indd-utils

Reads INDD files, the native document format of Adobe® InDesign®, without InDesign, and converts them to IDML. The crate and the command are both named indd.

Adobe and InDesign are trademarks of Adobe Inc. This project is not affiliated with or endorsed by Adobe.

Status

The converter turns INDD and INDT files into IDML packages. It writes the document structure, pages, page items, placed graphics, story text with its formatting, footnotes, endnotes, notes and tracked changes, styles, swatches, tables, hyperlinks and cross-references, fonts and XML structure. Its output validates against the IDML RelaxNG schemas, except for endnote markup and form fields anchored in text, which it writes as InDesign does and which the published schema rejects (docs/measurement.md). Some attributes are still missing (see Current numbers), so the output is not yet the same as InDesign's own IDML export.

The format is undocumented. Every fact the converter relies on was found by comparing INDD files with IDML files exported from the same documents, and is recorded with its evidence in docs/format/. CLEANROOM.md describes the method and its limits.

Use

Build from source with a recent stable Rust:

cargo build --release
./target/release/indd convert in.indd out.idml

Command	Output
indd convert <in.indd> <out.idml>	IDML package. Problems that do not stop the conversion are printed to stderr as warning: ….
indd info <file.indd>...	Header, master page and container summary
indd objects <file.indd>	One line per database object: UID, class, length, first bytes
indd object <file.indd> <uid>	One object's bytes, to stdout
indd dump [--full] <file.indd> <uid>...	The objects' chunks in hex
indd uids <file.indd>	One line per object: UID and class, without reading object data
indd xmp <file.indd>	The document's XMP packet
indd audit [--tsv] <file.indd>	What the converter does not read in the document

INDT templates are read the same way as INDD files.

As a library:

let conversion = indd::convert_file("brochure.indd")?;
std::fs::write("brochure.idml", &conversion.idml)?;
for warning in &conversion.warnings {
    eprintln!("warning: {warning}");
}

indd::convert(bytes, name) converts bytes in memory and indd::convert_into(bytes, name, writer) writes the package to a writer. A conversion keeps no global state, so several documents can be converted on different threads at once.

The library also builds for WebAssembly (wasm32-unknown-unknown): indd::convert needs no file system or other host function, and gives the same package as a native build (tools/check_wasm.sh).

Supported versions

Versions	Distinct samples	State
InDesign 1.x	8	Not supported (different container); reported as such.
InDesign 3.0 and 4.0	67	Convert. Some object layouts differ from later versions (big-endian.md). Document preferences, object style text frame settings, and the page number style (3.0) are left out with warnings.
InDesign 5.0 and 6.0	181	Convert. Document preferences are left out with a warning.
InDesign 7.0–7.5	566	Convert. Document preferences are left out with a warning.
InDesign 8.0–21.6	3,438	Convert. Tables from 9.2 are not read.

Files in both byte orders convert. The byte order applies to object data only; the database pages are little-endian in every file.

What converts

Each area below links to the document that holds the evidence.

Area	Converted
Document structure	Spreads, master spreads, layers, sections (start, length, continued numbering, Arabic and lower-case Roman page numbers), the story list. objects.md
Pages and masters	Page bounds, transforms, applied master, margins and columns, page names, ruler guides, master names and prefixes, primary text frames. objects.md
Page items and graphics	Text frames (with threading), rectangles, ovals, polygons, graphic lines and groups: transforms, paths, fill and stroke, corners, applied object style, text wrap, frame fitting, anchored objects, text on a path. EPS text. Placed images, PDF, EPS, SVG and InDesign pages with their bounds, clipping path settings and links; graphics pasted without a link keep their data. Transparency: blending, drop shadow, inner shadow, outer and inner glow and gradient feather where the samples tell the attributes apart. objects.md, attributes.md, transparency.md
Text and typography	Story text, paragraph and character ranges with their local formatting (about 200 text attributes, among them font, size, leading, tracking, indents, spacing, hyphenation, keeps, tabs, rules, underline and strikethrough, shading, borders, bullets and numbering lists, nested styles, span columns, directions, OpenType and CJK settings, languages), story settings, text frame settings, text variables and their instances, footnotes and endnotes with their options, notes, tracked changes, ruby (text, type, alignment, position, offsets, scaling and spacing), kenten kind and size, warichu, tate-chu-yoko. attributes.md, text-variables.md
Styles	Paragraph, character, object, cell and table styles, with style groups, BasedOn and NextStyle. objects.md, tables.md
Colours and swatches	Process, spot and registration colours, tints, gradients, inks, colour groups, the None swatch. objects.md
Tables	Tables in stories and in cells: rows, columns, headers and footers, cells with spans, cell text, cell and table formatting, applied styles. tables.md
Links and cross-references	Hyperlinks with text, paragraph and page item sources; text, page, URL and external page destinations; bookmarks; cross-reference sources and formats. hyperlinks.md, cross-references.md
Fonts	Font families and fonts with their names (including built-in and native-script names), styles, types, PostScript names and Typekit IDs; missing fonts; composite fonts and their entries. fonts.md
CJK	Kinsoku and mojikumi tables (with the spacing of custom mojikumi tables), composite fonts, grid alignment of paragraphs, layout grids of pages, named grids, frame grids and their stories. objects.md, fonts.md
XML	Tags, the XML structure with elements placed in story text, the backing story. xml.md
Index	The index with its topics, page references in the text, index options and header groups, index sort groups. index.md, preferences.md
Document lists	TOC styles, trap presets, numbering lists, the language list, named grids, document users, the document label. objects.md, idml-values.md
Preferences	Document setup, view, guides, grids, text, margins, pasteboard, colour settings, default styles and grids, print and booklet print settings, chapter numbering, dictionary, EPUB version and identifier, from the INDD. The other preference values, and the values every InDesign export has on the root styles, are written as observed in all reference IDML files. preferences.md, idml-values.md

