01Introduction
An image watermarking API adds a text or image overlay to a picture through an HTTP call, so the mark is applied automatically rather than by hand in an editor. Most APIs in this category handle visible overlays, and a smaller set handles invisible watermarks that survive as machine-readable provenance signals.
Both matter in 2026, for different reasons. Visible marks protect ownership and label confidential material. Invisible marks answer a regulatory question, because the EU AI Act's transparency rules have applied since 2 August 2026 and require AI-generated content to be marked in a machine-readable format. This post compares the main visible watermarking APIs, then covers what changes when the mark has to be invisible.
02What an image watermarking API does
Watermarking through an API covers three jobs:
- Labeling sensitive documents and screenshots with words like "Confidential" or "Draft"
- Protecting ownership on photography, product imagery and user-generated content
- Marking content provenance, which now includes identifying AI-generated or edited images
A call looks like this. Source image, watermark text, and the placement and styling options you want:
curl -X POST "https://api.eu.apyhub.com/apyhub/apply-watermark/url/link" \
-H "apy-token: $APY_TOKEN" \
-H "Content-Type: application/json" \
-d '{
"image_url": "https://assets.apyhub.com/samples/sample.jpg",
"watermark_text": "Confidential",
"watermark_position": "center",
"watermark_opacity": "40",
"watermark_angle": "45"
}'{
"data": "https://example.com/watermarked-image.jpg"
}That is the Apply Watermark on Images API, which also accepts base64 and multipart uploads, and takes watermark_image_url instead of watermark_text when you want a logo.
03The comparison
| ApyHub | imgix | NeutrinoAPI | QuickChart | Cloudinary | |
|---|---|---|---|---|---|
| Text watermarks | Yes | Yes (txt family) | No | No | Yes (l_text) |
| Image overlays | Yes | Yes (mark family) | Yes | Yes | Yes |
| Input methods | URL, base64, multipart upload | Asset served through imgix | Image URL | Public image URL | Upload or fetch URL |
| Documented size limit | 100 MB on the PDF endpoint | Not documented | Not documented | 1 MB free, 10 MB paid | 10 MB free, 20 MB on Plus |
| Free tier | 5 calls a day and 3,000 atoms a month, no card | 30-day trial, 100 credits | Permanent free account, roughly 10 to 50 requests a day | Community plan, rate limited | 25 credits a month |
| Entry paid price | One subscription covering the whole catalog, see pricing | $25 a month (Starter) | $10 a month (Tier 1) | $40 a month (Professional) | $99 a month (Plus) |
| Also does | PDF, video, invisible watermarking | Full image CDN and rendering | 40+ utility APIs | Charts and QR codes | Full DAM and media CDN |
Prices and limits read from each provider's own pricing and documentation pages on 23 September 2026.
imgix
An image CDN with a deep rendering pipeline. Watermarking is the mark parameter family, with 15 parameters covering alignment, alpha, fit, padding, rotation, scale and tiling, and text overlays come from a separate txt family.
Where it shines: conditional watermarking with mark-if-min-width and mark-if-min-height, SVG watermark support added in March 2026, and text overlays extended to video in July 2026. If images already flow through a CDN, the watermark is one parameter on a URL you already serve.
Where it lags: pricing moved to credits on 30 July 2025, so any comparison written before that is wrong. Watermarking bills as a standard render at 1 credit per 500 renders, which is cheap, though the credit model also covers storage at 2 credits per GB and delivery at 1 credit per GB, so cost tracks your traffic rather than your watermarking. Input format and file size limits are not documented.
NeutrinoAPI
A utility API suite where watermarking is one endpoint among many. Parameters are image-url, watermark-url, opacity, format, position, width and height, and it accepts GIF, ICO, JPEG, PNG and TIFF.
Where it shines: the cheapest paid entry point at $10 a month, a permanent free account, and SDKs in C#, Go, Java, Node.js, PHP, Python and Rust.
