — it enlarges pixels but doesn't add new detail
• For sharper large-format results, Nano Banana 2 at its native resolution produces better quality
• Same mask rules apply as dall-e-2

❌ Not ideal for

• Portrait retouching — same face identity limitation as dall-e-2
• Situations where you need genuine 2K resolution detail — GD bicubic upscaling simply enlarges existing pixels