Image Watermark Remover — Batch Remove Logos & Stamps

Image Watermark Remover takes up to 20 public JPG, PNG, or WebP image URLs per run, detects text watermarks, logos, date stamps, and caption strips, fills them in with inpainting, and returns a download link for each cleaned image plus a before/after preview. It is built for developers and teams who need to clean their own images in bulk through an API, and it costs USD 50 per 1,000 cleaned images (USD 0.05 per image) on the FREE tier, plus a USD 0.005 start fee per run.

How do I remove watermarks from images in bulk?

Pass a list of image URLs and the Actor processes them in parallel (two at a time by default, up to four with maxConcurrency), saves each cleaned file in the run's key-value store, and writes one dataset row per image with its download URL:

{
  "imageUrls": [
    "https://example.com/products/mug-front.jpg",
    "https://example.com/products/mug-side.jpg"
  ],
  "detectionMode": "corner_bottom_right",
  "inpaintMethod": "hybrid",
  "maskDilation": 5,
  "outputFormat": "png",
  "createComparison": true
}

For more than 20 images, split them across several runs or call the Actor from a loop in your own code. Images must be publicly reachable over HTTP(S); private network addresses, localhost, URLs with credentials, and animated GIFs are rejected.

Which detection mode should I use?

Use the mode that matches where the watermark sits; it is faster and safer than automatic detection.

• corner_bottom_right, corner_bottom_left, corner_top_right, corner_top_left, all_corners: logos, photographer credits, camera date stamps, and channel badges in corners.
• bottom_banner: caption or ticker strips along the bottom edge.
• center: large preview marks across the middle.
• manual_boxes: exact boxes in pixels ({x, y, width, height}) or percentages ({xPercent, yPercent, widthPercent, heightPercent}) when you know the position, for example a batch of images with the same stamp.
• auto (default when an API call omits detectionMode; the Console form starts on corner_bottom_right): scans the four corners, the bottom strip, and the center for the strongest text- or logo-like strokes and fills them.
• full_image: applies the same scan to the whole image.

Detection uses classical computer vision (edge, contrast, and stroke analysis), not a generative AI model, and it cannot tell whether a watermark exists. In auto and full_image modes it always marks the most contrasting strokes, so on a photo without a watermark it can still touch real detail. In a single-region mode, if nothing stands out, the whole region is filled. Check the before/after preview, and use maskDilation 7 or higher for watermarks with soft shadows or glow.

Three fill methods are available: hybrid (default; edge-preserving fill plus grain matching to avoid smooth smudges), telea, and ns.

What does the image watermark remover return?

One dataset row per cleaned image:

Field	What it tells you
sourceUrl	The input image URL
cleanedImageUrl, cleanedImageKey	Download URL and key of the cleaned image in the key-value store
comparisonImageUrl, comparisonImageKey	Side-by-side before/after JPEG, or null when disabled
originalWidth, originalHeight, outputWidth, outputHeight	Dimensions before and after the optional downscale
inputBytes, outputBytes	File sizes in bytes
watermarkDetected, detectedRegionsCount	Whether any region was filled, and how many
inpaintedPixels, inpaintedPercentage	How much of the image was reconstructed
detectionModeUsed, inpaintMethodUsed, outputFormat	The settings applied
processingTimeMs, sha256	Processing time and a hash of the cleaned file
base64Data	A data:image/... string when includeBase64 is on, otherwise null

An OUTPUT record summarizes the run: status, requested, delivered, failed, warnings, and result links.

This row is from the Store prefill run (2026-10-27): a beach photo with an orange camera date stamp in the bottom-right corner, cleaned with corner_bottom_right. The stamp was filled completely and nothing outside the corner was touched (8 regions, 0.67% of the pixels):

{
  "sourceUrl": "https://res.cloudinary.com/demo/image/upload/w_1200/l_text:Arial_36_bold:2024-06-18,co_rgb:F57C00,g_south_east,x_28,y_22/samples/landscapes/beach-boat.jpg",
  "cleanedImageUrl": "https://api.apify.com/v2/key-value-stores/0zOFLREzJmSUeMVfL/records/cleaned-0b2fcc38c368439f.jpg",
  "comparisonImageUrl": "https://api.apify.com/v2/key-value-stores/0zOFLREzJmSUeMVfL/records/comparison-0b2fcc38c368439f.jpg",
  "cleanedImageKey": "cleaned-0b2fcc38c368439f.jpg",
  "comparisonImageKey": "comparison-0b2fcc38c368439f.jpg",
  "originalWidth": 1200,
  "originalHeight": 900,
  "outputWidth": 1200,
  "outputHeight": 900,
  "inputBytes": 173058,
  "outputBytes": 212680,
  "watermarkDetected": true,
  "detectedRegionsCount": 8,
  "inpaintedPixels": 7276,
  "inpaintedPercentage": 0.674,
  "detectionModeUsed": "corner_bottom_right",
  "inpaintMethodUsed": "hybrid",
  "outputFormat": "jpeg",
  "processingTimeMs": 336,
  "sha256": "0b2fcc38c368439f08d2d84650d2df30e94bd8f8fc96d15ebf27c2ecf060b6d1",
  "base64Data": null
}

How much does it cost to remove watermarks from images?

