Remove thewatermarks, text, corner badges, coupon bars, decorative bordersfrom the image, and restore the obscured picture content.
The most common use ispreparing clean input for other skills: an image with text handed to fission-pattern or clothing-grass-planting, that text will be copied into every derived image.
Generation result examples
| Input: image with watermark text |
|---|
source-watermarked.jpg — Home textile promotion main image: red-gold border + title + event dates + price bubble + discount coupon bar + multi-color available corner badge |
Command actually executed:
dlazy gpt-image-2 \
--prompt 'Remove every piece of text, price tag, badge, ribbon and decorative promotional frame from this image, and reconstruct what was behind them. Delete the red-and-gold border frame, the large headline, the date line, the price bubble with the number, the coupon banner and the bottom-right colour label. Keep the photograph itself completely unchanged: same bedroom, same bed, same taupe and cream bedding set, same table lamp, same wall art, same perspective, same colour grading and same resolution. Inpaint the cleared areas so the room continues naturally — walls, headboard, floor and bedding must look seamless, with no ghosting, blur patches or leftover letter shapes. Clean product photo, absolutely no text.' \
--images docs/remove-watermark/source-watermarked.jpg \
--size 1024x1024 --quality medium --imageFormat jpeg \
--save docs/remove-watermark/example-output.jpg
Output
example-output.jpg — 1024×1024. The red-gold border, title, event dates, price bubble, discount coupon bar and corner badge are all removed; the wall, headboard, floor and bedding set connect naturally, with no glyph ghosting; the bedroom, bedding color scheme, table lamp and wall decorative painting are preserved.
Note: comparing with the original image, you can see the room perspective has been slightly reconstructed — this is the normal result of reconstruction rather than pixel-level retouching. To strictly preserve the original pixels, use professional retouching tools.
1. Capability boundaries
| Can remove | Description |
|---|---|
| Semi-transparent watermarks | Platform watermarks, store watermarks, anti-theft watermarks |
| Promotional text | Titles, selling-point copy, event dates |
| Price corner badges | Circular/irregular price bubbles, discount corner badges |
| Coupon bars | Bottom horizontal bars, color-block bars |
| Decorative borders | Borders, patterns, ribbons |
| Stickers | Emojis, arrows, indicator graphics |
Important limitation: This skill isreconstructionrather than pixel-level retouching. The content of the covered areas isinferredby the model, and the rest of the image may also be slightly redrawn (subtle changes in perspective and texture). To strictly preserve the original pixels, please use professional retouching tools.
Not for: not for erasing others' copyright watermarks and then misusing the image; not for erasing brand logos and then passing it off as one's own product; not for removing legally required markings (certification marks, warning statements, ingredient labels).
2. Input material rules
Before generating, self-check these hard constraints:
- Size:20KB ~ 15MB
- Resolution:greater than 400×400
- Format:jpg / jpeg / png / webp
Input recommendations
| Practice | Description |
|---|---|
| ✅ Simple background in the text area | Text on solid-color walls or solid-color countertops is easiest to restore |
| ✅ High resolution | Reconstruction quality improves with the original image's resolution |
| ✅ List the elements to delete one by one | Just saying "remove watermark" will miss borders and corner badges |
| ⚠️ Text pressed over the product subject | The parts of the product obscured by text are inferred and need manual verification |
| ❌ Text covers more than half the area | The reconstruction is basically a new image; it is recommended to redo the image |
3. Name items one by one + fidelity lock
The prompt for de-printing is two lists:deletion listandretention list。
【删除清单】Delete: the [边框描述], the [标题文字], the [日期行],
the [价格气泡], the [优惠券条], the [角标].
【保留清单】Keep the photograph itself completely unchanged: same [主体],
same [环境元素逐项], same perspective, same colour grading, same resolution.
【重建要求】Inpaint the cleared areas so the [场景] continues naturally —
[被遮挡的结构] must look seamless, with no ghosting, blur patches or leftover letter shapes.
【硬约束】absolutely no text anywhere in the output.
no ghosting, blur patches or leftover letter shapes This sentence is crucial—if not written, blurry glyph outlines are often left behind.
The retention list must be itemized; if you only write "keep the image unchanged," the model will reconstruct the perspective.
4. Tool invocation
This skill uses dLazy's gpt-image-2(image editing model; the essence of de-printing is "region removal + semantic reconstruction + keeping the rest unchanged," and it needs to understand the image structure to reasonably fill in the occluded parts).
