Pdftools Release Notes

Fixed

• Before this update, the Toolbox add-on for Java extracted its native libraries to the temporary directory, where they accumulated over time. As a consequence, disk usage in the temporary directory grew with repeated use. With this release, the Toolbox add-on stores the extracted native libraries in a persistent per-user cache; if the cache location is unusable, the SDK falls back to extracting into the temporary directory. As a result, the libraries no longer accumulate, and initialization is faster after the first run.
• Before this update, the Toolbox add-on ignored color key masks (transparency) when rendering indexed images with fewer than 8 bits per component. As a consequence, masked pixels rendered opaque instead of transparent. With this release, the Toolbox add-on applies color key masks to these images, and masked pixels render transparent as expected.

Added

• Indexed (palette-based) PNG output

Render PDF pages to smaller PNG files, suitable for thumbnails, web previews, and high-volume archiving. With this release, Pdftools SDK supports the PngColorSpace.Indexed color space for PNG output. Instead of a full color value, each pixel stores a short reference into a palette of up to 256 colors. Typical document pages, such as text, tables, and line art, contain only a few distinct colors. They keep their visual quality at a much smaller file size. Choose the palette size using the PngImageOptions.BitsPerPixel property. Indexed output is always opaque. For technical details, such as valid values, defaults, and error behavior, refer to the PngImageOptions API reference for your language:

• C
• Java
• .NET
• Python

Fixed

• Before this update, the Pdftools SDK for Java extracted its native libraries to the temporary directory, where they accumulated over time. As a consequence, disk usage in the temporary directory grew with repeated use. With this release, the Pdftools SDK stores the extracted native libraries in a persistent per-user cache; if the cache location is unusable, the SDK falls back to extracting into the temporary directory. As a result, the libraries no longer accumulate, and initialization is faster after the first run.
• Before this update, the Pdftools SDK ignored color key masks (transparency) when rendering indexed images with fewer than 8 bits per component. As a consequence, masked pixels rendered opaque instead of transparent. With this release, the Pdftools SDK applies color key masks to these images, and masked pixels render transparent as expected.

Added

• displayName field on the workflow objects returned by the Advanced API List available workflows endpoint, so the Playground and other API clients can show a workflow’s human-readable name instead of its technical name.

Fixed

This release fixes the following issues.

• The Standard Converter no longer omits watermarks defined in Word (.docx) documents from the converted PDF.
• Emails exported from Microsoft Word or Outlook, including messages forwarded through Gmail, no longer convert with a blank leading page caused by named-page markup in the exported HTML.
• The alt-text placeholder for a skipped external image no longer distorts the converted email’s layout when the image declares a width but no height.
• The Configurator Web displays the correct major.minor.patch version; earlier releases read the assembly version and always showed the patch component as zero.
• Profile import in both the Configurator Web and the Conversion Service Configurator compares only the major and minor version, so you can import a profile across patch releases within a minor line (for example, 6.12.1 into 6.12.0).
• The Conversion Service Configurator no longer becomes unresponsive to input on the splash screen when the Windows Defender troubleshooting check is slow or Windows Defender is unavailable, because the check runs asynchronously under a timeout.
• Certificate host name validation no longer fails on non-English operating systems (for example, French Windows), because the check reads Subject Alternative Name (SAN) DNS names independently of the display language and matches them against the host name following RFC 6125.

Fixed

• Before this update, the MRC optimizer removed the cmap and name tables from CID-only OpenType FontFile3 streams, producing font programs that strict OpenType validators rejected, although most viewers still rendered them correctly. With this release, table stripping is restricted to FontFile2 TrueType fonts, so embedded OpenType FontFile3 streams remain spec-compliant after optimization.

• Before this update, the ResolutionDPI and ThresholdDPI settings on the Web and MinimalFileSize optimization profiles affected only color and grayscale images, leaving bitonal images untouched. As a consequence, lowering the target resolution didn’t shrink files that contained high-resolution bitonal images. With this release, these settings also apply to bitonal images when explicitly configured.

• As of this release, Pdftools SDK no longer supports a glibc 2.12 build. The minimum supported glibc version is 2.17. Earlier releases of Pdftools SDK keep their glibc 2.12 build. If you need glibc 2.12 support in a newer releases of Pdftools SDK, contact Pdftools through the Contact page.

• Before this update, the revocation status of intermediate Certificate Authority (CA) certificates in the signing chain wasn’t reflected in the overall signature validity. As a consequence, a signature backed by a revoked intermediate CA could still be reported as valid. With this release, intermediate CA certificates are validated and their revocation propagates to the overall signature result.

