Repair & restore
Scratches, stains, and missing areas rebuilt in the original’s own style.
Qwen image-to-image · single-reference editing
Qwen Image Edit
Upload one photo. Describe the change. Keep everything else.
Qwen Image Edit applies exactly the change you ask for — restoration, annotations, restyles, background swaps — while preserving the subject, lighting, and composition of your original. Edit online below, free credits included.
The live tool runs Qwen image-to-image through the KIE backend — one reference image per edit, 4 credits per result. The labeled before/after examples below are official Qwen team outputs illustrating the same editing capability.
1 reference
Single-image faithful edits
4 credits
Per edited image
Text-aware
Readable labels & captions
Style-safe
Keeps identity & lighting
Editing that respects the original
Most AI editors regenerate the whole frame and hope for the best. Qwen image-to-image is built around a stricter contract: change what the instruction names, preserve everything it doesn't — which is what makes it usable for restoration and technical annotation, not just restyling.
Before / after · official Qwen examples
What a single instruction can do
Two edits published by the Qwen team that show the range: a damaged artwork rebuilt in its own style, and a plain photo upgraded into a publication-ready figure. The tool above runs the same image-to-image capability on your uploads.
Restore a damaged painting
Given a worn, spotted ink scroll, the model rebuilds missing areas while keeping the brushwork, gradients, and composition faithful to the original.
Instruction
Repair the damage and mold spots while preserving the original ink-wash style and brush strokes.
Before
After
Turn a photo into a labeled figure
From one insect photo, the model keeps the subject intact and adds taxonomy, morphology labels, a magnified inset, and a scale bar.
Instruction
Keep the insect, add taxonomic classification, morphology labels, a zoomed detail view and a scale bar.
Before
After
Annotate & label
Turn photos into figures: labels, insets, scale bars, readable captions.
Restyle faithfully
Watercolor, ink wash, poster looks — with the subject unmistakably intact.
Swap & adjust
Backgrounds, outfits, and props changed without touching identity.
How to edit with Qwen online
1
Upload your reference
The tool above is already in Edit Image mode — drop in one JPG, PNG, or WebP.
2
Describe the change
Name what changes and what stays: “swap the background to a rainy street, keep the pose and lighting”. Specific instructions beat vague ones.
3
Generate and compare
Run the edit, compare with your original, and refine the instruction. At 4 credits per pass, iteration stays cheap.
Qwen Image Edit next to the other editors here
Pick Qwen for faithful single-reference edits; switch models when you need multiple references or poster typography from scratch.
| Capability | Qwen Image Edit | Nano Banana Edit | Wan 2.7 Image | GPT Image 2 |
|---|---|---|---|---|
| Credits / edit | 4 | 4 | 6 | 10 |
| Reference images | 1 | 1 | Up to 9 | Up to 5 |
| Subject preservation | Excellent | Good | Strong | Strong |
| Text & annotations | Excellent | Basic | Good | Excellent |
| Restoration | Excellent | Good | Good | Good |
| Best fit | Repairs, labels, faithful restyles | Quick everyday retouches | Multi-reference composites | Typography-heavy posters |
Explore more image models
Qwen Image (T2I)
4 credits / image
The text-to-image side of the same family — dense layouts, posters, and multilingual typography from scratch.
OpenWan 2.7 Image
6 credits / image
Multi-reference editing with up to 9 input images and native-2K output.
Try nowZ-Image Turbo
2 credits / image
Need a fresh image instead of an edit? Draft photoreal concepts at the lowest cost here.
OpenQwen Image Edit FAQ
What is Qwen Image Edit?
Qwen Image Edit is the image-to-image mode of Alibaba’s Qwen-Image model family. You upload one reference photo, describe the change in plain language, and the model applies it while preserving the subject — restoration, restyling, adding annotations, or swapping backgrounds.
Can I use Qwen Image Edit online without installing anything?
Yes. The editor at the top of this page runs Qwen image-to-image online through the KIE backend. Upload a photo, type the instruction, and generate — no ComfyUI, no local GPU, and new Google sign-ups get free credits.
What edits is Qwen Image Edit best at?
Its standout skills are edits that must respect the original: repairing damaged photos and paintings, adding labels or annotations to turn a photo into a labeled figure, changing outfits or backgrounds while keeping identity, and restyling a picture into another medium like ink wash or watercolor.
How many reference images can I upload?
Qwen image-to-image works from a single reference image — that focus is why it preserves the subject so faithfully. If you need to combine multiple references in one edit, use Wan 2.7 Image on this site, which accepts up to 9 inputs.
How is Qwen Image Edit different from Nano Banana editing?
Both edit from references. Qwen’s edge is text-aware, structure-heavy edits — annotations, labels, layout-preserving changes — plus strong restoration. Nano Banana models are quick generalists for everyday retouches. At 4 credits per edit, Qwen is also the cheaper of the two here.
Does Qwen Image Edit change the whole image?
No — that is the point. Describe the specific change (“repair the damage”, “add taxonomy labels”, “swap the background to a beach at dusk”) and the model keeps everything else consistent: pose, lighting, identity, and composition.
How many credits does an edit cost?
Qwen Image Edit costs 4 credits per generated image on Wan27Image, the same as Qwen text-to-image. Free signup credits cover your first edits.
Can it render text inside the edited image?
Yes. The Qwen image family is known for legible multilingual text, so edits like “add a caption bar with the title at the bottom” or “label each part of the machine” come out readable — one of the main reasons to pick it over generalist editors.
Fix it, label it, restyle it
Upload a photo in the editor above, or open the full workspace for history, downloads, and a larger canvas.
Generating from scratch instead? Try Qwen text-to-image