Browser-only testing fixture

Safe sample PDF generator for testing PDF tools

Create a harmless test document instead of uploading a contract, tax form, medical record, or other real file. Every generated PDF has exact markers that make lost pages and broken conversions easy to spot.

Quick answer

Use a synthetic PDF when testing an unfamiliar converter. This generator creates 4–20 non-sensitive pages locally, including selectable text, a rotated page, transparent vector shapes, edge targets, and an image-only page for OCR. The document is generated in your browser and is never uploaded.

Nothing to upload

Build your test PDF

Generated on this device
Stress-test pages
Generated fixture contents
  • 6 uniquely numbered pages
  • Selectable text and exact reading-order markers
  • Page 2 rotated 90° clockwise
  • Page 3 with overlapping transparent vector shapes
  • Page 6 as one embedded image with no selectable text
  • Edge targets, a measurement ruler, and color swatches

Ready. No file has been created yet.

What each test page reveals

Fixture featureWhat to check after conversionTypical failure it exposes
Unique page markersEvery marker appears once, in numeric order.Missing, duplicated, shuffled, or incorrectly split pages.
Selectable textCopy READING ORDER 1 2 3 4 and search for the fixture marker.Flattened output, damaged fonts, or scrambled extraction order.
90° rotationThe marked page remains rotated unless you deliberately correct it.Rotation metadata ignored or applied twice.
TransparencyRed, blue, and yellow shapes overlap into visibly blended regions.Black boxes, opaque fills, or lost vector graphics.
Image-only pageNormal text extraction finds nothing; OCR should recover the visible marker.A tool claiming OCR when it only reads an existing text layer.
Edge targetsFour corner targets and the outer ruler remain visible.Unexpected cropping, page-size changes, or clipped margins.

Independent release validation

On July 10, 2026, the production generator code created a 10-page A4 fixture in Chrome using marker QA83B8. The 91,787-byte output was then checked outside the generating code with Poppler 26.05.0.

Validation checkMeasured result
PDF structureValid PDF 1.7; 10 pages; A4 page box measured 595.28 × 841.89 points; not encrypted.
Rotation metadataPage 2 reported 90°; pages 1, 3, and 10 reported 0°.
Selectable markersMarkers QA83B8-01-OF-10 through QA83B8-09-OF-10 were recovered by independent text extraction.
Image-only boundaryPage 10 contained zero extracted characters while its visible OCR target remained readable in the rendered PNG.
Visual renderingThe rotated layout, blended transparency regions, color swatches, rulers, and all four edge targets rendered without clipping or black boxes.

File size can vary slightly because PDF creation and modification timestamps change on each run. The structural expectations above remain the validation contract.

How to test an online PDF converter safely

  1. Generate a synthetic fixture.

    Choose the same page count and paper size as your real document, but do not include real names, addresses, signatures, or account information.

  2. Record what should survive.

    Note the page count, marker prefix, rotated page, transparency page, and image-only page before converting.

  3. Run the intended operation.

    Split, merge, rotate, compress, export, or OCR the harmless fixture first. A successful download is not enough.

  4. Inspect the output.

    Open it independently, count the pages, search the text layer, check page order and rotation, and zoom into the corner targets.

  5. Check privacy separately.

    A correct output does not prove local processing. Use the browser network panel and the two-minute no-upload test before trusting a sensitive document.

Which tools can this fixture test?

Split and extract

Page markers reveal missing, duplicated, or incorrectly ranged output files.

Rotate and organize

The deliberately rotated page and numeric sequence expose orientation and ordering errors.

Compress

Compare selectable text, vector edges, transparency, color swatches, and the raster page after compression.

OCR and text extraction

The final page has visible letters but no text objects, while the other pages contain a real text layer.

Page size and boxes

Inspect the generated paper size, rotation, MediaBox, CropBox, TrimBox, BleedBox, and ArtBox without uploading the fixture.

What this fixture does not test

It does not contain forms, annotations, digital signatures, embedded files, video, unusual fonts, right-to-left scripts, encryption, or intentionally malformed PDF structures. It is designed for common page-processing and conversion workflows, not as a complete PDF specification conformance suite.

The PDF-LIB code needed to create the file is fetched when you use the generator. The finished document bytes are assembled locally and downloaded directly by your browser. No generated PDF is sent to GoPDFConverter, analytics, or a conversion server.

Frequently asked questions

Can I make a 10-page sample PDF?
Yes. Set Total pages to 10. Every page receives a unique marker; page 2 tests rotation, page 3 tests transparency, and page 10 is image-only when all stress pages are enabled.
Is the generated sample PDF safe?
The default document contains only synthetic shapes, labels, and markers. Keep it safe by leaving the generic marker in place and never typing personal or confidential information into the marker field.
Why include both selectable text and an image-only page?
Text extraction and OCR are different jobs. A normal extractor should recover the selectable pages but not the image-only page. An OCR tool should recognize the visible text in that raster page.