Official Usage Guide

How to Use AI Image Editor

This guide explains the recommended AI Image Editor workflow, from preparing source images to writing stronger prompts, reviewing output quality, and deciding when to iterate.

Input Control

Up to 3 reference images

Runtime

Usually 1 minute

Billing

10-40 credits

Best Use

Prompt-led visual edits

Recommended Workflow

The best results usually come from a controlled sequence: prepare the right inputs, write a focused prompt, add exclusions, and then review the result critically before shipping it.

1

Prepare your image input

Choose a clear source image and, if needed, up to three reference images that show the style, context, or visual direction you want.

2

Describe the intended change

Write a direct prompt that explains what should change, what should stay consistent, and what quality or style outcome you expect.

3

Add guardrails with negative prompts

Use negative prompts to exclude artifacts, low-quality output, unwanted objects, or styling mistakes before you generate.

4

Review and iterate

Check the result, adjust your instructions, and run another pass if needed to improve realism, composition, or commercial readiness.

Before You Start

  • Use a source image with a clear subject and enough detail for the model to understand the scene.
  • Decide whether your prompt is asking for a precise edit, a style change, or a broader transformation.
  • Prepare reference images only when they materially improve the outcome; adding extra inputs can sometimes make intent less clear.

Prompt Writing Framework

State the main edit first

Lead with the highest-priority change, such as replacing a background, refining lighting, changing outfit details, or removing an object.

Anchor what should stay consistent

Mention any elements that should remain untouched, such as subject identity, product shape, framing, or brand colors.

Add output quality guidance

Include words that describe the final standard you need, such as clean, realistic, studio-ready, polished, natural, or e-commerce ready.

Prompt and Negative Prompt Examples

Use these examples as starting structures, then adapt them to your own subject and workflow.

Catalog image cleanup

Primary Prompt
Replace the background with a clean light-gray studio backdrop, keep the product shape exactly the same, improve edge separation, and make the final image look ready for an online catalog.
Negative Prompt
blurry edges, extra objects, warped product shape, harsh shadows, low detail

Lifestyle upgrade

Primary Prompt
Turn this product shot into a warm lifestyle scene on a modern desk, keep the product branding readable, preserve realistic lighting direction, and maintain a premium commercial look.
Negative Prompt
text distortion, unrealistic reflections, cluttered composition, cheap-looking colors

Portrait refinement

Primary Prompt
Refine the portrait with more balanced lighting, cleaner skin texture, and a subtle professional background blur while keeping the face natural and recognizable.
Negative Prompt
plastic skin, over-sharpening, distorted facial structure, artificial expression

Review Checklist

  • Check whether the model preserved the subject, composition, and key brand details you wanted to keep.
  • Zoom in on edges, hands, text, and product surfaces to catch artifacts before publishing.
  • If the output is close but not ready, reuse the result as a reference and tighten the next prompt instead of rewriting from scratch.
  • For business use, verify that the final image matches your visual standards before using it in ads, listings, or public materials.

Common Fixes

The result is too generic

Add more concrete direction about scene, material, lighting, or intended use. Generic prompts tend to produce generic edits.

Important elements changed unexpectedly

State what must remain unchanged and reinforce that instruction in both the main prompt and negative prompt.

The output feels unrealistic

Reduce the number of simultaneous changes and ask for one controlled improvement at a time, especially on faces, products, and text-heavy assets.