Sygal converts PDF, Word, and Excel to Markdown with the same engine as MarkItDown, Microsoft’s free tool, and in our testing the result was identical byte for byte. What you pay with its €25 is everything that surrounds the conversion: no Python installation, a desktop interface, batch export, explicit errors, and some formatting improvements. Whether it’s worth it depends on who’s going to use it.
yoDEV Tested · October 17, 2026 We tested MarkItDown 0.1.8 and sygal-cli 0.6.1 on Linux with six Spanish documents and two extra formats. The results we report as ours come from that test.
What is Sygal and why does it use MarkItDown under the hood?
Sygal is a desktop app and CLI that converts PDF, Office files, images, web pages, and YouTube links to Markdown and JSON. It’s developed by Cyrill Semah, an independent developer from Paris. His own FAQ names the converter he uses: MarkItDown. Its npm package is described as a “local MarkItDown pipeline, OCR for images”.
That’s not a shortcut: it’s the model Microsoft itself recommends. The MarkItDown README excludes desktop apps, web frontends, and graphical interfaces from the repository. It asks anyone who wants to build one to publish it as a separate project, dependent on markitdown from PyPI. Sygal is exactly that: a MarkItDown GUI, and a MarkItDown alternative for those who don’t want to touch Python, built on the same engine.
If you haven’t used MarkItDown yet, our guide explains what it is and how to install it: MarkItDown: Microsoft’s Tool That Converts Any Document into Context for Your AI.
Do Sygal and MarkItDown give the same result when converting PDF, Word, and Excel to Markdown?
Yes, for common office formats the result is identical. We generated six Spanish documents, full of accents and tildes, and converted each one with both tools on Linux:
- a
.docxreport with title and table - a
.xlsxbudget with two sheets - a
.pptxpresentation with two slides, with table and speaker notes - a PDF with text, headings, table, and list
- an
.htmlpage with navigation, list, table, and footer - a scanned contract saved as an image-only PDF
| File | MarkItDown 0.1.8 | sygal-cli 0.6.1 |
|---|---|---|
| Word, Excel, PowerPoint, PDF with text, HTML | Correct conversion | Identical byte for byte to MarkItDown |
| Scanned PDF | Empty file, no warning | Stops with scanned_pdf_detected and suggests --pdf-ocr |
| XML (extra test) | Passes XML raw | Structured Markdown table |
The accents and tildes arrived intact in all files with both tools. Both also share the same quirks, as expected from the same engine. A Word table without a marked header row comes out with an empty header and the actual header as the first row. And in the PDF, headings come out as normal text lines, not as #.
So for converting a PDF, Word, or Excel to Markdown, you get MarkItDown’s conversion with either one. The differences appear at the edges.
What does Sygal give you that MarkItDown doesn’t?
You don’t need Python. It’s Sygal’s strongest point, and we verified it. On Linux, npm installed a native engine that includes its own Python 3.12 runtime along with MarkItDown. The machine only needs Node.js 20 or higher. The tradeoff is size: the CLI took about 393 MB on disk.
A graphical interface. The desktop app adds a preview of the original alongside the result, editing before export, and a local library with full-text search. Sygal also presents it with drag-and-drop and with screenshots pasted via Cmd+V. It’s what you give a colleague who’s never going to open a terminal. (We didn’t test the desktop app: we describe it based on sygal.app.)
Errors you can see. With the scanned PDF, MarkItDown returned an empty file and exited without error. Sygal refused, named the problem, and told us which option to use. In a batch of hundreds of files, that difference matters.
A report designed for agents. With --json, Sygal reports for each file:
- the pipeline it used
- whether anything went out to the network (
networkUsed) - the token count of the result
- stable error codes
Its CLI page presents it for agent pipelines like Hermes and OpenClaw.
Some formatting improvements. Its changelog lists additional processing over MarkItDown. Our XML test confirmed one: Sygal produced a clean table where MarkItDown left the markup raw. The changelog also promises to rebuild tables in RTF, email, and TXT files, but a space-aligned TXT table came out unchanged in our test.
Extras. The CLI also offers --translate and a deferred OCR mode (--ocr defer, then sygal inject) for agents that can read images on their own.