• Before this update, post-analysis failures during PDF to PDF/A conversion were reported with a generic message that didn’t identify the object or page that caused the failure. As a consequence, applications had no diagnostic information to surface. With this release, the message includes the failing object ID, page number, and a count of any additional recorded errors, so applications can present actionable feedback.

Added

• Read the in-page scroll position of fit-height and fit-width bookmark destinations. The FitHeightDestination type exposes a Left property (the page location shown at the left border of the viewport), and the FitWidthDestination type exposes a Top property (the page location shown at the top of the viewport). Each returns null when the position before the jump is retained.

Changed

• As of this release, Toolbox add-on no longer supports a glibc 2.12 build. The minimum supported glibc version is 2.17. Earlier releases of Toolbox add-on keep their glibc 2.12 build. If you need glibc 2.12 support in a newer releases of Toolbox add-on, contact Pdftools through the Contact page.

Fixed

• Before this update, inspecting the font of an extracted text fragment through TextFragment.Font.BaseFont could crash the host application when the source PDF contained malformed font data, such as a missing FontDescriptor. As a consequence, applications that enumerated text fragments could terminate unexpectedly. With this release, the failure is reported as a regular SDK error, so applications can handle it gracefully.

• Before this update, creating a fit-height or fit-width destination discarded its horizontal (FitH) or vertical (FitV) scroll coordinate, so the viewport position was lost. With this release, FitHeightDestination.Create() accepts an optional left argument and FitWidthDestination.Create() accepts an optional top argument. As a result, the coordinate is preserved on the created destination and honored by viewers. Both parameters default to null, so existing code is unaffected.

Added

• A theming manager handles viewer CSS tokens with predefined light and dark themes. The theme option in ViewerConfig selects the active theme ('light', 'dark', or 'auto' for system preference). Developers can register custom themes using customThemes and override theme variables at runtime. Methods on PdfToolsViewer: addTheme(), clearThemeVariableOverrides(), getAvailableThemes(), getTheme(), overrideThemeVariables(), and setTheme().

• Viewer users can remove annotations using an eraser tool in the sidebar. Dragging a path deletes all overlapping annotations. A redaction-specific eraser removes only redaction annotations. Deleted annotations disappear from the document view, the annotation panel, and the redaction panel.

• The PDF Viewer SDK displays an alert when users close the tab or open a new document with unsaved changes. The promptOnUnsavedChange option in ViewerConfig controls this behavior (enabled by default).

• The disableMultipleImageInsert option in ViewerConfig deactivates the image annotation popover after each image is placed, preventing multiple insertions of the same image.

Changed

• Theme configuration replaced

The customCssVars configuration option is replaced by customThemes in ViewerConfig, letting developers register semantic theme definitions at initialization. A theme option selects the active theme.

Migration: Replace customCssVars: { '--var': 'value' } with customThemes: { myTheme: { interactive: { primary: '#value' } } } and set theme: 'myTheme'.

Fixed

• Before this update, annotations on rotated pages were positioned incorrectly and couldn’t be selected for editing. This affected free-text, stamp, sticky note, line, ink, shape, text markup, and redaction annotations. With this update, annotations respect page rotation. As a result, annotations display and behave correctly on rotated pages.

• Before this update, resizing an annotation to a small size and then zooming sometimes caused rendering errors and unstable behavior. With this update, viewer users can resize annotations to any size and zoom without rendering issues.

