The Apryse Fall 2026 Release is Here: Explore What's New

PDF Redaction SDKs Compared

Apryse is the only SDK offering one API across web, server, and mobile with regex in both the viewer and the API, and it has no built-in AI detection, which several competitors do. This page compares 20+ commercial and open-source tools on all six axes and says where another product fits better. Claims verified October 2026.

The Six Things That Separate Redaction Tools

Redaction tools differ on six things, and most product pages blur them:

Copied to clipboard
  1. Whether applying a redaction deletes the content from the file or only draws a box over it
  2. Whether a person can mark and review before applying
  3. Whether code can apply it unattended
  4. Whether the tool finds every match itself
  5. Whether matching supports regex and presets, and
  6. Whether a model can find sensitive data with nobody listing the words first

For the capability itself, start with PDF redaction. For why a drawn box fails, read true redaction rather than visual masking. For a buying checklist beyond tool selection, read the secure PDF redaction guide.

Term
What it means
True redaction
The content is deleted from the file itself, so nobody can copy, search, or recover it. A black box drawn on top only hides it. Turning pages into images (rasterizing) also removes it, but the text cannot be searched until OCR runs again.
User applied
A person marks text or areas in a viewer, reviews them, then applies. Good for human-in-the-loop work and manual review.
API
Code applies redaction automatically, usually on a server and often in batches. Sometimes it is a call to an external service.
Search & Redact
Find every match of a word or phrase and redact them all at once.
Pattern matching
Goes further than search, with regex and presets for common data such as email addresses or card numbers.
AI-powered
A model finds sensitive data, such as names, addresses, or account numbers, without anyone listing the words first.

Redaction Versus Sanitization, and Why the Distinction Decides the Comparison

Redaction removes visible content: the text, images, and vector regions inside the marked area. Sanitization removes hidden content from the file's internal structure: metadata, XFA data, XMP data, embedded scripts, form data, and hidden layers. A redacted file can still carry its hidden data, and a sanitized file still shows every word on the page.

Nutrient's documentation says metadata and hidden layers remain after redaction. Foxit offers Sanitize for hidden data, and Aspose and Adobe Acrobat Pro treat it as a separate step. A document headed outside your organization needs both operations. For the full treatment, see redaction versus sanitization.

Redaction is an add-on on the Apryse Web SDK, Apryse Server SDK, and Apryse Mobile SDK. Sanitization is in the Server SDK base package and runs on the Server SDK only, so a browser-only or mobile-only integration can redact but cannot sanitize. See PDF sanitization for more details.

PDF SDK Redaction Compared

Copied to clipboard

GroupDocs.Redaction is Aspose's sister brand, so those two rows are one company. Artifex MuPDF shares its engine with MuPDF.js and PyMuPDF, which appear in the open-source table.

How to read the tables: Yes means supported. Partial means limited: some content types, one platform, or only through your own code. No means not offered, or not in the vendor's public docs.

