What the benchmark contains
The current corpus contains 60 legally generated synthetic PDFs: 20 text documents, 20 vector documents, and 20 image-heavy documents. No private user files or scraped third-party documents are included.
Engine under test
qpdf-wasm-lossless. This report measures the lossless baseline, not the optional image-downsampling path.
Verification
Every output is parsed again and checked for page count, geometry, extractable text, content structure, and final byte size.
Targets
Documents are assigned one of four exact decimal targets: 200 KB, 500 KB, 1 MB, or 2 MB.
How to reproduce it
From the FitPDF project directory, run the benchmark command below. It regenerates the validated JSON report, this public summary, and the Markdown report.
npm run benchmarkWhat these numbers do not prove
Synthetic files make the run legal and repeatable, but they do not represent every real portfolio, scan, form, or PDF producer. Reaching an assigned target is not a promise that every user document will reach the same size. Competitor comparisons and the public production license for the lossy engine remain external approval gates.