Added

• With this release, you can install and run the Conversion Service as native Linux services, without Docker. The packages install systemd units for the Conversion Service, Connector Service, and Web Configurator, and are available for Debian and Ubuntu (.deb) and for RHEL and Rocky Linux (.rpm). They bundle the Xvfb virtual display and the fonts required for HTML and email rendering. For more details, review Conversion Service on Linux.
• With this release, the Conversion Service exposes a built-in Model Context Protocol (MCP) endpoint that lets AI agents convert documents, combine PDFs into dossiers, and read service status, workflows, and statistics through the same engine as the Advanced API. For more details, review MCP server.
• Added an in-app Conversion Service client (Playground) to the Web Configurator. You can upload a file, choose a workflow and profile, run the conversion, and download the result directly in the browser, without configuring a connector first. Available on Linux and Docker. For more details, review Playground.
• Added support for the Extract XML from PDF and Convert to XML workflows in the Web Configurator. Previously, these workflows shipped with the service, but you could configure them only in the Windows-based Configurator. For more details, review the Convert to XML workflow documentation.
• Added connectors that integrate the Conversion Service with AI Smart Redact for AI-based detection and redaction of personally identifiable information (PII). The Schedule Smart Redact output connector submits processed PDFs to the Smart Redact Orchestrator, and the Poll Smart Redact input connector retrieves the redacted PDFs once they’re ready and forwards them to a workflow. The connectors track in-flight jobs in a local database inside the Connector Service.
• Added a Preserve Input Filename toggle to the Folder.Output connector. When you enable this toggle, the connector renames each result file to match the watched-folder input file. This renaming preserves the original file identity, for example, when you convert a single-file archive (ZIP). For more details, review Output folder.
• The Conversion Service now registers a built-in trial license automatically when you don’t supply a license key. As a result, you can start and evaluate the Conversion Service without providing a license key first. To use your own license, set the LICENSEKEY or LICENSEKEY_FILE environment variable as before, or configure them in the Conversion Service Configurator. For more details, review Trial license.

Changed

• New installations report license validation and consumption directly to the Pdftools Licensing Service rather than the License Gateway Service (LGS). The transport is controlled by the license.useLgs setting in appsettings.json, introduced in this release (container override LICENSE__USELGS). New installations default to false (direct), so no LGS container is required. Existing installations are migrated to true, which preserves their current LGS behavior.
• Improved the Web Configurator user experience with a range of fixes. Connection names are now validated both when adding a connection and when saving changes, path pickers apply their changes immediately and clear stale validation errors, and the Save button shows a clear indication when changes are blocked by validation. The Gmail connectors now provide clearer guidance, including Docker Compose hints on the App-Password connectors and a Windows-only callout with immediate validation on the OAuth2 connector. Stamp configuration validation is stricter (unit required, with mutual exclusion between left/right and top/bottom offsets), and the trust-store file upload re-validates when you clear or replace a file.
• Updated font substitution behavior, including PDF/A conversion. With this release, the underlying PDF engine applies improved font substitution. Document layout remains consistent, but when a required font is not embedded or otherwise available, the visual appearance of the substituted font may differ from that of earlier versions. To guarantee an exact visual match, install the specific original fonts in the environment where the Conversion Service runs. The service no longer ships the ZapfDingbats font and falls back to the URW++ D050000L.ttf font instead. As a consequence, documents that rely on ZapfDingbats may show minor visual differences. To preserve the original appearance, install ZapfDingbats in the environment where the Conversion Service runs.

Fixed

• Before this update, emails whose headers contained embedded carriage-return or line-feed characters (for example, .eml files produced by Microsoft Power Automate) used invalid header continuation lines. As a consequence, the converted PDF had an empty Subject, To, and Cc, and the page showed the raw email and its base64-encoded attachments as visible text. With this update, the service collapses embedded CR/LF runs in header values to a single space before serialization. As a result, these emails convert correctly, and the service extracts their attachments as expected.
• Before this update, Archive PDF/A workflows failed when combining a destructured PDF whose parent had to be rasterized during PDF/A conversion, because the service could no longer match the embedded child files to their destructure markers. As a consequence, affected jobs failed. With this update, the Pdf.Combine step reports a Child Removed warning and continues. As a result, these jobs complete instead of failing.
• Before this update, the Connector Service and Web Configurator containers could fail to access their mounted volumes in Docker because of file ownership and permission mismatches. With this update, the containers set up the required volume permissions at startup. As a result, the Connector Service and Web Configurator start and operate correctly against their mounted volumes.
• With this release, the Conversion Service logs errors returned by the Pdftools OCR Service with the OCR job ID and ties them to the originating Conversion Service job and input file. As a result, you can correlate an OCR failure back to the specific job and document that triggered it. For more details, review Service log.

VERSIONS EARLIER THAN 6.12.0 ARE NO LONGER AVAILABLE

Pdftools removed Conversion Service versions earlier than 6.12.0 from the customer portal. Versions 6.12.0 and later remain available.

Changed

• HEVC decoding library (libde265) is dynamically linked

The libde265 library, which reads HEVC-encoded HEIF images, ships as a separate dynamically linked library. The native library requires the HEIF/HEIC decoding library at load time, even if your application doesn’t use HEIF/HEIC functionality. If Pdftools SDK can’t find the libde265 decoding library, it fails to load. Pdftools SDK includes the required binaries.

Action:

After updating, verify that your deployment can still decode HEVC-encoded HEIF images.

