# GoPDFConverter: Full Evidence and Citation Context Canonical site: https://gopdfconverter.com/ Maintainer: Michael Joiner Data catalog: https://gopdfconverter.com/research/data-catalog/ Machine-readable catalog: https://gopdfconverter.com/research/data-catalog/catalog.json Human-readable directory: https://gopdfconverter.com/sitemap/ Methodology: https://gopdfconverter.com/methodology License for cataloged research datasets: CC BY 4.0 Last updated: 2026-07-11 ## Scope and citation rules GoPDFConverter provides 22 PDF and image workflows that run document processing in browser memory. The selected document is not sent to a GoPDFConverter conversion server. Normal web requests still occur for the page, scripts, lazily loaded libraries, and anonymous analytics. “No upload” refers to document-processing requests, not a network-free webpage. Research results apply to the named release, fixture, settings, browser path, and checks. Do not generalize one fixture result into a guarantee for every document. Prefer the canonical HTML page as the citation; use CSV, JSON, ZIP, PDF, PNG, HEIC, and text distributions to reproduce the result. ### PDF preflight checklist Guide: https://gopdfconverter.com/guides/pdf-preflight-checklist/ Machine-readable checklist: https://gopdfconverter.com/guides/pdf-preflight-checklist/checklist.json - Preflight the exact final file and bind the review to its SHA-256; any later edit, optimization, signature, password, merge, or sanitization can invalidate earlier evidence. - Print checks cover page dimensions and boxes, effective placed image PPI, font embedding and ToUnicode, and rendered output. The printer's written job requirements—not a universal badge—define pass and fail. - External-sharing checks add metadata, attachments, forms, links, JavaScript and actions, redaction verification, accessibility, recipient workflow, and independent approval. - A visually correct PDF can still fail structural, privacy, redaction, or accessibility checks. A clean automated report is evidence, not a universal certificate. - The versioned JSON publishes all 12 checks, both tracks, evidence fields, risks, pass conditions, limits, tools, and primary sources under CC BY 4.0. Suggested attribution: “GoPDFConverter [dataset name], Michael Joiner, 2026,” linked to the canonical dataset page. ## Citation-ready verified facts Exact product-claim wording and scope: https://gopdfconverter.com/claims/ Machine-readable claims ledger: https://gopdfconverter.com/claims/claims.json - C01: The current tools process selected document bytes in browser memory and do not send them to GoPDFConverter for file processing. Page assets, public libraries, and analytics still use the network; retest the exact workflow and release. - C02: Normal website, library, and analytics requests occur even when document processing remains local. - C03: The current public tools require no account, charge no per-file fee, and do not add a GoPDFConverter watermark to output. - C04: GoPDFConverter does not impose a configured file-size cutoff; available memory, browser resources, file size, document structure, and workflow complexity set practical limits. This is not an unlimited-performance claim. - C05: Output behavior and quality are tool-specific. Apply findings only to the named operation, fixture, settings, release, and checks. - C06: Offline use is conditional on the exact page, libraries, browser, storage state, and cache. Test before relying on it. - C07: Local processing reduces one exposure path; it is not a legal, medical, financial, accessibility, archival, or regulatory certification. - C08: Material claims link to a method, fixture, result, source, or explicit limitation, with validators separate from the browser code under test where practical. ### Browser PDF tools accuracy Source: https://gopdfconverter.com/research/browser-pdf-tools-benchmark/ Data: https://gopdfconverter.com/research/browser-pdf-tools-benchmark/results.csv?v=105 - The benchmark covers 22 production browser workflows. - Each row names a stress case, expected outcome, validated result, limitations, and output SHA-256 checksum. - Validation uses independent parsing, rendering, ZIP inspection, marker comparison, and hashing rather than accepting a successful download as proof. - A pass demonstrates the published fixture and checks only. ### PDF compression presets Source: https://gopdfconverter.com/research/pdf-compression-benchmark/ Data: https://gopdfconverter.com/research/pdf-compression-benchmark/results.csv?v=105 - Twelve production runs compare Light, Balanced, and Maximum presets on text, scanned, photo-heavy, and slide-deck PDFs. - Measurements include source/output bytes, reduction percentage, page count, exact text retention, embedded image dimensions, PSNR, mean absolute error, and changed