Converting HWPX, DOCX, and Markdown without leaving your browser

Moving between HWPX, DOCX, and Markdown can preserve a document’s semantic structure, but it cannot reproduce every editor’s page layout exactly. This guide brings together the five conversion directions, the Markdown ZIP shape, table and image handling, safety limits, and the checks to make afterwards.

The five supported directions

The available paths are HWPX → DOCX, HWPX → Markdown, DOCX → Markdown, Markdown → DOCX, and Markdown → HWPX. Headings, paragraphs, lists, tables, and images move through a shared semantic model and are then written in structures the target format can express.

Supported paths by input and output
Input	Supported output	Status
HWPX	DOCX, Markdown ZIP	Beta
DOCX	Markdown ZIP	Stable
Markdown ZIP	DOCX	Stable
Markdown ZIP	HWPX	Beta

There is no direct DOCX → HWPX path. Converting through Markdown ZIP lets you inspect the intermediate result and its warnings instead of inventing matches between two editor-specific layout systems. The older .hwp binary HWP format is different from HWPX and is neither read nor written here.

What a Markdown ZIP looks like

Markdown input is a Markdown ZIP rather than a lone file. The archive must contain exactly one .md or .markdown document, and images are resolved at paths relative to that document. Exports to Markdown use the same shape so text and assets travel together.

report.md — the one document to convert
assets/chart.png — an image referenced as !Quarterly chart
assets/logo.svg — another image that can live under the same assets/ folder

If the archive contains two or more .md files, the tool stops instead of guessing which one is primary. Absolute paths, ../ paths that leave the folder, and paths carrying query strings are not read as package assets either.

GFM tables and safe HTML tables

A simple rectangular table is exported as a GFM pipe table. Structures that GFM cannot express, such as merged cells, use a restricted safe HTML table and carry a warning. On import, only allowed table elements and attributes are interpreted; scripts and arbitrary HTML are never executed.

Markdown representation by table structure
Source table	Markdown result	What to check
No merged rows or columns	GFM pipe table	Cell reading order
Merged rows or columns	Safe HTML table	Spans and target editor support
Unsupported nested object	Visible text or omission	The reported warning

The HTML fallback is not a promise of pixel-perfect appearance. It is a way to retain the table’s semantic relationships, so check how it renders in the Markdown editor where the result will be used.

Image names and images that are not fetched

Images extracted from a document receive collision-free names based on nearby text and their original names, then go under assets/. When importing a Markdown ZIP, the tool uses only relative-path images that actually exist in the archive and whose bytes match a supported image format.

Duplicate names gain a number, so one image never silently overwrites another.
A missing or unsupported image keeps its alternative text and produces a warning.
A remote image at an https: address or a data: URL is not fetched by the browser; only its visible text remains.

Declining to fetch remote images protects privacy and makes the conversion repeatable. Put every image to be converted under assets/ in the ZIP and reference it with a relative path.

Safety limits for files and archives

Everything runs inside the browser, but a compressed document can become far larger when opened. These limits keep a tab responsive and prevent an unexpected archive bomb from being expanded.

Document conversion input limits
Item	Limit	Behaviour
Files selected at once	5 files	Extra files are not accepted
Large-file notice begins	Above 50 MiB	You choose whether to continue
One input file	200 MiB	Larger files are rejected
Total expanded archive	1 GiB	Larger archives are rejected
Archive entries	10,000	More entries are rejected
Compression ratio	200:1	Over the limit is rejected once an entry exceeds 10 MiB

50 MiB is a notice, not a failure threshold. The 200 MiB input, 1 GiB expanded size, 10,000-entry, and 200:1 ratio limits are safety boundaries checked before package contents are read.

Read the warnings, then check the target editor

A result warning says which structure was simplified and what was retained instead. It never exposes raw parser details; it uses information you can act on, such as a file name or format. A successful conversion with warnings deserves a focused check at those locations.

Files stay on your device and run in a dedicated worker inside the current browser.
The three paths that read or write HWPX are Beta, so keep the original file.
Open the download in the target editor and check fonts, page breaks, merged tables, and image order.

The goal is to move meaning safely and make differences visible. Include a side-by-side check of the source and result before publishing or submitting the converted document.

Back to PrismMesh