Fixed

• Before this update, the Web and MinimalFileSize optimization profiles didn’t downsample or recompress bitonal (scanned black-and-white) images. As a consequence, optimizing a scanned PDF with these profiles produced a file roughly the same size as the input. With this release, these profiles downsample bitonal images and re-encode them with JBIG2. As a result, scanned documents optimize to substantially smaller files.

Changed

• HEVC decoding library (libde265) is dynamically linked

The libde265 library, which reads HEVC-encoded HEIF images, ships as a separate dynamically linked library. The native library requires the HEIF/HEIC decoding library at load time, even if your application doesn’t use HEIF/HEIC functionality. If Toolbox add-on can’t find the libde265 decoding library, it fails to load. Toolbox add-on includes the required binaries.

Action:

After updating, verify that your deployment can still decode HEVC-encoded HEIF images.

Fixed

• Before this update, extracting text in C# on Linux dropped or replaced non-ASCII characters in the output. As a consequence, Toolbox add-on returned content with accented or extended Unicode characters incorrectly. With this release, text extraction preserves all UTF-8 characters. As a result, extracted text matches the source document on Linux.

Added

• Viewer users can fill in PDF form fields, toggle checkboxes and radio buttons, type into text boxes, and select options from combo boxes and list boxes. The PDF Web SDK exposes these form fields through the Forms namespace with CheckBoxWidget, RadioButtonWidget, TextBoxWidget, ComboBoxWidget, and ListBoxWidget classes.
• Developers can configure automatic navigation to a PDF’s initial destination using the useInitialDestination option. Set this option in ViewerConfig during initialization or pass it to Document.open(). When enabled, the viewer navigates to the PDF’s initial destination (if defined) when it opens.
• The hidePasswordDialogCancelButton option in ViewerConfig lets developers hide the cancel button in the password dialog. As a result, users must enter a password to proceed.
• New localization methods on PdfToolsViewer enable runtime language management:
  • addLanguage(): adds a new language at runtime.
  • customizeTranslations(): overrides specific translation strings.
  • getAvailableLocales(): lists all available locales.
  • getLocale(): returns the current locale.
  • setLocale(): switches the active locale (supports 'auto' for browser detection).
• The npm package includes third-party open-source license texts in the licenses/ directory. See licenses/MANIFEST.txt for the full list of components.

Changed

• The hideComponents(), showComponents(), and overrideButtonBehavior() methods accept component names instead of path-based references. A type alias provides strict typing for all valid component names.
• The file upload control is disabled while a document is loading, preventing users from uploading another file during the process.
• The viewer preserves the zoom level, fit mode, and page layout when opening a new document.
• The note annotation dialog has an improved layout and full keyboard accessibility.
• The password dialog has an improved layout and full keyboard accessibility.
• The print dialog dots per inch (DPI) selector displays a tooltip explaining each DPI value on hover.

Breaking changes

Localization configuration renamed

The customLocalisationFile configuration option is replaced by customTranslations in ViewerConfig, letting developers provide translation overrides directly in code instead of loading external files. The lang property is replaced by locale, which supports built-in and custom locales, as well as automatic browser detection.

Migration:
Replace customLocalisationFile: 'path/to/file.json' with customTranslations: { en: { key: 'value' } }. Replace lang: 'en' with locale: 'en'.

AnnotationType enum values renumbered

The AnnotationType enum values have shifted to accommodate form widget members (CheckBox 19, RadioButton 20, TextBox 21, ListBox 22, ComboBox 23). Affected values: Screen 19→24, PrinterMark 20→25, TrapNet 21→26, Watermark 22→27, Redact 23→28, 3D 24→29.

Migration:
Use the named enum members (AnnotationType.Redact) instead of raw integer values.

Fixed

• Before this update, printing on mobile devices failed silently or produced blank output when triggered from the print dialog. With this update, the print flow correctly renders document content on iOS and Android devices.
• Before this update, embedding the viewer in CSS flex or grid containers caused layout overflow and scrollbar issues. As a consequence, the viewer did not size correctly within modern CSS layouts. With this update, the viewer respects container constraints. As a result, the viewer displays correctly in flex and grid containers.
• Before this update, running multiple viewer instances on the same page caused style conflicts and rendering errors. With this update, viewer instances are properly isolated. As a result, multiple viewers can run simultaneously on the same page.

