PDF creation and document layout for .NET, under the MIT licence.
FreePDF is io7's independent MIT continuation of PDFsharp and MigraDoc. PDFsharp provides PDF drawing, reading, and modification; MigraDoc provides document layout with paragraphs, tables, headers, and footers.
The starting point is empira's MIT-licensed v6.2.4 release. FreePDF preserves
that source history and attribution. It is maintained independently by io7
and is not an official empira release.
PDFsharp 6.5.0 and 7.0.0 Preview 2 (October 2026) are licensed under the AGPL or
a commercial licence, and the application must select a licence in code before
first use. FreePDF continues the MIT line from 6.2.4: replace the PDFsharp
packages with the matching io7.FreePDF packages and rebuild.
In a survey of 26 public projects
that use PDFsharp 6.2.4, 23 compiled against FreePDF with no changes. The
migration guide lists
what the other three needed, each a one-line change or a font file you supply.
Applications that fill forms through the legacy PdfSharp.Pdf.AcroForms model
move over with renames; see Filling forms.
One call turns a document into a hybrid e-invoice: PDF/A-3 with your UN/CEFACT
Cross Industry Invoice XML embedded as factur-x.xml and declared the way
Factur-X 1.0 and ZUGFeRD 2.x require (associated file, XMP extension schema).
using var document = new PdfDocument();
FacturXManager.ForDocument(document).SetInvoice(invoiceXml, FacturXProfile.EN16931);
var page = document.AddPage(); // then draw the human-readable invoice
document.Save("invoice.pdf");Call SetInvoice before adding pages. FreePDF checks that the XML is a
well-formed Cross Industry Invoice; checking its content against EN 16931 and
the profile's business rules is up to your invoicing software. CI validates
the generated PDF/A-3 file with veraPDF.
var form = document.Catalog.GetOrCreateAcroForm();
form.AddTextField("name", page, new XRect(150, 70, 220, 22));
form.AddCheckBox("agree", page, new XRect(150, 110, 14, 14));
form.AddRadioButtonGroup("size", page, new[] { ("S", new XRect(150, 150, 14, 14)), ("M", new XRect(190, 150, 14, 14)) });
form.AddComboBox("country", page, new XRect(150, 190, 150, 22), new[] { "Norway", "Sweden", "Denmark" });
form.AddSignatureField("signature", page, new XRect(150, 240, 200, 50));
document.Save("form.pdf"); // a fillable form
// Later: fill it in and certify it, so that only form filling keeps the signature valid.
using var filled = PdfReader.Open("form.pdf", PdfDocumentOpenMode.Modify);
((PdfFormTextField)filled.Catalog.GetAcroForm()!.Fields["name"]!).Text = "Kari Nordmann";
var signer = new PdfSharpDefaultSigner(certificate, PdfMessageDigestType.SHA256);
DigitalSignatureHandler.ForDocument(filled, signer, new DigitalSignatureOptions
{
FieldName = "signature",
Certification = PdfCertificationLevel.FormFilling,
});
filled.Save("certified.pdf"); // the signature is made when the file is savedRectangles are in drawing coordinates, as for XGraphics: points from the top
left corner of the visible page. document.Flatten() turns filled fields into
page content. To fill in a signed or certified file without breaking its
signature, open it with PdfDocumentOpenMode.Append. PDF/A documents take
check boxes, radio buttons and signature fields, but not text and choice fields
(their Helvetica is not embedded); PDF/UA documents take no new fields yet.
QR codes, Data Matrix, Code 128 (including GS1-128), EAN-13, EAN-8 and UPC-A are drawn as vector shapes, so they stay sharp in print. No image library is needed.
