Z-Image-Edit Online
Edit Images with Natural Language Instructions
Transform your images using simple text commands. Z-Image-Edit understands what you want to change and applies edits precisely, from background replacement to object removal and text addition.
What is Z-Image-Edit?
Z-Image-Edit is an instruction-based image editing model that allows you to modify existing images using natural language prompts. Unlike text-to-image generation, Z-Image-Edit preserves your original image while making targeted changes based on your instructions.
What Z-Image-Edit Can Do
Z-Image-Edit excels at precise, localized edits that preserve the original image context:
Replace
Swap backgrounds, change clothing, replace objects while keeping the subject intact
Remove
Clean up unwanted objects, text, watermarks, or people from your images
Add
Insert new elements like text, logos, objects, or accessories into existing scenes
Extend
Outpaint and expand your image canvas while maintaining visual consistency
When to Use Edit vs Turbo (Editing vs Generation)
Choose the right tool for your task:
Use Z-Image-Edit when:
- You have an existing image you want to modify
- You need to preserve specific elements (faces, products, layouts)
- You want to make localized changes without affecting the whole image
- You're doing retouching, cleanup, or targeted adjustments
Use Z-Image-Turbo when:
- You're creating images from scratch
- You need completely new compositions
- You want maximum creative freedom
- Speed is critical and you don't have a source image
How to Use Z-Image-Edit Online (Step-by-Step)
Get started with Z-Image-Edit in three simple steps:
Upload an Image and Write an Edit Instruction
Start by uploading the image you want to edit. Then write a clear, specific instruction describing what you want to change. Be explicit about what should stay the same and what should change.
- • Use high-quality source images for best results
- • Be specific: 'change the sky to sunset' is better than 'make it look better'
- • Mention what should remain unchanged: 'keep the person exactly as is'
Use a Mask or Region Edit for Precise Changes
For more control, define the area you want to edit using a mask or region selection. This tells the model exactly where to apply changes.
Mask Tips: Soften Edges for Natural Blending
- Use soft, feathered edges on your mask to avoid hard cutoffs
- Extend the mask slightly beyond the object you're editing
- For complex edges (hair, fur), use a more generous mask boundary
- Consider using grow_mask_by parameter to automatically expand mask edges
Export, Reuse, and Iterate
Download your edited image and use it as input for further refinements. Iterative editing often produces the best results for complex changes.
- • Save intermediate results for comparison
- • Use edited images as input for multi-step edits
- • Adjust your prompts based on results and re-run
Prompt Templates for Common Edits
Use these proven prompt structures as starting points for your edits:
Change Background Prompts (Keep Subject Unchanged)
Background replacement is one of the most common editing tasks. The key is explicitly telling the model what to preserve.
"Replace the background with a beach sunset, keep the person exactly as they are""Change the background to a modern office interior, preserve the subject's pose and appearance""Put a mountain landscape behind the subject, do not modify the person""Keep the Person's Face and Pose Unchanged" Variants
- ...maintain the subject's exact facial features and expression
- ...preserve the person's identity and body position
- ...keep all human elements identical to the original
- ...do not alter the person in any way, only change the surroundings
Remove Object Prompts (Clean Removal)
Object removal requires clear identification of what to remove and how to fill the space.
"Remove the person on the left side of the image, fill with natural background""Erase the text watermark in the corner, reconstruct the underlying image""Delete the power lines from the sky, maintain the cloud pattern""Remove the coffee cup from the table, extend the table surface naturally"Add Text or Logo Prompts (Placement + Lighting + Perspective)
Adding text or logos requires attention to perspective, lighting, and integration with the scene.
"Add the text 'SALE 50% OFF' in bold red letters on the storefront sign""Place a company logo on the laptop screen, match the screen's angle and lighting""Write 'Welcome' in elegant script on the wall, blend with the interior lighting""Insert product label text on the package, maintain correct perspective"- • Specify font style: bold, italic, script, sans-serif
- • Mention color and size relative to the scene
- • Describe placement precisely: 'top-center', 'on the sign', 'across the banner'
- • Request lighting matching: 'match the ambient lighting of the scene'
Bilingual Instruction Prompts
Z-Image-Edit understands both English and Chinese instructions, and can render text in both languages within images.
"Add Chinese text '欢迎光临' on the shop sign in traditional calligraphy style""Replace the English menu text with Chinese: '今日特价'""Write both English 'Welcome' and Chinese '欢迎' on the entrance banner"Bilingual capability is particularly useful for cross-market campaigns, localization projects, and multi-language content creation.
Common Problems (and Fixes)
Troubleshoot common issues with Z-Image-Edit:
Why the Edit Didn't Apply
If your edit instruction doesn't seem to have any effect, try these solutions:
| Problem | Solution |
|---|---|
| Prompt too vague | Be more specific about what to change and where |
| No region specified | Use a mask to indicate exactly which area to edit |
| Conflicting instructions | Simplify your prompt to focus on one change at a time |
| Low denoise strength | Increase the editing strength to allow more modification |
Why It Changed Too Much
If the model is modifying more than intended, try these approaches:
| Problem | Solution |
|---|---|
| Prompt too broad | Add explicit anchors: 'only change X, keep everything else identical' |
| High denoise strength | Lower the editing strength for more subtle changes |
| Mask too large | Use a tighter, more precise mask around the edit area |
| Missing preservation instructions | Explicitly state what must remain unchanged |
Edge Artifacts and Blending Issues
Hard edges and unnatural blending can be fixed with these techniques:
- Use soft, feathered mask edges instead of hard boundaries
- Apply mask expansion (grow_mask_by) to extend edit boundaries
- Reduce edge sharpness in your mask selection
- Run a second pass focusing on blending problem areas
- Use lower denoise values at edges for gradual transitions
Frequently Asked Questions
Is Z-Image-Edit good at Chinese and English text in images?
Yes, Z-Image-Edit has strong bilingual capabilities. It can understand prompts in both English and Chinese, and can accurately render text in both languages within images. This makes it particularly effective for international marketing materials, localized content, and multilingual designs.
Can it follow creative editing instructions reliably?
Z-Image-Edit is designed to follow natural language instructions with high fidelity. For best results, be specific and explicit in your prompts. Complex creative edits may require iterative refinement, using the output of one edit as input for the next.
What image formats are supported?
Z-Image-Edit supports common image formats including JPEG, PNG, and WebP. For best results, use high-resolution source images with clear subjects.
How is Z-Image-Edit different from Photoshop or manual editing?
Z-Image-Edit uses AI to understand and execute edits based on natural language, eliminating the need for manual selection, masking, and manipulation tools. It can make intelligent decisions about how to implement your requested changes while maintaining image coherence.
Can I use Z-Image-Edit for commercial projects?
Yes, images edited with Z-Image-Edit can be used for commercial purposes. Check our terms of service for specific licensing details and usage guidelines.
What's the best way to handle complex multi-step edits?
Break complex edits into sequential steps. Edit one element at a time, save the result, and use it as input for the next edit. This iterative approach gives you more control and produces more reliable results.
Ready to Edit Your Images?
Start using Z-Image-Edit today and transform your images with natural language instructions.
Get Started