What does not convert

Reasons:

• Not in public samples: no sample with an IDML has the feature, so there is nothing to compare against.
• Not provable: the samples have the feature, but every sample has the same value, so the fields cannot be told apart.
• Not started: the samples have it, and no one has worked on it yet.

Feature	Reason
Conditional text, buttons and forms	Not in public samples
Ruby and kenten fonts, colours, weights and Y scales; kenten placement and characters; warichu characters before and after a break	Not provable (every sample has the default value)
XML attributes, comments, processing instructions, DTDs; XML elements that cross paragraphs	Not in public samples (left out with a warning)
Hyperlink sources that cross a style range	Not in public samples with an IDML (left out with a warning)
Most transparency effects (bevel, satin, feathers other than gradient, some glow values), drop shadow offsets	Not provable
Which table border attribute is which side (all four sides are equal in every sample)	Not provable
Font Status, variable font design axes	Not provable (Status depends on the exporting machine)
Many attributes with one value in nearly every sample (Visible, Locked, Name of page items, …)	Not provable
Tables from InDesign 9.2	Not started (different layout)
Document preferences from InDesign 3.0–7.5, object style text frame settings from 4.0	Not started (different layouts)
Transparency of placed graphics and object styles	Not started
Index title in languages other than Japanese or Chinese without stored index options, endnote title without stored endnote options; BitmapPrinting of the print settings	Not provable (they follow the exporting application or computer)
QR codes	Not started

How fidelity is measured

The local test corpus holds INDD and INDT files made by other people for other purposes. It is not redistributed; its openly licensed part can be downloaded with tools/fetch_corpus.py. 803 distinct files come with an IDML that InDesign exported from the same document at the same major version. In 606 of these pairs the IDML shows the same save as the INDD (trustworthy pairs); the others were saved again after the export (docs/measurement.md).

• tools/compare.py converts each paired file and compares the output with InDesign's IDML. Elements are matched by Self (the INDD UID). The headline is value coverage: the share of the reference's values that the output reproduces.
• tools/compare.py --all also converts the files without a usable IDML.
• With --schemas and --jing, every output is validated against the IDML RelaxNG schemas with Jing (tools/validate.sh).
• cargo test runs unit tests, smoke tests on the open-licensed samples listed in tests/fixtures/manifest.json (fetch them first with python3 -I tools/fetch_fixtures.py) and, if corpus/ exists, tests over the corpus.

Current numbers

From tools/compare.py --all with schema validation, without the privately held samples:

Measure	Result
Conversion failures	0 of 1,460 paired files; 0 of 4,303 other files
Rejected files	0 of 1,460 paired files; 22 of 4,303 other files (11 not INDD files, 8 InDesign 1.x files, 2 truncated, 1 without an object database)
Schema validation failures	0 paired files, 12 other files
Value coverage, trustworthy pairs	98.96 % (28,546,513 of 28,847,647 values)
Value coverage, all pairs	97.77 %
Story text, trustworthy pairs	37,129 of 37,150 stories exact, 3 differ, 18 missing

The biggest remaining gaps are elements the converter does not write yet (hyperlinks in some documents, page references, footnote options of text frames, graphic layers) and attributes that have one value in nearly every sample, so that their INDD fields cannot be located. compare.py lists them, and writes per-key tables to target/compare/.

Development

• CONTRIBUTING.md: the clean-room rules, the checks to run, and how to add an attribute, element or constant value end to end. Read CLEANROOM.md first.

• ARCHITECTURE.md: layers, data flow, module map, and how the generated value files work.

• python3 -I tools/fetch_fixtures.py downloads the test fixtures.

• cargo test, cargo clippy --all-targets, cargo fmt.

• python3 -I tools/compare.py [--all] [--detail TAG --show N] [--schemas DIR --jing DIR] after every change. The schemas and Jing are not part of the repository; see tools/validate.sh.

• python3 -I tools/diff_outputs.py [OLD [NEW]] compares the output of two revisions over the corpus; a refactoring must change none.

• python3 -I tools/readme_numbers.py --schemas DIR --jing DIR runs the full measurement and updates Current numbers.

• python3 -I tools/inventory.py corpus/ lists each sample's InDesign version and whether it has a matching IDML file.

Licence

Dual-licensed under MIT or Apache-2.0, at your option. Copyright (c) 2026 aiongg and the indd-utils contributors. The repository contains no sample files. The test fixtures are downloaded from their sources and keep their own licences; see tests/fixtures/README.md.