using PdfSharp; // DrawBarCode and DrawMatrixCode are extension methods here
using PdfSharp.Drawing.BarCodes; // CodeQR, CodeDataMatrix, Code128, Ean13, Ean8, UpcA
var qr = new CodeQR("https://example.com/pay/2026-118", new XSize(90, 90)); // level M, version chosen to fit
gfx.DrawMatrixCode(qr, new XPoint(40, 40));
gfx.DrawBarCode(new Code128("PO-2026-000123", new XSize(160, 40)) { TextLocation = TextLocation.Below },
XBrushes.Black, new XFont("Arial", 9), new XPoint(40, 160));
gfx.DrawBarCode(new Ean13("590123412345", new XSize(113 * 1.5, 80)), new XPoint(40, 230)); // check digit addedIn MigraDoc, add a barcode to a section, cell, text frame, header or footer:
var barcode = section.AddBarcode();
barcode.Type = BarcodeType.QRCode; // also DataMatrix, Barcode128, Ean13, Ean8, UpcA, Barcode39, Barcode25i
barcode.Code = "https://example.com/pay/2026-118";
barcode.Width = barcode.Height = Unit.FromCentimeter(3);The size of a QR code includes its quiet zone, and the size of an EAN or UPC
code includes the quiet zones that hold its outer digits; in MigraDoc the
shape's size includes the quiet zones of every type. Text that is not
ASCII goes into QR codes as UTF-8 with the standard's ECI marker; set
Utf8Eci = false for payment apps that expect plain UTF-8. Each code type was
checked against an independent decoder; see the support policy.
The layout engine writes tagged PDF for PDF/UA-1, for documents that must be accessible, for example under the European Accessibility Act:
document.Info.Title = "Invoice 2026-001"; // required
document.Culture = CultureInfo.GetCultureInfo("nb-NO"); // the document language
logo.AlternativeText = "Example Ltd logo"; // every image and chart, or Decorative = true
var renderer = new PdfDocumentRenderer { Document = document, TaggedPdf = true };
renderer.RenderDocument();Headings (paragraphs with an outline level), paragraphs, lists, tables, images,
charts and links are tagged; headers, footers, borders and shading become
artifacts that screen readers skip. Rendering stops with an explanation if an
image has neither alternative text nor Decorative, rather than writing a file
that does not conform. CI validates a tagged multi-page invoice with veraPDF.
See the support policy for what is covered.
The native PdfAccessibilityScanner.Scan("document.pdf") API inspects existing
PDFs without changing them. It reports structural failures, incomplete checks and
required human review, with explicit partial PDF/UA-1 coverage. The freepdf-a11y
CLI prints a plain-English summary and exports JSON, SARIF and offline HTML reports.
This is an accessibility preflight,
not ADA certification; see the scanner guide.
The producer corpus defines 84,000 synthetic PDFs from
FreePDF, ReportLab and Chromium, with tiered CI checks for text, rendering and
accessibility regressions.
using var merged = new PdfDocument();
foreach (var path in files)
{
using var source = PdfReader.Open(path, PdfDocumentOpenMode.Import);
merged.AppendPages(source, new PdfPageImportOptions { BookmarkTitle = Path.GetFileName(path) });
}
merged.Save("merged.pdf");
using var report = PdfReader.Open("report.pdf", PdfDocumentOpenMode.Import);
using var summary = report.ExtractPages(0, 2); // the first two pages
summary.Save("summary.pdf");Form fields stay a form you can fill or flatten; a field name that is already
taken gets a suffix (name_2), and AppendPages tells you which. Links,
buttons and bookmarks lead to the pages in the new file, and a split file does
not carry hidden copies of the pages it left out. AddPage(page) keeps PDFsharp 6.2.4's
behaviour; see Merging and splitting.
using var document = PdfReader.Open("signed.pdf", PdfDocumentOpenMode.Append);
((PdfFormTextField)document.Catalog.GetAcroForm()!.Fields["name"]!).Text = "Ada Lovelace";
DigitalSignatureHandler.ForDocument(document, signer, options); // optional: sign it again
document.Save("signed-filled.pdf");In Append mode, saving keeps the original file byte for byte and adds only the
new and changed objects after it, as an incremental update. The signatures
already in the file stay valid, and a new signature covers the whole file,
earlier signatures included. A document with no changes is saved as it was.
A change that a certification signature or a field lock does not permit is
refused before anything is written. An encrypted document (opened with its
owner password, or its user password and AllowModifyWithoutOwnerPassword)
keeps its encryption, and the update is encrypted the same way.
To sign a form's empty "sign here" field instead of adding a new one, set
FieldName = "Approval" in the DigitalSignatureOptions; its lock and required
seed values are honoured.