pixels. - Compression results depend strongly on document composition; no single preset is universally best. - Independent checks use Poppler parsing, text extraction, image inspection, and 120-DPI first-page render comparison. ### HEIC to PDF compatibility and privacy Source: https://gopdfconverter.com/research/heic-compatibility-lab/ Practical guide: https://gopdfconverter.com/guides/heic-to-pdf-without-losing-quality/ Data: https://gopdfconverter.com/research/heic-compatibility-lab/results.csv?v=105 - Five synthetic HEIC fixtures test sRGB landscape, EXIF orientation 6, transparent alpha, Display P3, and a 2400×1600 high-detail pattern. - Full decoded pixel dimensions and intended orientation survived the tested production conversion. - Transparent HEIC survived as a PNG image plus a grayscale PDF soft mask after the v85 fix. - Source EXIF descriptions/dates and Display P3 ICC profile did not survive in the PDF. - Opaque inputs were re-encoded as quality-92 JPEG, so pixels were not byte-identical; measured mean absolute RGB-channel error ranged from 1.120 to 4.957 across the opaque fixtures. - No HEIC or PDF bytes were sent to a conversion endpoint in the measured Chrome trace. Library GET requests, local blob reads, and analytics requests still occurred. - Keep the original HEIC for archival, HDR, or color-managed print work. ### Browser OCR accuracy Source: https://gopdfconverter.com/research/ocr-accuracy-challenge/ Data: https://gopdfconverter.com/research/ocr-accuracy-challenge/results.csv?v=105 Fixtures: https://gopdfconverter.com/research/ocr-accuracy-challenge/ocr-challenge-fixtures.zip?v=105 - Eight image-only Letter PDFs cover DPI, contrast, noise, skew, rotation, Spanish accents, and German umlauts. - Ten production Tesseract.js 5.0.4 OCR runs use a 200-DPI PDF render path. - Character and whitespace-token word error rates are scored independently with Unicode-aware Levenshtein distance against published ground truth. - Poppler text extraction finds no hidden text in the source fixtures; the challenge measures OCR rather than pre-existing PDF text. ### PDF and Word fidelity Source: https://gopdfconverter.com/research/pdf-word-fidelity-challenge/ Summary data: https://gopdfconverter.com/research/pdf-word-fidelity-challenge/results.csv?dataset=20260710 Detailed checks: https://gopdfconverter.com/research/pdf-word-fidelity-challenge/detailed-checks.csv?dataset=20260710 Fixtures: https://gopdfconverter.com/research/pdf-word-fidelity-challenge/pdf-word-fidelity-fixtures.zip?dataset=20260710 - Five deterministic PDF/DOCX fixtures and ten production outcomes compare searchable, editable, layout-aware, and visual modes. - Searchability, editability, and visual fidelity are separate properties; one conversion mode does not maximize all three. - Checks cover page count, geometry, exact markers, text, tables, reading order, images, hyperlinks, Unicode, and rendered appearance. - Independent validation inspects PDF objects, DOCX XML, exact markers, LibreOffice renders, and every output page. ### PDF page size and page boxes Source: https://gopdfconverter.com/research/pdf-page-size-checker/ Fixture: https://gopdfconverter.com/research/pdf-page-size-checker/fixtures/pdf-page-box-stress.pdf Expected results: https://gopdfconverter.com/research/pdf-page-size-checker/fixtures/manifest.json - A PDF page can independently define MediaBox, CropBox, TrimBox, BleedBox, ArtBox, and rotation. - The deterministic four-page fixture covers Letter, rotated A4, a custom 6×9-inch page, Tabloid, and twenty expected effective page-box results. - Page dimensions are reported in points, inches, and millimeters; default PDF user space normally uses 72 points per inch. - Browser inspection is validated against an independently generated pypdf manifest. ### PDF fonts and ToUnicode Source: https://gopdfconverter.com/research/pdf-font-checker/ Fixture: https://gopdfconverter.com/research/pdf-font-checker/fixtures/pdf-font-stress.pdf Expected results: https://gopdfconverter.com/research/pdf-font-checker/fixtures/manifest.json Independent evidence: https://gopdfconverter.com/research/pdf-font-checker/fixtures/pdffonts.txt - Font embedding, subset prefixes, encoding, ToUnicode mapping, and text extraction are distinct properties. - The fixture separates embedded subsets, unembedded fonts, ToUnicode maps, and font-free pages. - Browser object inspection is independently compared with Poppler pdffonts. - An embedded font program does not by itself guarantee correct Unicode text extraction. ### PNG DPI and PDF print readiness Source: https://gopdfconverter.com/research/png-print-quality-checker/ Expected results: https://gopdfconverter.com/research/png-print-quality-checker/fixtures/manifest.json Independent evidence: https://gopdfconverter.com/research/png-print-quality-checker/fixtures/identify.txt