Pay-per-event prices, active since October 6, 2026, with no change scheduled. Apify platform usage is included.

Event	FREE	BRONZE	SILVER	GOLD	PLATINUM	DIAMOND
apify-default-dataset-item: one cleaned image delivered (primary)	$0.05	$0.04875	$0.0475	$0.04	$0.04	$0.04
apify-actor-start: run start	$0.005	$0.005	$0.005	$0.005	$0.005	$0.005

The start fee is charged once per started GB of allocated memory; the default 1,024 MB run pays one start event.

• One image, FREE: $0.005 + 1 × $0.05 = $0.055.
• A full batch of 20 images, GOLD: $0.005 + 20 × $0.04 = $0.805.

1,000 images take 50 runs of 20, which is 50 × ($0.005 + 20 × $0.05) = $50.25 on FREE. Each image that is saved to the dataset is charged once, including images where nothing was filled; images that fail to download or decode are listed as warnings in OUTPUT and not charged. A run with both saveToKeyValueStore and includeBase64 turned off is rejected before any image is processed, so only the start fee applies. Duplicate URLs in one run are processed and charged once, and when your maximum charge covers fewer images than you sent, the extra images are skipped, listed in OUTPUT.warnings, and not charged.

When should I use it, and when not?

Use it to clean your own images, or images you are licensed to edit, in bulk from code or an automation: product photos with an outdated supplier or brand stamp, your own photos with a burned-in camera date, or screenshots with a logo you are allowed to remove.

Do not use it to strip copyright notices from other people's photos, stock previews, or listings you do not own. It does not process video, animated GIFs, SVG, or PDF, and it cannot rebuild a face or intricate detail fully hidden under a large watermark. For restoring old or damaged photos use Vintage Photo Restorer, and for reading text in images use Image OCR Scraper.

Use it from an AI assistant

Image Watermark Remover works from Claude, ChatGPT, Cursor, or any MCP client through Apify's MCP server. Add it as a single tool with https://mcp.apify.com/?tools=muhammadafzal/ai-image-watermark-remover, then ask, for example: "Remove the date stamp in the bottom-right corner of these three product photos and give me PNG download links."

How do I run it from the API?

JavaScript (apify-client):

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('muhammadafzal/ai-image-watermark-remover').call({
  imageUrls: ['https://example.com/products/mug-front.jpg'],
  detectionMode: 'corner_bottom_right',
  createComparison: true,
});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items[0]?.cleanedImageUrl);

Python (apify-client 3.x):

from apify_client import ApifyClient

client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("muhammadafzal/ai-image-watermark-remover").call(run_input={
    "imageUrls": ["https://example.com/products/mug-front.jpg"],
    "detectionMode": "corner_bottom_right",
    "outputFormat": "png",
})
for item in client.dataset(run.default_dataset_id).iterate_items():
    print(item["cleanedImageUrl"], item["comparisonImageUrl"])

For n8n, Make, or Zapier flows, set includeBase64: true to get the cleaned image inline in each row.

Limits and reliability

• 1–20 images per run, up to 25 MB and 36 megapixels each, at least 32 × 32 pixels. Downloads time out after 30 seconds.
• maxDimension (512–6,000, default 4,096) downscales larger images before processing; smaller images are never upscaled.
• Results depend on the background: fills work best over sky, walls, or even texture, and marks over faces, fine text, or busy patterns can leave visible traces.
• An image that fails to download or decode does not stop the run; it is listed in OUTPUT.warnings. If no image is delivered at all, the run ends as failed and only the start fee is charged.

FAQ

Is it legal to remove a watermark?

It depends on who owns the image. Removing a watermark from your own image, or one you are licensed to edit, is generally fine; removing copyright information from someone else's work can break copyright law and site terms. You are responsible for having the right to edit every image you submit.

Is this an AI watermark remover?

It is an automatic watermark remover that uses classical computer-vision detection and inpainting, not a generative AI model. That keeps it fast and predictable, but it does not invent new detail.

Can it remove a logo from an image?

Yes, when the logo sits in a corner, the bottom strip, or the center, or when you give its position with manual_boxes.

Will it change parts of my image without a watermark?

In auto and full_image modes it can, because it always fills the strongest strokes it finds. Use a corner mode or manual_boxes to limit changes to one area.

Does it remove watermarks from video?

No. It processes still JPG, PNG, and WebP images only.

Where do I download the cleaned images?

From cleanedImageUrl in each dataset row, or from the run's key-value store under keys starting with cleaned-.

Related Actors

• Vintage Photo Restorer: restore old, faded, or damaged photos.
• Image OCR Scraper: extract text from images.

Legal and responsible use

Submit only images you own or are licensed to modify. Do not use this Actor to remove watermarks, signatures, or other copyright information from third-party content, and respect the terms of the sites that host your images.

Last verified: 2026-10-27