PDF Object Explorer

PDF Object Explorer is a browser-based inspection tool that displays the internal objects and resources inside a PDF. It lets you browse pages, indirect objects, fonts, images, content streams, annotations, forms, metadata and object references, and links those structures back to the rendered page — without modifying the original file.

A PDF object is a unit of data inside a PDF file. Objects can represent numbers, strings, arrays, dictionaries, streams, pages, fonts, images, annotations and other document resources. Many important objects are indirect objects identified by an object number and a generation number.

An indirect object has an object number and a generation number, which lets other objects refer to it. A reference written as 24 0 R points to object 24, generation 0. That indirection is what makes reuse possible — one image object can be drawn on every page.

The explorer lists every object it can parse, including objects stored inside compressed object streams. Encrypted data, unsupported filters or severely damaged structures may limit what can be displayed. Very large documents are capped at a stated number of objects, and the report tells you when a limit was reached.

Yes — this is the core of the tool. Click any text run, image, vector path, form XObject, annotation or form field on the rendered page and the inspector shows the object behind it: the font that drew the text, the XObject that holds the image pixels, the content stream that issued the instruction. Some vector graphics are drawing instructions inside a content stream rather than separate indirect objects, and the inspector says so when that is the case.

Yes. Content streams are decoded where the filter allows and shown as text, along with the operator profile for the page — how many operators in total, how many text operations, and which operators are most used. Raw bytes are also shown as a hex dump. There is no syntax highlighting or operator-by-operator timeline in this version.

Yes. Each font object shows its base name, subtype, whether the font program is embedded, whether it is a subset, its encoding, whether it carries a ToUnicode map, its descriptor, and the pages that actually use it. For deeper font diagnosis — glyph coverage, rendering risk and print readiness — use the dedicated PDF Font Inspector.

Yes. Every image XObject is listed with pixel dimensions, colour space, bits per component, compression filter, transparency and soft mask, stored size, and the pages it is drawn on. This version does not render image thumbnails in the list — you locate an image by clicking it on the page instead.

Yes. The explorer hashes the stored bytes of each image and reports images whose encoded data is identical, along with roughly how much file size the duplication costs. Note that an image shared correctly — one object drawn on twelve pages — costs nothing extra and is reported separately from an image stored twelve times over.

Yes. The Metadata section shows the document information dictionary — title, author, subject, keywords, creator, producer, creation and modification dates — plus the raw XMP metadata packet when the file carries one.

Yes. Links, comments, highlights, stamps, ink annotations, shapes and form widgets are listed with their object ID, page, subtype, author, bounds, flags and appearance stream, and each is drawn as a box over the rendered page.

Yes. AcroForm fields are shown with their full and partial names, field type, current and default values, flags, parent and child fields, the widgets that render them, and the page each appears on. Fields with no appearance stream are flagged, because their values can be invisible in some viewers.

The explorer counts and describes embedded-file objects and file specifications so you know they are there. It never opens, extracts or executes embedded content.

JavaScript lives in objects that the explorer lists like any other, so you can see that it is present and what references it. This version does not single out or pretty-print JavaScript source, and it never executes it.

No — it does not produce a safety verdict. It surfaces structures worth a second look, such as embedded files or unusual actions, but it is an inspection tool and is not a substitute for antivirus or dedicated malware-analysis software.

No. The explorer is strictly read-only. It parses your document and reports what it finds, and never rewrites, repairs, optimises, normalises or re-saves it. Embedded JavaScript, launch actions, attached files and rich media are never executed or opened.

No. The explorer diagnoses and explains structural problems — broken references, unparseable streams, a damaged cross-reference table — but it deliberately does not modify the original file. Repair would mean writing to your document, which this tool never does.

A text editor cannot reliably decode compressed streams, read objects packed inside object streams, interpret binary image data, or follow a cross-reference stream. It also cannot connect an object to the place it appears on the page. You will usually see some readable fragments surrounded by binary noise.

A PDF content stream is a sequence of instructions that tells a viewer how to draw text, images and vector graphics on a page. It references fonts, images and other resources that are stored elsewhere in the document and named in the page’s resource dictionary.

A filter defines how stream data is encoded or compressed. Common examples are FlateDecode, LZWDecode, DCTDecode (JPEG), JPXDecode (JPEG 2000), CCITTFaxDecode and JBIG2Decode. This tool decompresses the general-purpose filters; image codecs are reported and previewed as raw bytes rather than decoded to text.

An object stream is a compressed stream that packs many non-stream indirect objects together. PDF 1.5 introduced them to reduce file size, and they are a common reason raw text inspection fails — the objects simply are not there as readable text. The explorer reads inside them and marks which objects were stored that way.

