tbytes_pdf_flutter #

Fill, flatten, decrypt, and sign PDF forms in Flutter — parse AcroForm fields, fill them, flatten filled forms into static content, decrypt password-protected PDFs, and place signature images onto PDF pages. No native PDF SDK, no platform channels, pure Dart.

This package is deliberately not a PDF viewer. Pair it with whichever viewer you already use (pdfrx, flutter_pdfview, syncfusion_flutter_pdfviewer, etc.) — see example/ for a working combination with pdfrx + flutter_pdfview.

There was no solid native option for this on Flutter — Android had no good package for AcroForm filling, and iOS's PDFKit annotations would have meant heavy per-platform method-channel work with no shared data model. So this package's PDF engine was built from Adobe's openly published spec instead of wrapping an existing library — see How this was built for the full story, and Contributing if you'd like to help extend it.

Features #

Getting started #

dependencies:
  tbytes_pdf_flutter: ^0.1.0

import 'package:tbytes_pdf_flutter/tbytes_pdf_flutter.dart';

Usage #

The examples below are trimmed for readability. See example/lib/pdf_feature_demo_screen.dart for a complete, runnable demo of every feature with a full Flutter UI around it.

1. Load a PDF and read its AcroForm fields #

final bytes = await File(path).readAsBytes();
final doc = PdfDoc.load(bytes);
final reader = AcroFormReader(doc);

if (reader.hasForm) {
  final fields = reader.readFields();
  for (final field in fields) {
    print('${field.name}: ${field.type} = ${field.value}');
  }
}

2. Fill form fields #

final reader = AcroFormReader(doc);

reader.setFieldValue('full_name', 'Jane Doe');           // text field
reader.setFieldValue('subscribe', true);                  // checkbox
reader.setFieldValue('plan', 'Pro');                       // radio button

final bytes = PdfWriter().write(doc);
await File(outputPath).writeAsBytes(bytes);

3. Place a signature image on a field #

// signaturePng: Uint8List of a PNG, e.g. rendered from a drawing canvas.
// SignatureImagePlacer auto-trims transparent padding and preserves
// aspect ratio when fitting the image into maxWidth x maxHeight.
final field = reader.findField('signature_1')!;

SignatureImagePlacer(doc).placeSignatureOnField(
  field: field,
  imageBytes: signaturePng,
  maxWidth: 200,
  maxHeight: 60,
  offsetX: 10,
  offsetY: 5,
  transparentBackground: true,
);

final signedBytes = PdfWriter().write(doc);

4. Flatten filled fields into static content #

Once a form is filled (and optionally signed), flattening bakes the appearance streams directly into page content so the result is no longer an editable form — useful before archiving or sending a final copy.

final flattener = PdfFlattener(doc);
final result = flattener.flatten(
  quality: FlattenQuality.high,     // standard (72dpi) | high (150dpi) | ultra (300dpi)
  target: FlattenTarget.all,        // stampsOnly | formsOnly | all
);

print(result); // FlattenResult(pages: 3, flattened: 5, skipped: 0)

final flatBytes = PdfWriter().write(doc);

5. Decrypt a password-protected PDF #

final decryptor = PdfDecryptor(doc);

if (decryptor.isEncrypted) {
  final ok = decryptor.tryDecrypt('the-password'); // or '' for owner-only encryption
  if (!ok) {
    throw Exception('Wrong password or unsupported encryption scheme');
  }
}

final decryptedBytes = PdfWriter().write(doc);

6. Read/write document metadata #

final info = PdfInfoStore(doc);
info.setString('Title', 'Signed Agreement');
info.setString('Author', 'Jane Doe');

final title = info.getString('Title');

Running the example #

cd example
flutter pub get
flutter run

The example lets you paste a PDF URL or local path and exercises every feature above through a tabbed UI (Fields → Fill → Sign → Flatten → Decrypt → Result), using pdfrx and flutter_pdfview purely to view the output — those two packages are not dependencies of tbytes_pdf_flutter itself, only of the example app.