Table A: The six redaction axes

Axis
Apryse
Nutrient
Foxit PDF SDK
ComPDFKit
Syncfusion
Aspose.PDF
GroupDocs.Redaction (Aspose's sister brand)
IronPDF
Artifex MuPDF
Datalogics Adobe PDF Library
MESCIUS Document Solutions
DevExpress
True redaction
Yes. Removes text, images, and vector paths in the area (Web SDK, Server SDK, and Mobile SDK).
Yes. Removes text, images, and vectors; docs say metadata and hidden layers remain.
Yes. Removes marked text, images, and paths; Sanitize clears hidden data.
Yes. Removes text, images, and paths; a Windows sample calls this a paid add-on.
Partial. .NET removes text and graphics; the JS library cannot redact images.
Yes. Removes text and images; hidden data is a separate step.
Yes. Rasterizes by default; can keep the original format.
Partial. Removes text; unclear whether images under a region are removed.
Yes. The most documented controls, down to single characters and image pixels.
Partial. Removes text and alters images.
Yes. Removes text and images.
Yes. Clears text, images, vectors, and annotations by default.
User applied
Yes. WebViewer, Android, and iOS redaction tools.
Yes. Built-in redaction tools on Web, iOS, and Android.
Yes. Web, Android, and iOS SDK tools.
Yes. Web Redact Text, Redact Area, and Search and Redact.
Yes. JS, Blazor, and WPF viewer tools.
No. API only.
No. API only.
No. API only.
Yes. MuPDF viewer and commercial MuPDF WebViewer.
No. API only.
Yes. Viewer Redact tools (Professional license).
No. Viewer redaction is done in code.
API
Yes. Web SDK, Server SDK, and Mobile SDK.
Yes. Web, mobile, .NET, Document Engine, and cloud API.
Yes. Redaction add-on for C++, C, Java, Python, C#, Go, Node.js, and Web.
Yes. Web, mobile, Windows, and server (Java, .NET, C++, PHP, Python).
Yes. .NET library and a JS add-on.
Yes. .NET, Java, and Python, one area at a time.
Yes. .NET, Java, and Python; PDF, Office, and image files.
Yes. .NET.
Yes. C, Java, JavaScript, and Python.
Yes. C/C++, .NET, and Java.
Yes. .NET and JS.
Yes. .NET (v25.1 and later) and Java.
Search & Redact
Yes. In the UI and the API.
Partial. API only, no search-and-redact UI.
Yes. In the Web SDK UI and the API.
Partial. In the Web UI; the API needs search, mark, then apply.
Partial. Code samples only.
Partial. Your code finds each match, then redacts it.
Yes. Exact phrase or regex across the document.
Yes. One call redacts a phrase on every page.
Partial. Search returns positions; you mark and apply.
Partial. Your code finds each match, then redacts it.
Partial. Your code finds each match, then marks it.
Partial. Your code loops over matches.
Pattern matching
Yes. Regex, wildcards, and 3 presets (email, card, phone).
Yes. Regex and 13 Web presets (card, email, SSN, VIN, and more).
Partial. Regex (SDK 10.1 and later); no presets in the SDK.
No. Regex and presets not documented.
Partial. Case and whole word; regex in your own code.
Partial. Regex; no presets.
Partial. Regex; no presets.
Partial. Case and whole word; regex in your own code.
Partial. Experimental regex and built-in OCR; no presets.
Yes. Regex; its pdfRest API adds 12 presets.
Partial. Regex, case, and whole word; no presets.
Partial. Case and whole word only.
AI-powered
No. No built-in AI detection.
Yes. Cloud API and AI Assistant.
No. Foxit's AI Smart Redact is not in the SDK.
Partial. Offered on its AI platform; categories not documented.
Yes. Smart Redaction in Blazor, using Azure OpenAI.
No.
No.
No.
No.
No.
No.
No.

Table A2: Hidden data, licensing, and better fit

Product
Hidden data and sanitization
Licensing
Better fit when
Apryse
Server SDK: Sanitization removes metadata, XFA, XMP, scripts, form data, and hidden layers, and is in the base package. Web SDK and Mobile SDK: not available.
Commercial. Redaction is an add-on on the Web SDK, Server SDK, and Mobile SDK.
You need one API across web, server, and mobile, with regex in the viewer and the API, and documents that stay in your environment.
Nutrient
Not removed by redaction: docs say metadata and hidden layers remain.
Commercial. Redaction is an add-on.
You want AI detection or 13 Web presets, or you build mobile-first products.
Foxit PDF SDK
Sanitize clears hidden data.
Commercial. Redaction is an add-on.
You need desktop and server signing at PAdES B-LTA or live dynamic XFA form filling alongside redaction.
ComPDFKit
Not documented in the source.
Commercial, from $1,300. A Windows sample calls redaction a paid add-on.
You want a low entry price and a community license program.
Syncfusion
Not documented in the source.
Commercial. Free for small companies.
You want Smart Redaction in Blazor, already standardize on Syncfusion UI, or need 24x7 enterprise support.
Aspose.PDF
A separate step.
Commercial, from $1,199.
You want self-serve pricing, niche formats such as Visio and OneNote, or deep programmatic DOCX editing.
GroupDocs.Redaction
Not documented in the source.
Not documented in the source.
You redact Office and image files as well as PDFs through one API.
IronPDF
Not documented in the source.
Commercial, $999 to $5,999 per developer.
You work in .NET and want one call to redact a phrase on every page.
Artifex MuPDF
Not documented in the source.
Commercial license.
You need character-level or image-pixel control over what is removed.
Datalogics Adobe PDF Library
Not documented in the source.
Commercial, by quote.
You want regex plus 12 presets through its pdfRest API.
MESCIUS Document Solutions
Not documented in the source.
Commercial, from $999 per developer.
You work in .NET or JS and want viewer redact tools.
DevExpress
Not documented in the source.
Commercial subscription.
You already use DevExpress on .NET and want annotations cleared by default.