See
Signed documents.
var signer = new PdfSharpDefaultSigner(certificate, PdfMessageDigestType.SHA256, new Uri("https://tsa.example.com/"))
{
Format = PdfSignatureFormat.Pades, // ETSI.CAdES.detached
};
DigitalSignatureHandler.ForDocument(document, signer, new DigitalSignatureOptions { Rectangle = new XRect(36, 700, 200, 50) });
document.Save("signed.pdf");A PAdES signature (ETSI EN 319 142-1) binds the signer's certificate with the
signing-certificate-v2 attribute and has no signing-time attribute: the claimed
time is the signature dictionary's /M. Without a time-stamp authority it is
PAdES B-B; with one, the signature gets a signature time-stamp and is B-T.
Set TimeStampClient to send the RFC 3161 request yourself, for example with
authentication; TimeStampReserve (16384 bytes by default) is the room kept
for the authority's token. Time-stamps need .NET 8 or later. PAdES needs
SHA-256 or stronger. Below PDF 2.0 the catalog declares the ESIC extension.
The default format is still adbe.pkcs7.detached.
The signer calls the authority while the document is saved, so prefer
await document.SaveAsync(...) in UI and ASP.NET code; Save blocks until
the authority answers (at most 30 seconds over HTTP). A signed document saved
to a path is written only once the signature is made.
// B-LT: the chains of the signatures and time-stamps, with an OCSP response or the CRLs for each certificate.
using (var document = PdfReader.Open("signed.pdf", PdfDocumentOpenMode.Append))
{
var report = await PdfDocumentSecurityStore.AddValidationDataAsync(document,
new PdfValidationDataOptions { ExtraCertificates = [intermediate, root] });
if (report.Problems.Count > 0)
Console.WriteLine(string.Join(Environment.NewLine, report.Problems));
await document.SaveAsync("signed.pdf");
}
// B-LTA: a document time-stamp over the document and its validation data.
using (var document = PdfReader.Open("signed.pdf", PdfDocumentOpenMode.Append))
{
DigitalSignatureHandler.ForDocument(document, new PdfTimeStampSigner(new Uri("https://tsa.example.com/")), new DigitalSignatureOptions());
await document.SaveAsync("signed.pdf");
}AddValidationDataAsync asks the OCSP responders the certificates name and
falls back to their CRL distribution points; give it intermediate and root
certificates the signatures do not carry, and an HttpClient of your own for
a proxy. The addresses come from the certificates, which whoever signed the
document chose: the default client reaches only public addresses (set
AllowPrivateNetworks for an internal CA), checks every redirect and the
address it connects to (on .NET Framework and .NET Standard it does so for
HTTP addresses only, the usual case for OCSP and CRLs), uses no proxy (pass
your own HttpClient behind one), takes at most
MaxResponseSize bytes per answer (32 MiB) and 30 seconds per request, and
stops at the cancellation token. It does not
judge the answers: Problems lists what it could not get and certificates
OCSP reports as revoked. To add data you fetched yourself, use
PdfDocumentSecurityStore.Add. Both kinds of update leave the signed bytes as
they are and are permitted under every certification level, "no changes"
included. A document time-stamp is an invisible field; it needs .NET 8 or
later.
var results = await PdfSignatureVerifier.VerifyAsync(File.ReadAllBytes("signed.pdf"),
new PdfSignatureVerificationOptions { TrustedRoots = [root] });
foreach (var signature in results)
Console.WriteLine($"{signature.FieldName}: {signature.Status} {signature.PadesLevel} {string.Join("; ", signature.Reasons)}");For each signature and document time-stamp, PdfSignatureVerifier reports
whether the bytes it covers are intact, the signer and the chain, whether the
chain ends at a trusted root (the ones you give, or the system's), the claimed
and the time-stamped signing time, revocation (from the DSS by default, online
on request), the first later change its certification or field locks forbid,
and the PAdES level reached. A certificate that has expired since counts if
it was valid at the signature's time-stamp, and only a time-stamp from a
trusted time-stamp authority counts. After an approval signature, a later
update that changes more than form fields, signatures and validation data
(a comment, for example) is reported as UnexpectedChange. Status is Valid,
Invalid (changed bytes, a revoked certificate, a forbidden change) or
Indeterminate (trust, revocation or later changes not established as
harmless), with the reasons. It is a technical
check: EU trusted lists, qualified status and signature policies are not
checked. Time-stamps are checked on .NET 8 or later.