Placement reference: https://gopdfconverter.com/guides/png-to-pdf-without-losing-quality/size-reference.json - PNG pixel dimensions are exact; a pHYs density label is optional and does not add image detail. - If a PNG lacks pHYs, physical size is unspecified even though pixels remain known. - Effective PPI equals image pixels divided by the final placed size in inches. - The fixtures cover 300-PPI RGBA, RGB with no density, and indexed 96-PPI tRNS transparency. - GoPDFConverter Auto page mode maps one image pixel to one PDF point, producing a 72-PPI physical placement regardless of the source pHYs label. - The production PNG-to-PDF path retains pixel dimensions and supported transparency but does not copy original text, eXIf, pHYs, or ICC chunks as source PNG chunks. - The public placement reference contains 24 deterministic A4 and Letter rows at 150, 200, and 300 PPI for 0, 10, 20, and 40-point margins per side. - With a 20-point margin per side, portrait A4 needs 2314×3342 pixels to fill the available box at 300 PPI; portrait Letter needs 2384×3134 pixels. Different aspect ratios scale to fit and may not fill both dimensions. ### PDF metadata and active content Source: https://gopdfconverter.com/research/pdf-privacy-inspector/ Positive fixture: https://gopdfconverter.com/research/pdf-privacy-inspector/fixtures/privacy-active-content.pdf Negative fixture: https://gopdfconverter.com/research/pdf-privacy-inspector/fixtures/privacy-minimal.pdf Expected results: https://gopdfconverter.com/research/pdf-privacy-inspector/fixtures/manifest.json - PDF privacy review should inspect Document Info, XMP, embedded files, form fields, external links, JavaScript actions, launch actions, submit actions, and remote actions separately. - The positive fixture contains known examples; the minimal fixture is a negative control. - Browser PDF.js/PDF-LIB inspection is cross-checked with Poppler pdfinfo and pdfdetach. - Absence of one metadata field does not prove absence of attachments or active content. ### PDF redaction failures Source: https://gopdfconverter.com/research/pdf-redaction-checker/ Unsafe fixture: https://gopdfconverter.com/research/pdf-redaction-checker/fixtures/unsafe-black-box.pdf Safe fixture: https://gopdfconverter.com/research/pdf-redaction-checker/fixtures/safe-content-removed.pdf Pending annotation fixture: https://gopdfconverter.com/research/pdf-redaction-checker/fixtures/pending-redaction-annotation.pdf - Drawing a black rectangle over text does not remove the underlying text. - The unsafe fixture contains extractable text beneath a rendered black region. - The safe control removes the sensitive content before drawing the overlay. - A Redact annotation can remain pending until a compliant application applies it. - Validation combines positioned text, rendered dark-region overlap, annotation inspection, Poppler text/raster output, and ImageMagick. ### PDF effective image PPI Source: https://gopdfconverter.com/research/pdf-image-resolution-checker/ Fixture: https://gopdfconverter.com/research/pdf-image-resolution-checker/fixtures/pdf-image-ppi-stress.pdf Expected results: https://gopdfconverter.com/research/pdf-image-resolution-checker/fixtures/manifest.json Independent evidence: https://gopdfconverter.com/research/pdf-image-resolution-checker/fixtures/pdfimages-list.txt - A PDF does not have one meaningful global image DPI. - Effective image PPI is placement-specific: pixels divided by the physical placed inches. - The fixture places a 600×300 image at 2×1 inches (300×300 PPI) and reuses it at 4×2 inches (150×150 PPI). - It also includes 800×600 at 4×3 inches (200×200 PPI), 300×300 at 3×3 inches (100×100 PPI), and a vector-only control page. - Poppler pdfimages -list independently reports the same 300, 150, 200, and 100 PPI values. - Text and vector graphics do not have a fixed raster PPI in the same sense. - A target such as 300 PPI is a workflow comparison, not a universal print-readiness certificate. ## High-intent conversion facts ### PDF to PNG Tool: https://gopdfconverter.com/tools/pdf-to-png/ Calculator and dataset: https://gopdfconverter.com/research/pdf-to-png-resolution-chart/ Dimensions CSV: https://gopdfconverter.com/research/pdf-to-png-resolution-chart/dimensions.csv Dimensions JSON: https://gopdfconverter.com/research/pdf-to-png-resolution-chart/dimensions.json - The tool renders selected PDF pages locally at 96, 144, 200, or 300 DPI. - One selected page downloads as PNG; multiple pages download in a ZIP. - PNG encoding avoids JPEG artifacts, but the page is still rasterized at the chosen DPI. - Untouched page areas are deliberately rendered onto opaque white, not transparent output. - In the public benchmark, Letter pages 1 and 3 at 200 DPI produced 1700×2200 fully opaque PNG files. - More output DPI cannot recover