On the user-applied axis, Apryse offers redaction tools in WebViewer, Android, and iOS, so a person marks, reviews, and then applies. enterprise-grade redaction covers review at scale. On the API axis, one API spans the Web SDK, Server SDK, and Mobile SDK, with language pages for redaction in C# and redaction in Java. Search and redact is where the closest competitor differs: Nutrient is API only, with no search-and-redact UI. Implementation details live in the docs: the Web SDK redaction guide and the Server SDK redaction guide.

Cloud AI Redaction Services Compared

Copied to clipboard

Three cloud AI services detect sensitive data automatically. All three process documents on the vendor's servers, so none keeps documents in your environment. Two of them do not redact PDFs directly.

Service (licensing)
True redaction
User applied
API
Search & Redact
Pattern matching
AI-powered
Azure AI Language, Document PII (pay as you go)
Partial. Returns a redacted PDF but does not say how the text is removed.
No. API only.
Yes. PDF, DOCX, and TXT, async (GA May 2026).
Yes. Redacts every detected entity automatically.
Partial. Custom regex only in the Text PII container.
Yes. 79 languages, OCR for scans.
AWS Comprehend (pay as you go)
No. Plain text only; PDF only in samples that turn pages into images.
No.
Partial. Redacts text in English and Spanish.
Partial. Automatic, text only.
No. Regex not found.
Yes. 36 entity types.
Google Sensitive Data Protection (usage-based)
No. Images and text only; docs say PDF and DOCX are not supported.
No.
Partial. Image and text APIs, no PDF.
Partial. Automatic, images and text only.
Yes. Custom regex and dictionaries; OCR on images.
Yes. Built-in detectors.

Adobe: Acrobat Pro Versus the Developer API

Copied to clipboard

Adobe Acrobat Pro is a desktop app for people redacting by hand, not a developer tool, so it sits outside the SDK tables. Adobe's developer API, PDF Services, has no redaction feature, and an AI Auto-Redact API is at beta sign-up.

Product (licensing)
True redaction
User applied
API
Search & Redact
Pattern matching
AI-powered
Adobe Acrobat Pro (commercial desktop app)
Yes. Removes visible text and graphics; Sanitize Document clears hidden data as a separate step.
Yes. Redact tools and Find text and redact.
No. Scripting works only inside Acrobat; an AI Auto-Redact API is in beta sign-up.
Yes. Words, phrases, lists, or patterns, plus "Redact all" in the AI chat.
Partial. Presets (phone, card, email, SSN, dates); no regex.
Yes. "Redact text with chat" finds content by meaning (paid, English only).

Open-Source Redaction Compared

Copied to clipboard

Four open-source projects ship working redaction. Only Stirling-PDF adds AI, and only in its proprietary parts. Only Stirling-PDF searches and redacts across many files. MuPDF.js and PyMuPDF share an engine with Artifex MuPDF and are AGPL or commercial, which carries obligations if you ship them in a closed-source product.

Project (license)
True redaction
User applied
API
Search & Redact
Pattern matching
AI-powered
EmbedPDF (MIT for v2; Apache-2.0 for the v3 prerelease)
Yes. Removes text, images, drawings, and overlapping annotations.
Yes. Browser viewer redaction tools.
Yes. JavaScript, in the browser and Node.
Partial. Only in the v3 prerelease.
Partial. Regex in v3; case and whole word in v2.
No.
Stirling-PDF (MIT core; some parts proprietary)
Partial. True removal only in auto-redact; manual redaction draws boxes on top.
Yes. Web app, manual redaction in the browser.
Yes. REST API with batch files.
Yes. Word lists across multiple files.
Partial. Regex, case, whole word; PII presets are proprietary.
Partial. Only in the proprietary parts.
MuPDF.js (AGPL or commercial)
Yes. Same engine as MuPDF; needs a full save, since an incremental save keeps the original.
No. API only.
Yes. JavaScript, in the browser and Node.
Partial. Search returns positions; you mark and apply.
Partial. Experimental regex.
No.
PyMuPDF (AGPL or commercial)
Yes. Its scrub() function also clears metadata and hidden data.
No. API only.
Yes. Python.
Partial. Your code searches, then marks each match.
Partial. Literal text only; Tesseract OCR.
No.