FreePDF 7.0 adopts the verified MIT 7.0.0-preview-1 source, carries
our earlier fixes, and adds independently implemented reader, image, signature,
font-cache and layout corrections. 7.0.1 gives the assemblies io7's own
strong-name key and FreePDF/io7 product metadata; FreePDF is not an empira
release. 7.1.0 adds the GDI and WPF packages, form filling, Factur-X / ZUGFeRD
invoices, tagged PDF/UA-1 output and the fixes from a real-PDF corpus round trip
and parser fuzzing; 7.1.1 measures filled form text with Adobe's standard font
metrics. 7.2.0 draws text with installed fonts when no font resolver is set,
flattens forms and tags multi-page tables as one table. 7.3.0 merges and splits
documents with their form fields, bookmarks and links, and names FreePDF as the
producer of the files it writes. 7.4.0 saves documents as incremental updates,
so signed PDFs can be filled in and signed again; 7.4.1 fixes cases where that
broke earlier signatures and opens encrypted files with incremental updates.
7.5.0 refuses Append-mode changes that a document's signatures do not permit
and opens encrypted documents in Append mode; 7.5.1 encrypts empty streams and
strings in AES-encrypted documents, and 7.5.2 reads encrypted metadata streams in
documents that leave metadata unencrypted. 7.6.0 signs a form's existing, empty
signature field instead of adding a new one; 7.6.1 reads multi-select list boxes
and fixes how metadata is encrypted. 7.7.0 adds, removes and moves pages in
Append mode when the page tree is nested, and fixes metadata in encrypted
documents. 7.8.0 opens damaged files that viewers open, fixes MigraDoc crashes
and hangs, and keeps PDF/A files conforming. 7.9.0 draws QR, Data Matrix, Code 128,
EAN and UPC codes as vectors, renders MigraDoc barcodes, creates form fields and
certifies documents when signing. 7.10.0 signs PAdES B-B and B-T signatures,
adds validation data and document time-stamps (B-LT, B-LTA) and checks the
signatures of a file; 7.10.1 hardens reading, flattening and signature checking
against hostile files; 7.10.2 corrects which revocation data the signature
check accepts. 7.11.0 adds accessibility checking through a native API and the
freepdf-a11y CLI, improves coordinate precision and defines an 84,000-PDF
regression corpus. In 7.12.0 the CLI lists the Matterhorn 1.1 requirements and
can run veraPDF as an independent PDF/UA-1 check. In 7.13.0 the native scanner
checks content inside nested Form XObjects and resolves named marked-content
properties. 7.14.0 bounds the scan's memory and time on hostile files and adds
a scan time limit. In 7.14.1 the time limit and cancellation also stop the
reading of the PDF. See the release notes for 7.14.1, 7.14.0, 7.13.0,
7.12.0, 7.11.0, 7.10.2, 7.10.1, 7.10.0, 7.9.0, 7.8.0, 7.7.0, 7.6.1, 7.6.0, 7.5.2, 7.5.1, 7.5.0, 7.4.1, 7.4.0, 7.3.0, 7.2.0, 7.1.1, 7.1.0,
7.0.1 and 7.0.0, the
changelog and
upstream feature tracking.
Published versions are listed in GitHub Releases
and NuGet; CI artifacts alone are not a public release.
Replace each upstream package with its FreePDF counterpart. Namespaces and assembly names remain PdfSharp
and MigraDoc, so an application references exactly one implementation:
| Upstream package (6.2.4) | FreePDF package | Targets |
|---|---|---|
PDFsharp |
io7.FreePDF |
.NET 8, 9, 10, .NET Standard 2.0 |
PDFsharp-MigraDoc |
io7.FreePDF.Layout |
.NET 8, 9, 10, .NET Standard 2.0 |
PDFsharp-GDI |
io7.FreePDF.Gdi |
Windows .NET 8, 9, 10, .NET Framework 4.6.2 |
PDFsharp-MigraDoc-GDI |
io7.FreePDF.Layout.Gdi |
Windows .NET 8, 9, 10, .NET Framework 4.6.2 |
PDFsharp-WPF |
io7.FreePDF.Wpf |
Windows .NET 8, 9, 10, .NET Framework 4.6.2 |
PDFsharp-MigraDoc-WPF |
io7.FreePDF.Layout.Wpf |
Windows .NET 8, 9, 10, .NET Framework 4.6.2 |
Each layout package depends on its matching core package. The separate
io7.FreePDF.Accessibility.Tool package supplies freepdf-a11y for .NET 10.