Invocation method
Two equivalent ways to write it; choose one. The unified entry point automatically selects the backend, retries on failure, creates directories and writes to disk, and estimates cost:
# A. 统一入口（推荐）：可切任意后端，加 --dry-run 不计费空跑
node scripts/gen.mjs --task remove-watermark \
--prompt '<见下方 Prompt 模板>' \
--images <按下表顺序> \
--save output/remove-watermark-<sku>.jpg
# B. 直接用 dLazy CLI（不想引入 Node 依赖时，效果等价）
dlazy gpt-image-2 --prompt '...' --images ... --save output/remove-watermark.jpg
Parameter conventions (fixed usage for this skill)
| Parameter | Value | Reason |
|---|---|---|
--images |
[原图] |
Single-image input |
--size |
Same aspect ratio as the original image | De-printing should not change the composition ratio |
--quality |
high(large reconstruction area) / medium(only a small watermark) |
Reconstruction quality is positively correlated with the tier |
--imageFormat |
jpeg |
General format |
--batch |
2 ~ 3 |
Reconstructed content is random; pick the most reasonable one |
--save |
docs/remove-watermark/output-<sku>-clean.jpg |
Archive the cleaned version separately |
Command Examples
# basic call: 去掉水印和文字
dlazy gpt-image-2 \
--prompt 'Remove every piece of text, watermark, badge and decorative frame from this image and reconstruct what was behind them. Keep the photograph itself unchanged: same subject, same environment, same perspective, same colour grading. Inpaint the cleared areas so the scene continues naturally, with no ghosting, blur patches or leftover letter shapes. Absolutely no text.' \
--images docs/remove-watermark/source-watermarked.jpg \
--size 1024x1024 --quality medium
# complex call: 逐项点名 + 保真锁定 + 出 3 张挑最合理的重建
dlazy gpt-image-2 \
--prompt 'Remove every piece of text, price tag, badge, ribbon and decorative promotional frame from this image, and reconstruct what was behind them. Delete: the red-and-gold border frame, the large headline, the date line, the price bubble with the number, the coupon banner and the bottom-right colour label. Keep the photograph itself completely unchanged: same bedroom, same bed, same bedding set, same table lamp, same wall art, same perspective, same colour grading and same resolution. Inpaint the cleared areas so the room continues naturally — walls, headboard, floor and bedding must look seamless, with no ghosting, blur patches or leftover letter shapes. Clean product photo, absolutely no text.' \
--images docs/remove-watermark/source-watermarked.jpg \
--size 1024x1024 --quality high --imageFormat jpeg \
--batch 3 --save docs/remove-watermark/output-bedding-clean.jpg
# 批量洗图：作为其他技能的前置步骤
for f in docs/remove-watermark/raw/*.jpg; do
dlazy gpt-image-2 \
--prompt 'Remove all watermarks, text, badges and decorative frames; reconstruct the covered areas seamlessly. Keep the photograph unchanged in subject, perspective and colour grading. No ghosting, no blur patches, no leftover letter shapes. Absolutely no text.' \
--images "$f" --size 1024x1024 --quality high --imageFormat jpeg \
--save "docs/remove-watermark/clean/$(basename $f)"
done
# 先估价不真跑
dlazy gpt-image-2 --dry-run --prompt '...' --images a.jpg --size 1024x1024 --quality high
Further reading
| What to look up | Where to go |
|---|---|
| Authentication, multi-backend configuration, output structure, error codes | references/provider-cli.md |
gpt-image-2 all available parameters of |
references/model-flags.md |
| All options of the unified entry point | node scripts/gen.mjs --help |
5. Prompt template
Remove every piece of text, watermark, badge, ribbon and decorative promotional frame
from this image, and reconstruct what was behind them.
Delete: the [边框], the [标题文字], the [日期/说明行], the [价格气泡],
the [优惠券条], the [角标/贴纸].
Keep the photograph itself completely unchanged: same [主体], same [环境元素逐项],
same perspective, same colour grading and same resolution.
Inpaint the cleared areas so the [场景] continues naturally — [被遮挡的结构逐项]
must look seamless, with no ghosting, blur patches or leftover letter shapes.
Clean product photo, absolutely no text.
Correction sentences appended by issue
| Issue | Append to the end of the prompt |
|---|---|
| Blurry glyphs left behind | No residual glyph shapes, no smudged rectangles where text used to be. |
| Only part was deleted | Add the missed items one by one to the deletion list, and append Every one of the listed elements must be gone. |
| The image was reconstructed | Do not re-compose the photograph: keep the same camera angle, the same object positions and the same wall/floor lines. |
| The texture in the reconstructed area does not connect | Continue the existing [墙面/木纹/织物] texture across the cleared area with matching scale and direction. |
| The part of the product occluded by text is wrong | Reconstruct the occluded part of the product from symmetry and the visible portion; do not invent new features. |
6. Execution process
- List the deletion list: nameeverytext/graphic element in the image (borders, titles, dates, prices, coupon bars, corner badges, stickers).
- List the retention list: subject + environmental elements item by item + perspective + color tone + resolution.
- Write the reconstruction requirements: specify which structures need to connect, and add the sentence "no glyph residue."
--quality high(when the reconstruction area is large) →--batch 2~3pick the most reasonable one.- Quality check: whether the text is completely removed, whether there are glyph afterimages, whether the perspective has been changed, and whether the occluded parts of the product are reasonable.
- Explain at delivery: the reconstructed area is inferred content; if pixel-level fidelity is required, use professional retouching.
7. Common issues
| Symptom | Cause | Handling |
|---|---|---|
| Blurry glyph outlines left behind | No no-residue constraint was written | Append no residual glyph shapes sentence |
| Only part of the elements were deleted | The deletion list is incomplete | Fill in item by item + Every one of the listed elements must be gone. |
| Perspective/composition changed | Composition not locked | Add a sentence prohibiting composition reconstruction; this is an inherent limitation of this skill |
| Texture discontinuity in the reconstructed area | Texture continuity not required | Add a texture continuity sentence specifying the texture type and direction |
| The product obscured by text was filled in incorrectly | That part was inferred | Add a symmetry inference sentence; manually verify key areas |
| Text covers more than half | Beyond capability boundary | Redo the image, don't wash it |
Tips
Visit https://dlazy.com for more information.