Microsoft Presidio:

Detection Without Redaction

Presidio finds personal data in text but cannot change a PDF, so it needs a PDF library to do the redaction.

It is a free, open-source (MIT) toolkit that detects names, card numbers, locations, and Social Security Numbers, and returns each match with its position and a confidence score. Its own README warns it will not find everything.

Microsoft built it, and it is now community-owned under Data Privacy Stack: see the Presidio repository and the project transition notice.

The pairing with a PDF library runs in five steps:

Copied to clipboard
  1. The PDF library extracts the page text, with the position of each character.
  2. Presidio scans that text and returns each match with its character positions.
  3. Those positions are mapped back to boxes on the page.
  4. The PDF library applies true redaction to those boxes and saves the whole file.
  5. A cleanup tool clears metadata and hidden data, for example pikepdf or PyMuPDF's scrub().

Scanned pages need OCR first, and the redaction must remove both the image pixels and the OCR text. See redacting scanned documents for that workflow, and redaction in Python for the Apryse route in Python.

Likeliest pairings

Library
Why it fits
Watch out for
PyMuPDF (Python)
Gives each character's position, so one match can be redacted without touching the rest. scrub() clears metadata.
AGPL or a commercial license.
MuPDF.js (browser and Node)
Same engine as PyMuPDF, in JavaScript.
AGPL; needs a full save, not an incremental one.
EmbedPDF v3 (browser)
Presidio's positions map to text ranges on the page; Apache-2.0.
Still a prerelease.
Stirling-PDF (REST API)
Simplest setup: send Presidio's matches to its auto-redact endpoint.
Matches by text, so every occurrence of a word is redacted, which can over-redact.

Weaker fits: pdf-redactor (text and metadata only, unmaintained since 2019) and LibreOffice (driven through a dialog, image-only output).

Tools With No Redaction Feature

Some products have no redaction feature:

Sanity Image

Commercial and cloud:

Telerik, TX Text Control, SimplePDF, Adobe PDF Services and Embed APIs (Auto-Redact is a beta sign-up), Reducto, and LlamaParse.

Open source:

PDF.js, upstream PDFium, OpenPDF, PdfPig, Docling, qpdf, and generation libraries such as pdf-lib, jsPDF, and QuestPDF.

Deployment Control and Where Documents Are Processed

The cloud AI services process documents on the vendor's servers. If a document cannot leave your perimeter, that architecture is ruled out. Apryse runs entirely in your environment. The same APIs work in a VPC, on-premises, or air-gapped, and documents never leave your perimeter. In the Web SDK client only deployment, redaction runs in the browser through WebAssembly (WASM) with no server involved. Air-gapped operation needs a license configured for offline use.

The Server SDK has no built-in work queue, job scheduler, or batch orchestration, so your pipeline owns the loop when you are redacting documents in bulk. Confirm deployment for any SDK vendor on your shortlist.

Licensing Models

Copied to clipboard

Add-on licensing is the norm in this field. Redaction is a separately licensed add-on for Apryse, Nutrient, and Foxit, and a Windows ComPDFKit sample calls it a paid add-on.

Apryse licenses Redaction as an add-on on the Web SDK, Server SDK, and Mobile SDK. Sanitization is in the Server SDK base package. The cloud services bill per use, and several open-source options are AGPL. Apryse pricing starts with a sales conversation, while several competitors publish prices or offer free tiers. Web SDK and Server SDK trial keys include every add-on.

When Another Tool Is the Better Fit

Copied to clipboard

Apryse has no built-in AI detection, and that is the first reason to choose something else. Several tools score better on other axes too.

AI detection:
Copied to clipboard

Nutrient (local in .NET, cloud API, and AI Assistant) and Syncfusion (Smart Redaction in Blazor, using Azure OpenAI) detect sensitive data with no word list, as do Azure, AWS, and Google. ComPDFKit and Stirling-PDF offer it in part.