Where it lags: image overlays only, with no text watermark parameter, which rules it out for "Confidential" style labeling. The free daily quota is published as a range across APIs rather than a number for this endpoint, and no file size limit is documented.
QuickChart
Primarily a chart and QR code rendering service, with a watermark endpoint at quickchart.io/watermark that composites two images.
Where it shines: a simple GET or POST interface, sensible placement controls including markRatio and per-axis positioning, and clearly published size limits.
Where it lags: image overlays only, no text watermarks, both URLs must be publicly accessible image files, and the free Community plan caps images at 1 MB with a rate limit that is enforced but not published. Watermarking is a side feature next to the charting product.
Cloudinary
A media platform where watermarking is a transformation layer rather than a dedicated endpoint: l_text or l_<image> with opacity, gravity and offsets.
Where it shines: the most complete media stack of the five, with text and image layers, drop shadows, borders, and SDKs across every major backend, frontend and mobile framework.
Where it lags: the highest entry price at $99 a month, a free tier measured in credits that are shared across transformations, storage and bandwidth, and a transformation syntax that takes longer to learn than a request body.
ApyHub
Watermarking is four endpoints in a catalog that covers the neighboring jobs too: images, PDFs, videos and an asynchronous video job API for long files.
Where it shines: text and image watermarks on one endpoint, three input methods including base64 and multipart, and control over position, opacity, angle, font size, font color, padding, alignment, line spacing and text background. The same subscription covers PDF and video watermarking, so a document pipeline and a media pipeline do not need two vendors. Every endpoint is callable over ApyHub MCP, so an AI agent can discover the watermark endpoint, read its schema and run it without a hand-written wrapper or a tool definition. Uptime is published on a public status page.
Where it lags: the free tier is 5 calls a day, which is enough to evaluate and too small for a bulk test. There is no asset CDN behind it, so if you want watermarking applied at the edge on images you already serve through a CDN, imgix or Cloudinary fit that shape better. And the image endpoint does not publish a file size limit, unlike the PDF endpoint's documented 100 MB.
04Invisible watermarking, and why 2026 changed the question
Every API above applies a visible mark. The newer requirement is the opposite: a mark a viewer cannot see, that a machine can still read.
Three things drive this.
Regulation. The EU AI Act's transparency obligations apply from 2 August 2026. Article 50 requires providers to ensure that outputs of a generative system are "marked in a machine-readable format and detectable as artificially generated or manipulated," using solutions that are "effective, interoperable, robust and reliable as far as this is technically feasible" (European Commission, transparency rules fact page, 29 July 2026).
The standard. C2PA's Technical Specification reached version 2.4 in April 2026, and its Soft Binding API exists precisely because metadata does not survive real pipelines. The spec defines a soft binding as a content identifier "embedded as an invisible watermark in the identified digital content," used to recover provenance when "the metadata containing the C2PA Manifest has been stripped" (C2PA Specification 2.4). The standards body treats invisible watermarking as the layer underneath metadata, not as an alternative to it.
The honest caveat. Invisible watermarking is not solved. A robustness benchmark published on 15 September 2026 evaluated five local invisible watermarking methods against 55 image transformations and found every method vulnerable to some class of edit, concluding that "signal distortions are often tolerated by the strongest methods, while geometric misalignment and generative local edits, such as inpainting and outpainting, can completely impair payload recovery" (Yao et al., arXiv preprint, September 2026). Google's SynthID covers images, video, audio and text with the same imperceptible approach, and its detector portal is still in limited testing.
If you need this today, ApyHub has four endpoints for it: embed an invisible image watermark, extract one, and the same pair for text hidden in images and its extraction. None of the other four providers in this comparison offer invisible watermarking.
05Choosing
Pick imgix or Cloudinary when images already run through a media CDN and watermarking is one transformation among many. You are buying a delivery pipeline, and the watermark comes with it.
Pick NeutrinoAPI or QuickChart when you need a cheap image overlay, have no text watermark requirement, and your images are already at public URLs.