The cross-reference table records where each indirect object is stored so a reader can jump straight to it instead of scanning the file. Newer PDFs use a compressed cross-reference stream instead of a plain text table. The explorer parses both, shows each entry’s byte offset and generation, and follows the /Prev chain so every revision stays visible.

The page tree is the hierarchy that organises pages. Properties such as resources, page size and rotation can be set on a parent node and inherited by the pages beneath it, which is why a page can appear to have no resources of its own and still render.

A resource dictionary maps the short names a content stream uses — /F1, /Im1 — to the actual font, image, pattern, shading, colour space or graphics-state objects. It is the lookup table that connects drawing instructions to the objects they draw with.

An XObject is a reusable object referenced from page content. The two common kinds are image XObjects, which hold pixels, and form XObjects, which hold a reusable set of drawing instructions.

A form XObject is a reusable fragment of page content — a logo, a stamp, a repeated header — with its own resources and its own coordinate system. It is not the same thing as a PDF form field. Content inside one is real page content, just stored a level down, which is why text can be hard to find until you look inside it.

The explorer counts and lists optional content groups (layers) so you can see they exist. Content can also be effectively hidden by clipping, transparency, off-page placement or invisible text rendering modes — the presence of an object never guarantees it is visible.

PDFs saved incrementally keep earlier versions of changed objects in the file. The explorer counts the revisions, lists each cross-reference section separately, and marks entries that a later revision superseded.

An incremental save appends the new and changed objects to the end of the file instead of rewriting it. The previous versions of those objects stay in the file. Several rounds of editing can therefore leave a document much larger than its visible content warrants.

Yes. Objects are reported with their stored and decoded sizes, images are sorted by size, duplicated image data is flagged, uncompressed streams are counted, unreferenced objects are listed, and the number of incremental revisions is shown. The optimisation report names the likely causes — it never changes the file.

Yes, when you supply the correct password and the encryption method is supported. The password is used only to open the document for this analysis and is not stored. The tool does not bypass or recover passwords.

The explorer lists signature fields and the signature objects behind them, including the signer name, reason, location and signing time where present. It does not perform cryptographic verification or evaluate certificate trust — that needs a dedicated signature-validation tool.

Not always. Structural parsing, standards conformance and visual rendering are related but separate concerns. A file can be structurally sound and still render differently across viewers — for example when a font is not embedded and a substitute is used instead.

Yes. You can export the full analysis as JSON, the object list as CSV, or the diagnostics as CSV. The exports contain the analysis only — they are not a copy of your document. A formatted PDF report is not available in this version.

Yes. Search accepts an object number and matches object types and subtypes, and results link straight to the object with its references in both directions.

Yes. Searching a font name returns the font object, and selecting it shows which pages use it and which objects reference it.

Yes. Text extracted from content streams is searchable, and a match takes you to that exact text run on the page with its box highlighted. Text drawn with a font that has no ToUnicode map cannot be mapped back to characters, so it is not searchable — the tool tells you when that is why.

No. Every object carries a plain-language explanation of what it is, why it exists and whether it affects editing, so designers, print professionals, QA teams, support staff and anyone curious about how PDFs work can use it without knowing PDF syntax.

No. Preflight checks a PDF against a defined standard or rule set and reports conformance. An object explorer helps you browse and understand the file’s internal structure. They complement each other — preflight tells you a rule was broken, an explorer helps you find the object responsible.

Yes. You can inspect supported PDFs online with no account and no watermark. The file-size and page limits are shown before processing, and your upload is removed after your session.

A direct object is written inline wherever it is used and has no identity of its own. An indirect object is stored separately with an object number and generation, appears in the cross-reference table, and is referenced elsewhere as “24 0 R” — which is what lets one font or image serve every page.

The trailer is the block at the end of the file that points to the catalog and to the cross-reference table. Readers start there rather than at the beginning, which is why a damaged trailer can break a file whose objects are all perfectly intact.

The catalog is the document root. It holds references to the page tree, the outlines, the AcroForm, the metadata stream and the optional-content configuration, so almost every other object is reachable from it.

An appearance stream is the form XObject that defines what an annotation actually looks like when drawn. When one is missing, viewers fall back to their own rendering or show nothing at all — which is the usual reason an annotation appears in one application and not another.

Because the file leaves something ambiguous and each renderer resolves it differently. The common causes are a missing appearance stream, a font that is not embedded, an unusual stream filter, or a transparency group — and each of those is a specific object you can open in the report.

The Info dictionary is the older key–value block holding title, author, producer and dates. XMP is an XML metadata stream that can carry far richer, standardised properties. A file can contain both, and they can disagree — the report shows each one with its object number.

An image XObject is a numbered object referenced by name from the page’s resource dictionary, so it appears in the images list with its own object number. An inline image is written directly into the content stream between BI and EI operators and has no object of its own.