PDF → PNG → TeX · Equation-Aware Extraction
Each page is rasterized in your browser (no upload until the PNG is ready), then sent to a vision-capable LLM instructed to transcribe — not "fix" — the math. Output is raw LaTeX, ready for amsmath.
Drop a PDF here
or click to browse · math papers, textbooks, problem sets
Downloaded before the zip-builder fix? Those archives won't open. Upload the old .zip here to rebuild it as a valid archive with the same .tex contents.
No Fibs — Honest Limitations
Equation integrity is best-effort, not guaranteed. This lab uses a vision LLM to OCR rendered pages — it does not read the PDF's internal math objects (which are often flattened to glyphs with no recoverable structure). State-of-the-art PDF→TeX extraction routinely misidentifies structurally similar symbols (subscripts vs. exponents, ℓ vs. 1, ∂ vs. δ), drops alignment in multi-line derivations, and struggles with hand-set or older typefaces. Always verify extracted TeX against the source before using it. Illegible glyphs are marked [illegible] rather than silently guessed. No fibs.