Does Sygal work on Linux?
Yes for the command line, no for the desktop app. The CLI installs with a single command:
npm install -g sygal-cli
It includes a native binary for Linux x64 that requires glibc 2.28 or higher (Ubuntu 18.04+, Debian 10+, RHEL 8+). It doesn’t work on Alpine or musl images. We ran all our tests on it.
The desktop app is only listed for macOS 12+ and Windows 10+. One more limit: on macOS, the CLI’s native engine is only for Apple Silicon, so Macs with Intel are left out.
MarkItDown works on any system with Python 3.10 or higher, including Linux.
Can you convert a scanned PDF to text with Sygal or MarkItDown?
Yes, with OCR, and on neither one is the OCR local. Sygal makes it clear: document conversion happens on your machine, but OCR for images and scanned pages goes to the provider you choose, with your own API key (BYOK OCR). The options are OpenAI, Claude, Mistral, NVIDIA NIM, Gemini, or a custom provider. The site details exactly what can leave the machine: an image or scan you send for OCR, and a URL you decide to import.
MarkItDown follows the same path. Its markitdown-ocr plugin adds OCR to PDF, DOCX, PPTX, and XLSX files via a vision model you provide. In OCR there’s parity, not a weakness. If your scanned documents can’t leave the company, neither of these solves it out of the box: you need a local OCR model. We analyzed one in Chandra 2: OCR for the Documents That Break Everything Else. (We didn’t test OCR: it requires an API key from a provider and outgoing network access.)
Free MarkItDown or paid Sygal? The comparison
| MarkItDown | Sygal | |
|---|---|---|
| Price | Free, MIT license | €25 one-time payment, launch offer (regular price €29), 2 seats with CLI included; CLI only €22, 1 seat* |
| Engine | Own | MarkItDown (included) |
| Requires Python | Yes (3.10+) | No (includes 3.12) |
| Installation | pip install 'markitdown[all]' |
App download, or npm install -g sygal-cli |
| Linux | Yes | CLI yes; desktop app no |
| Graphical interface | No | Yes (macOS, Windows) |
| Scanned PDF without OCR | Empty result, no warning | Explicit error |
| OCR | Plugin markitdown-ocr, with your model |
Built-in, BYOK |
| Extensibility | Plugins, markitdown-mcp server |
Not indicated |
| Version | 0.1.8 | 0.6.1 |
| Trial | Not applicable | 14 days, no account; refund within 14 days |
*Prices published on sygal.app as of October 17, 2026. The launch offer does not indicate an end date.
Both are still below version 1.0. MarkItDown has been in 0.x for nearly two years. Sygal has released 19 versions since July 2026.
How do you test MarkItDown and Sygal with your own files?
Install each one, convert the same folder, and compare the results with diff.
pip install 'markitdown[all]'
markitdown informe.docx -o informe.md
npm install -g sygal-cli
sygal convert carpeta/ --format both -o ./salida
Testing Sygal starts when you install it. It doesn’t ask for an account, credit card, or network activation; sygal license status shows the days remaining. If your documents are Office, PDF with text, or HTML, expect them to match.
Is it worth paying €25 for Sygal?
Yes, if:
- you use macOS or Windows and want a graphical interface
- you’re going to hand off document conversion to colleagues who don’t use the terminal
- you process large batches and prefer explicit errors and a JSON report to silent empty files
No, if:
- you’re already comfortable with
pip: conversions are the same, free - you only work on Linux and wanted the app: there Sygal is a CLI, and MarkItDown already is one
- you need plugins or an MCP server: MarkItDown has them, Sygal doesn’t
- your OCR documents can’t leave your infrastructure: neither solves it out of the box
yoDEV has no relationship with Sygal. We tested it with the free trial version.
What about you? What kind of document do you find hardest to convert to Markdown for your AI: scanned PDFs, Excel, or something more unusual?
- MarkItDown
- A paid app or service
- I pass the file directly to the model
- I prefer another option (tell us which)
Comment below, or if this article reached you by email, reply directly to the email: your reply gets published here.
Related: MarkItDown: Microsoft’s Tool That Converts Any Document into Context for Your AI