Marker
Converts PDF documents to markdown and JSON with high accuracy.
Tested on LinuxKnown vulnerabilities
Use it
Before you install: known vulnerabilities
pickfit found 38 known high or critical vulnerabilities in the packages a fresh install brings (OSV.dev, 5 Nov 2026). Weigh this before you install it.
pypi:pillow 10.4.0Pillow: Heap buffer overflow in _tiffWriteProc when saving TIFF to BytesIO (GHSA-332p-7hrc-92wf, critical, fixed in 13.0.0)pypi:filelock 3.32.7filelock ReadWriteLock: SQLite reopens the lock database by name, so a planted symlink or journal file makes the lock user's process overwrite or unlink its own files (GHSA-v4x8-3wr5-3wq8, high, fixed in 4.2.1)
See 8 more
pypi:pillow 10.4.0— GHSA-5x8w-8jj5-q52h (critical), fixed in 13.0.0pypi:pillow 10.4.0— GHSA-8g5g-8vmr-9782 (critical), fixed in 13.0.0pypi:pillow 10.4.0— GHSA-w889-685w-c8fh (critical), fixed in 13.0.0pypi:pillow 10.4.0— GHSA-22hg-r758-rrw9 (high), fixed in 13.0.0pypi:pillow 10.4.0— GHSA-27h8-gf7j-v5hc (high), fixed in 13.0.0pypi:pillow 10.4.0— GHSA-3j72-p6qg-6xp7 (high), fixed in 13.0.0pypi:pillow 10.4.0— GHSA-45hq-cxwh-f6vc (high), fixed in 12.3.0pypi:pillow 10.4.0— GHSA-58r4-6rv3-gfh8 (high), fixed in 13.0.0
28 more findings not listed here.
Linux x86_64
You need first
- Runtimes:
python 3.10+
Install marker-pdf via uv tool
uv tool install marker-pdf
Check that it works
marker_single --help 2>&1The output should contain “Show this message and exit.”.
First thing to try
cat > /tmp/makepdf.py << 'PY'
# minimal PDF writer with correct xref, stdlib only
def make_pdf(path, lines):
objs = []
objs.append("<< /Type /Catalog /Pages 2 0 R >>")
objs.append("<< /Type /Pages /Kids [3 0 R] /Count 1 >>")
objs.append("<< /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792] /Contents 4 0 R /Resources << /Font << /F1 5 0 R >> >> >>")
# content stream
y = 750
stream_lines = ["BT", "/F1 12 Tf"]
for text in lines:
esc = text.replace("\\", "\\\\").replace("(", "\\(").replace(")", "\\)")
stream_lines.append(f"50 {y} Td ({esc}) Tj")
# reset position trick: move back
# use Td relative; simpler: use Tm absolute per line
# redo with Tm:
stream_lines.pop()
stream_lines.append(f"BT /F1 12 Tf 50 {y} Td ({esc}) Tj ET")
y -= 20
# Actually wrap: we already have BT open; simplify: join
stream = "\n".join(stream_lines[2:]) if len(stream_lines)>2 else ""
# rebuild cleanly
parts = []
for i, text in enumerate(lines):
esc = text.replace("\\", "\\\\").replace("(", "\\(").replace(")", "\\)")
yy = 750 - i*20
parts.append(f"BT /F1 12 Tf 50 {yy} Td ({esc}) Tj ET")
stream = "\n".join(parts) + "\n"
objs.append(f"<< /Length {len(stream.encode())} >>\nstream\n{stream}endstream")
objs.append("<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>")
out = bytearray()
out += b"%PDF-1.4\n"
offsets = [0]
for i, body in enumerate(objs, start=1):
offsets.append(len(out))
out += f"{i} 0 obj\n{body}\nendobj\n".encode()
xref_pos = len(out)
out += f"xref\n0 {len(objs)+1}\n0000000000 65535 f \n".encode()
for off in offsets[1:]:
out += f"{off:010d} 00000 n \n".encode()
out += f"trailer\n<< /Size {len(objs)+1} /Root 1 0 R >>\nstartxref\n{xref_pos}\n%%EOF".encode()
open(path, "wb").write(out)
print(f"wrote {path} {len(out)} bytes")
lines = [
"Hello Marker Test Document",
"This is a paragraph for conversion testing.",
"Name Age City",
"Alice 30 Paris",
"Bob 25 London",
"def hello(): print('hi')",
]
make_pdf("/tmp/sample.pdf", lines)
PY
python3 /tmp/makepdf.py && ls -l /tmp/sample.pdfmkdir -p /tmp/mout && marker_single /tmp/sample.pdf --output_dir /tmp/mout --disable_ocr 2>&1What it does
- Convert documents to markdown JSON chunks HTMLTested
- Convert PDF image PPTX DOCX XLSX HTML EPUBNot shown working
- Format tables forms equations and code blocksNot shown working
- Extract and save imagesNot tested
How far pickfit got
Linux x86_64
- Installs
- Starts
- 1 of 4 features work
- Works offline
Good for
- convert documents to markdown and JSON
- convert tables and equations to markdown
- extract tables as HTML blocks
An alternative to MinerU and docling.
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
Its main function works
- Replayed from scratch
- Works offline
- Checked on 4 Nov 2026
- Version 2.1.0
Differences the test machine hit
- README: marker_single /path/to/file.pdf
In the test: pass --disable_ocr on this CPU-only box so no VLM/llama.cpp server is started; conversion then uses pure text-layer extraction - README: log reports saved format-specific output
In the test: correct sample.json/sample.html/chunks files are written despite the log line always saying Saved markdown