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
- 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 feature | What to check after conversion | Typical failure it exposes |
|---|---|---|
| Unique page markers | Every marker appears once, in numeric order. | Missing, duplicated, shuffled, or incorrectly split pages. |
| Selectable text | Copy READING ORDER 1 2 3 4 and search for the fixture marker. | Flattened output, damaged fonts, or scrambled extraction order. |
| 90° rotation | The marked page remains rotated unless you deliberately correct it. | Rotation metadata ignored or applied twice. |
| Transparency | Red, blue, and yellow shapes overlap into visibly blended regions. | Black boxes, opaque fills, or lost vector graphics. |
| Image-only page | Normal text extraction finds nothing; OCR should recover the visible marker. | A tool claiming OCR when it only reads an existing text layer. |
| Edge targets | Four 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 check | Measured result |
|---|---|
| PDF structure | Valid PDF 1.7; 10 pages; A4 page box measured 595.28 × 841.89 points; not encrypted. |
| Rotation metadata | Page 2 reported 90°; pages 1, 3, and 10 reported 0°. |
| Selectable markers | Markers QA83B8-01-OF-10 through QA83B8-09-OF-10 were recovered by independent text extraction. |
| Image-only boundary | Page 10 contained zero extracted characters while its visible OCR target remained readable in the rendered PNG. |
| Visual rendering | The 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
- 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.
- Record what should survive.
Note the page count, marker prefix, rotated page, transparency page, and image-only page before converting.
- Run the intended operation.
Split, merge, rotate, compress, export, or OCR the harmless fixture first. A successful download is not enough.
- Inspect the output.
Open it independently, count the pages, search the text layer, check page order and rotation, and zoom into the corner targets.
- 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.