Why is there a vendored image package subset inside this package? #

SignatureImagePlacer needs to decode PNG signature images and auto-crop transparent padding. Rather than depending on package:image directly — which many apps already depend on transitively at a different, pinned version, causing dependency resolution conflicts — this package vendors only the ~85 files actually needed for that specific job, under lib/src/image_decoder/, with its own internal namespace. It cannot conflict with your app's own image dependency, however you have that pinned.

Full details, exactly what's included vs. excluded, and license attribution: lib/src/image_decoder/README.md.

Third-party licenses #

This package depends on archive, crypto, and encrypt, and vendors a trimmed subset of image. The example app additionally uses pdfrx and flutter_pdfview for PDF viewing. Full attribution and license text for all of these: THIRD_PARTY_NOTICES.md.

How this was built #

I couldn't find a good native option for this on Flutter. On Android there was no solid package for filling AcroForm fields directly. On iOS, PDFKit exposes annotations, but wiring that up properly would have meant heavy method-channel work per platform, keeping two native implementations in sync, and still not getting a shared Dart-side data model to reason about fields, values, and coordinates the same way on both platforms. So instead of going down the method-channel route, I decided to actually learn the PDF format and write a pure-Dart engine that works identically everywhere.

I went through Adobe's PDF Reference, version 1.7 — the openly published specification of the file format — to understand PDF from the byte level up, rather than from the outside-in shape of an existing library's API. A PDF isn't really "one format" so much as a byte stream that has to be pulled apart layer by layer, and building PdfTokenizer and PdfParser meant reading it exactly like that: a PDF loads as a Uint8List, and the tokenizer walks it byte-by-byte, comparing each byte against the hex codes for PDF's structural punctuation — 0x3C 0x3C for << (dictionary open), 0x2F for / (a name like /AcroForm or /FT), 0x5B/0x5D for array brackets, and so on — to figure out where dictionaries, names, arrays, and streams start and end, with no assumptions about what "kind" of PDF it is.

Once the tokenizer could turn raw bytes into a stream of typed tokens, the next problem was finding the form itself. An AcroForm isn't a special file section — it's just a /AcroForm dictionary hanging off the document's root /Catalog, pointing at a /Fields array, where each field dictionary carries a /FT (field type — Tx for text, Btn for button/checkbox/radio, Ch for choice, Sig for signature) and often inherits attributes like /FT or /DA from a /Parent field rather than declaring them itself. AcroFormReader walks that structure recursively, resolving indirect references (PdfRef → the actual object via the cross-reference table) along the way, since a field's kids, parent, and even its own value can each live in a completely different part of the file. Getting field values in and back out correctly also meant handling PDF's own string encodings (literal strings, hex strings, and the octal/ UTF escape sequences inside them) so that filled-in text round-trips correctly instead of corrupting on write.

The last piece — placing a signature on top of an existing AcroForm field rather than replacing it — meant generating a PDF content stream by hand: creating an Image XObject from the decoded signature PNG, adding it to the target page's /Resources, and emitting the cm/Do operators that position and draw it at the field's /Rect, converted into PDF's Y-up coordinate space. PdfFlattener builds on the same idea for baking filled values and stamped signatures permanently into page content afterward.

None of this wraps an existing PDF library — it's the spec, read end-to-end and implemented piece by piece, which is also why I'm sharing the source rather than keeping it as a black box.

Contributions are very welcome. If you spot a place where the implementation deviates from the spec, or want to extend coverage (more encryption revisions, more filter types, more field types), please open an issue or PR — see Contributing below.

Contributing #

Issues and pull requests are welcome — please open an issue first for anything beyond a small fix, so we can discuss the approach. If you're proposing a change to the core PDF parsing/writing logic, a reference to the relevant section of the PDF Reference 1.7 in your PR description is very helpful for review.

License #

MIT — see LICENSE. Third-party code and dependencies are separately licensed; see THIRD_PARTY_NOTICES.md.