Small defects block ad launch: a specular highlight washes out a label, a stray prop enters frame, or a generated lifestyle scene needs warmer light without rebuilding the entire batch. Rhinobird’s Image Editor skill applies targeted visual changes through natural-language instructions in Agent Chat. This guide shows how to open the skill, phrase edits merchants can audit, and save revised assets from History.
Open Image Editor
Launch Image Editor. Confirm the session shows Image Editor as the active skill. If you land in a general chat, reopen the skill URL so edit intents route to the retouch workflow and credit display on the skill button.
Upload the right source file
Attach the most recent approved image—exported from History, from AI E-commerce Image Sets, or from your DAM. Editing an outdated intermediate wastes credits when you must redo changes on the newer parent file.
For transparency or glass packaging, note in text which regions must stay untouched (barcode, nutrition panel, regulatory icons). The editor follows boundaries better when you name them explicitly.
Write precise edit instructions
Effective prompts name location, desired change, and constraints:
- “Reduce specular highlight on front label; keep text sharp.”
- “Remove coffee cup on the right; fill with blurred countertop.”
- “Warm overall white balance slightly; do not shift product red.”
- “Crop to 1:1 centered on bottle; preserve top cap visibility.”
Avoid contradictory requests in one message (“make background busy but minimal”). Stack edits sequentially: fix glare, review download, then request crop so you can roll back a bad step via History versions.
Run the skill and review fidelity
Send the prompt, then trigger the skill action with displayed credits. Sign in via Login / Sign up if prompted. After completion, compare product identity against the upload—logo geometry, cap color, fill level for liquids.
If the model invents label text, revert to the prior History entry and retry with “do not alter printed text” plus a tighter crop on the label zone. For heavy structural changes, regenerating from source tools may cost less than chained edits.
Combine with matting and upscale
When edits expose rough edges after background changes, run AI Background Removal on the intermediate, then return to Image Editor for color work. If final exports need sharper micro-detail for print, follow with Image Enhancement after edits are approved.
When not to use Image Editor
Full scene replacements belong in Background Generation. New handheld model shots belong in Model Handheld. Image Editor excels at surgical fixes on existing pixels, not net-new campaign concepts from scratch.
Failure states
- Over-edited product shape — reduce prompt scope; attach a higher-resolution source.
- Login loop — complete sign-in on the same tab, rerun skill.
- No visible change — specify measurable targets (“20% darker shadow under base”).
- Wrong file saved — confirm History timestamp before uploading to ads manager.
Natural-language editing keeps PDP and ad iterations inside Rhinobird when prompts are localized, sequential, and validated against product identity before media spend goes live.
Keep a short log of prompts that worked for each SKU category—cosmetics glare behaves differently from brushed metal appliances—and reuse those phrases in Agent Chat so reviewers see consistent, auditable edit language across campaigns.