Pick ApyHub when watermarking is a processing step rather than a delivery feature: text and image marks on one endpoint, the same subscription covering PDFs and video, invisible watermarking available when provenance matters, and agent access over MCP.
For the neighboring formats, see our comparisons of PDF watermarking APIs and video watermarking APIs, or the step-by-step Node.js guide if you want working code. For the wider image toolkit, see the top image editing APIs.
06Conclusion
Visible watermarking is a solved problem, and the five APIs here differ mainly in shape: a CDN parameter, a utility endpoint, or a processing call that also covers your other formats. The question worth thinking about in 2026 is whether a visible mark is all you need, given that the EU AI Act now requires machine-readable marking of AI-generated content and C2PA treats invisible watermarking as the recovery layer when metadata is stripped.
Pick the visible watermarking API that matches where your images already live, and check whether provenance is a requirement you are about to inherit.
07FAQ
What is an image watermarking API? An HTTP API that applies a text or image overlay to a picture and returns the processed file or a link to it. It replaces manual editing and runs inside your application or pipeline.
Which image watermarking API is best? It depends on where your images live. imgix and Cloudinary suit teams already serving images through a media CDN, NeutrinoAPI and QuickChart suit simple image-on-image overlays, and ApyHub suits teams that want text and image marks plus PDF and video watermarking under one subscription.
Can I add a text watermark rather than a logo? On ApyHub, imgix and Cloudinary, yes. NeutrinoAPI and QuickChart composite two images and have no text parameter, so a text mark has to be rendered as an image first.
Is there a free image watermarking API? Several have free tiers. ApyHub gives 5 calls a day and 3,000 atoms a month with no card, NeutrinoAPI offers a permanent free account with a small daily quota, QuickChart's Community plan is free with a 1 MB image limit, Cloudinary gives 25 credits a month, and imgix offers a 30-day trial.
Can I control watermark position, opacity and rotation? Yes. The ApyHub image watermark endpoint takes watermark_position, watermark_opacity, watermark_angle, watermark_font_size, watermark_font_color, watermark_text_padding and watermark_text_alignment, among others.
What is an invisible watermark, and do I need one? An invisible watermark is embedded in the image data so a person cannot see it and software can still read it. You need one if you generate or distribute AI-created images in the EU, where transparency rules requiring machine-readable marking apply from 2 August 2026, or if you want provenance to survive metadata being stripped.
Can AI agents call these APIs? ApyHub endpoints are callable over ApyHub MCP, so an agent can find the watermark endpoint, read its input schema and run it directly. The other providers are REST only, so an agent needs a wrapper you write.
How do I test a watermarking API before integrating it? Every ApyHub service page has a Playground where you can run the endpoint with your own image. Outside the browser, Voiden is an offline, Git-native API client that stores each request as a Markdown file you can commit next to your code, so the exact call that produced a result stays reproducible for the whole team.
What about watermarking PDFs or videos? Both are separate endpoints: the PDF Watermark API marks every page of a document up to 100 MB, and the video watermark API handles video, with an asynchronous job version for long files.
08About ApyHub
ApyHub is a curated API catalog and trusted operational layer for developers and AI agents. One subscription in atoms covers a catalog of over 1,500 capabilities that keeps growing, each endpoint carrying machine-readable certification for GDPR, SOC 2 and ISO 27001, and each callable by AI agents over ApyHub MCP without a hand-written wrapper. ApyHub is headquartered in Amsterdam with offices in the Netherlands, Greece and India, and serves 65,000+ monthly developer workspaces. The free tier needs no card. If you build APIs, you can publish one in under 10 minutes.
09Sources
- European Commission, Quick facts: transparency rules for AI systems, 29 July 2026.
- C2PA, Technical Specification 2.4, Soft Binding API, April 2026.
- Yao et al., What Breaks Local Watermarks? A Robustness Benchmark for Local Invisible Image Watermarking, arXiv, 15 September 2026.
- Provider pricing and documentation: imgix, NeutrinoAPI, QuickChart, Cloudinary, read 23 September 2026.