Presets:
Copied to clipboard

Nutrient offers 13 Web presets against 3 in Apryse, and Datalogics' pdfRest API adds 12.

Fine-grained control:
Copied to clipboard

Artifex MuPDF has the most documented controls in the field, down to single characters and image pixels.

Office and image files:
Copied to clipboard

GroupDocs.Redaction handles PDF, Office, and image files through one API.

Open source:
Copied to clipboard

Stirling-PDF searches and redacts across many files.

Beyond redaction:
Copied to clipboard

Apryse has no live co-editing of documents, no built-in PDF/UA validator, no public VPAT, and no 24/7 support tier. Pricing starts with a sales conversation rather than a published price or a free tier.

When Apryse Is the Better Fit

Copied to clipboard
You need one API across surfaces.
Copied to clipboard

The Web SDK, Server SDK, and Mobile SDK share one API, and regex works in both the viewer and the API.

You need search and redact in the UI and the API.
Copied to clipboard

Nutrient, the closest match, offers it through the API only.

You need hidden-data removal in the same SDK.
Copied to clipboard

Sanitization is in the Server SDK base package.

Your documents cannot leave your environment.
Copied to clipboard

Redaction and sanitization both run inside your perimeter, in a VPC, on-premises, or air-gapped.

Your security team wants certification evidence.
Copied to clipboard

Apryse has held ISO 27001 certification since 2018, now ISO/IEC 27001:2022, and completes a SOC 2 Type II audit yearly.

For industry workflows:

Compare Apryse With a Specific Vendor

Copied to clipboard

Company-level pages go deeper on one vendor than this hub does:

For the same comparison on document editing, see the DOCX Editor comparison hub.

Frequently Asked Questions

It depends which of six axes matters most: true removal, mark-and-review before applying, unattended API use, search and redact, pattern matching, and AI detection. Apryse is the only SDK with one API across web, server, and mobile plus regex in both the viewer and the API. Nutrient is the closest match and leads on presets and AI detection.

True redaction deletes the content from the file itself, so it cannot be copied, searched, or recovered. A black box drawn on top only hides it. Turning pages into images also removes the text, but the document is unsearchable until OCR runs again.

Redaction removes visible page content: text, images, and vector regions inside the marked area. Sanitization removes hidden content: metadata, XFA and XMP data, embedded scripts, form data, and hidden layers. In the Apryse SDK, sanitization is in the Server SDK base package and is Server SDK only.

No. Apryse has no built-in AI detection: redaction is driven by what you mark, by search, or by regex, wildcard, and preset patterns. Nutrient, Syncfusion, and the cloud AI services offer model-based detection, but the cloud services process your documents on their servers.

Four open-source projects ship working redaction: EmbedPDF, Stirling-PDF, MuPDF.js, and PyMuPDF. Stirling-PDF is the only one with search-and-redact across many files. MuPDF.js and PyMuPDF are AGPL or commercial, and parts of Stirling-PDF are proprietary.

Not on its own. Presidio finds personal data in text but cannot modify a PDF, so it must be paired with a PDF library that applies true redaction. It is now a community-owned project rather than a Microsoft one.

Not currently. Adobe PDF Services and the Embed APIs have no redaction feature, and an AI Auto-Redact API is at beta sign-up. Adobe Acrobat Pro does redact, but it is a desktop application for redacting by hand.

With Apryse, yes: the same APIs run in your own environment, in a VPC, on-premises, or air-gapped, and documents never leave your perimeter. The cloud AI services process documents on the vendor's servers.

Add-on licensing is the norm: redaction is a separately licensed add-on for Apryse, Nutrient, and Foxit, and a paid add-on in at least one ComPDFKit configuration. The cloud services bill per use, and several open-source options are AGPL.

When model-based detection with no word list is the requirement, several competitors have it and Apryse does not. Also when you need live co-editing, a built-in PDF/UA validator, a public VPAT, a 24/7 support tier, or a published price and free tier. The better-fit column names the strongest alternative per axis.

Test Redaction on Your Own Documents

Looking to test redaction on your own documents? Try it instantly in our showcase, choose the free trial that best fits your deployment, or connect with a solutions engineer to discuss your needs.

Sanity Image

© Apryse Software Inc.