Added

• Standard Converter for Office documents With this release, the Conversion Service converts Word, Excel, and PowerPoint files to PDF without requiring a Microsoft Office installation, Microsoft 365 subscription, or Office license feature. The Standard Converter is enabled by default and works on both Windows and Docker deployments. For highest visual fidelity, configure a local Microsoft Office installation or the Office for the web service.
  • The Standard Converter supports the following profile options:
    • Track Changes (Word): Renders revision marks (insertions, deletions, formatting changes) in the converted PDF.
    • Allow Scaling (Excel): Scales worksheet columns to fit on one page.
• Added a Force line break option to preserve Rich Text Format (RTF) line breaks in email text. Generic RTF control-word stripping prevents leftover control characters in the output.
• Added an Include external stylesheets option, separate from Include external images, so you can skip external CSS independently.
• Added missing features to the Configurator Web, reducing the gap with the Windows-based Configurator. Additions include:
  • Signing, Stamping, Annotations, and OCR configuration in profiles.
  • Connector configuration in the Integrations tab for all connector types, including the ability to create connectors and edit existing ones.
• Added a Flatten Form Fields toggle in the Compression and Optimization section of the Configurator profile settings. Turn it on to flatten AcroForm fields into static graphics so that filled-in form values render consistently across PDF viewers. The toggle works in the PDF/A-1, PDF/A-2, and PDF/A-3 workflows. The runtime job option is FLATTEN.FORM.FIELDS. For details, see Compression and optimization settings.

Fixed

• Before this update, PPTX files with restricted embedded fonts caused conversion to hang. As a consequence, jobs stalled and didn’t return a result. With this update, the service strips restricted embedded fonts before conversion. As a result, these PPTX files convert without hanging.
• Before this update, the Pdf.Resize step failed with an ArgumentException on PDFs that contain AcroForm fields, which caused affected jobs to fail. With this update, the resize step catches the exception across the full resize block and skips the resize gracefully when the document is incompatible. As a result, jobs that include AcroForm PDFs complete without failing.
• Before this update, deployments that ran both the Connector Service and the Configurator Web Docker containers failed to start when host permissions on the shared connector volume (which holds Connections.xml) didn’t match the container user. As a consequence, operators had to adjust the volume permissions manually before the containers could read or write Connections.xml. With this update, the container entry points adjust the permissions at startup. As a result, the Connector Service and Configurator Web start reliably across host configurations.

Added

• The Annotation class supports storing custom data through the privateData property of type string. The property persists across save and load cycles, letting applications attach domain-specific metadata to each annotation.
• The Document API emits a pageRendered event when a page finishes rendering. The event payload is the rendered page number.

Changed

• The addAnnotation(), deleteAnnotation(), and updateAnnotation() methods on viewer.document accept either a single annotation or an array of annotations, supporting batch operations with a single call.
• The viewer.document.getPage() method returns a Page synchronously instead of a Promise. Update existing code to call viewer.document.getPage(pageNumber) directly, without await.

Fixed

• Before this update, setting ViewerConfig.devicePixelRatio caused the viewer to crash at browser zoom levels other than 100% because the calculated canvas dimensions didn’t match the rendered buffer. As a consequence, viewer instances that configured devicePixelRatio failed whenever users zoomed the browser. With this update, the viewer derives the effective device pixel ratio from the rendered buffer size. As a result, the viewer renders correctly at any browser zoom level.
• Before this update, text selection under certain circumstances ended one character before the end of the selected word. As a consequence, viewer users couldn’t include the final character when selecting a complete word. With this update, the text-selection logic includes the last glyph in the range. As a result, viewer users can select complete words.

Added

• Dithering control for Fax image conversion The Fax conversion profile exposes a dithering option, letting you switch between dithering modes when rendering PDFs to fax-ready images. For more details, refer to Fax conversion profile.

Fixed

• Before this update, PDF/A conversion incorrectly reported a device-color-space error on documents containing nested form XObjects. As a consequence, valid documents couldn’t be converted to PDF/A. With this release, the converter resolves color spaces correctly. As a result, these documents convert to PDF/A as expected.
• Before this release, optimizing PDFs with certain internal structures caused the optimizer to perform many redundant object comparisons. As a consequence, optimization appeared to hang and didn’t complete on the affected documents. With this release, the optimizer avoids the redundant comparisons. As a result, optimization finishes in the expected time on the affected documents.
• Before this update, validating a malformed PDF containing nested object streams caused unbounded recursion. As a consequence, the process crashed with a stack overflow on the affected documents. With this release, the validator detects the cycle and rejects the document with a clean validation error. As a result, malformed input can no longer crash the validator.