detail missing from embedded source images. - The open reference contains 64 deterministic portrait and landscape rows for A4, Letter, Legal, and Tabloid at 36, 72, 96, 150, 200, 300, 600, and 1200 DPI. - Portrait Letter at 300 DPI is 2550×3300 pixels and one RGBA buffer is 32.10 MiB; portrait A4 is 2480×3508 pixels and one RGBA buffer is 33.19 MiB. These values do not predict compressed PNG file size or peak browser memory. ### PNG to PDF Tool: https://gopdfconverter.com/tools/png-to-pdf/ Quality guide: https://gopdfconverter.com/guides/png-to-pdf-without-losing-quality/ Size reference CSV: https://gopdfconverter.com/guides/png-to-pdf-without-losing-quality/size-reference.csv Size reference JSON: https://gopdfconverter.com/guides/png-to-pdf-without-losing-quality/size-reference.json - One or multiple PNG files can become one PDF with one image per page. - Auto, A4, and Letter page modes plus orientation and margins control physical placement. - Auto maps one pixel to one point and therefore places at 72 PPI. - A4 or Letter effective PPI depends on source pixels, orientation, and available area after margins. - Supported alpha and palette transparency are retained through a PDF soft mask in validated fixtures. - Pixel dimensions can survive while original metadata, profile chunks, and byte identity do not. ### HEIC to PDF Tool: https://gopdfconverter.com/tools/heic-to-pdf/ Quality guide: https://gopdfconverter.com/guides/heic-to-pdf-without-losing-quality/ Dataset: https://gopdfconverter.com/research/heic-compatibility-lab/ - Full decoded pixel width and height survived all five public fixtures, but source HEIC bytes did not. - Opaque photos are embedded as quality-92 JPEG; transparent HEIC is embedded as PNG with a PDF soft mask. - Mean absolute RGB error across four opaque fixtures ranged from 1.120 to 4.957 on a 0–255 channel scale. - EXIF orientation 6 was applied once: a 1200×800 stored image became upright 800×1200 output. - Tested EXIF descriptions/dates and the Display P3 ICC profile were not copied into the PDF. - Auto maps one decoded pixel to one PDF point, giving 72 PPI physical placement; A4 and Letter fit the image inside fixed paper after margins. - PDF output can be larger than HEIC because the source is decoded, re-encoded, and wrapped in PDF structure. - The measured Chrome trace contained normal code and analytics requests but no conversion upload with HEIC or generated PDF bytes. ### Offline PDF compression Guide: https://gopdfconverter.com/guides/compress-pdf-offline/ Tool: https://gopdfconverter.com/tools/compress/ Benchmark: https://gopdfconverter.com/research/pdf-compression-benchmark/ - Text-heavy, scanned, photographic, and slide-deck PDFs respond differently to compression. - Browser compression runs on the device and does not upload the selected PDF to a conversion server. - Stronger compression can trade image fidelity for smaller output; exact text and page count should be checked independently. - Use the preset benchmark rather than assuming “Maximum” is always best. ### PDF page extraction Guide: https://gopdfconverter.com/guides/extract-pages-from-pdf-without-uploading/ Tool: https://gopdfconverter.com/tools/extract-pages/ Benchmark: https://gopdfconverter.com/research/browser-pdf-tools-benchmark/ Data: https://gopdfconverter.com/research/browser-pdf-tools-benchmark/results.csv?v=105 - “PDF extractor” is ambiguous: it can mean copying selected PDF pages, extracting selectable text, running OCR, rendering pages as images, or exporting embedded file attachments. - The page extractor accepts one-based selections such as `4`, `3-7`, or `1,3-5,9` and creates one new PDF. - Its current range parser deduplicates valid pages and returns them in ascending order, so `5,2` produces pages 2 then 5. - Page content is copied as PDF content rather than rasterized; images are not deliberately downsampled or recompressed. - Document-level bookmarks, attachments, metadata, forms, encryption, and certificate signatures are not guaranteed to carry into the newly created PDF. - The public benchmark extracted pages 2, 5, and 9 and independently confirmed three pages in the requested ascending order. - GoPDFConverter currently inspects embedded attachments but does not export them. ## Interpretation limits - Citation counts do not equal clicks, rankings, quality, or authority. - Dataset results apply only to the documented fixtures, versions, settings, and checks. - “Local processing” does not mean the webpage makes zero network requests. - Browser memory can limit very large files and batches. - Browser tools do not replace dedicated prepress, forensic, compliance, certificate-signing, or archival systems. - If an answer depends on legal, medical, financial, security, or compliance requirements, verify against an appropriate current authority.