Build and test all seven packages with PowerShell 7:
./eng/Test-Packages.ps1 -OutputDirectory /absolute/path/to/candidate-packages
# Windows also runs Framework consumers and the GDI/WPF consumers (net8.0-windows and net472 by default):
./eng/Test-Packages.ps1 -OutputDirectory D:/test/freepdf-packages -Frameworks net8.0,net9.0,net10.0,net462,net472The Candidate packages workflow publishes build artifacts, not NuGet releases.
See package and migration notes. The inherited upstream
packaging projects are blocked to prevent output under upstream identities.
The v7 migration guide,
support policy and
dependency audit define the release's scope.
CI additionally rehearses unpublished plain stable packages, checks public APIs
against pinned MIT baselines, and validates generated output with qpdf, Poppler
and veraPDF.
The manual publication workflow promotes verified CI
artifacts without rebuilding them and authenticates through NuGet Trusted Publishing.
- The inherited code remains under the MIT licence.
- New contributions to FreePDF are accepted under MIT.
- Original copyright notices and third-party licences remain intact.
- PROVENANCE.md records the exact source baseline and checksums.
- CONTRIBUTING.md explains the contribution policy.
Later upstream releases use different licensing. Their changes are not automatically eligible for inclusion here. Each imported change must have documented MIT-compatible provenance; otherwise, fixes must be implemented independently from behavioural requirements and the permitted source baseline.
Install the .NET 10 SDK. From the repository root:
dotnet build src/foundation/src/MigraDoc/src/MigraDoc.Rendering/MigraDoc.Rendering.csproj --configuration Release -p:TargetFrameworks=net10.0 -p:TargetFramework=net10.0
dotnet build src/foundation/src/MigraDoc/src/MigraDoc.RtfRendering/MigraDoc.RtfRendering.csproj --configuration Release -p:TargetFrameworks=net10.0 -p:TargetFramework=net10.0These builds include the PDFsharp and MigraDoc dependencies. CI also compiles the core libraries for .NET 8 and .NET Standard 2.0. Compilation does not prove PDF conformance or layout correctness.
Run the synthetic regression suite without external fonts or documents:
dotnet test src/tests/FreePDF.RegressionTests/FreePDF.RegressionTests.csproj --configuration Release --framework net10.0CI runs these regressions on Linux and Windows with .NET 8/10 and on Windows with .NET Framework 4.7.2 (the test runner's minimum). The comparison lab also executes console consumers targeting Framework 4.6.2 on Windows. CI compiles the GDI and WPF variants for Framework 4.6.2 and modern Windows .NET. These checks do not certify every inherited feature; the feature matrix records remaining gates. See 7.0 validation for the observed checks and limits.
The comparison lab generates identical authored fixtures in separate MIT PDFsharp and FreePDF processes. A read-only browser workbench shows the PDFs, raster comparisons, overlays, differences and structural checks. Known reference defects remain visible for review. The lab can run permanently on Windows alongside other applications; its installer requires operator approval.
The research and roadmap records the reviewed upstream issue inventory, Pdfxport.Net findings, competitor features and release priorities. Source availability and pixel similarity are not conformance certification.
The full upstream solution includes Windows projects and tests requiring external assets. See the archived upstream build notes. External assets are not part of this source baseline and require a separate licence/provenance check before importing or redistributing them.
Open issues and pull requests at https://github.com/io7/FreePDF. Include a small, synthetic reproduction and the expected result. Do not upload customer documents, credentials, or other confidential information.
PDFsharp and MigraDoc were created and maintained by empira Software GmbH and their contributors. Their original author credits and release README are preserved in docs/upstream/README-v6.2.4.md. The embedded BigGustave PNG reader retains its own licence in the source tree.