A PDF content stream is syntactically simple. Each instruction consists of zero or more operands followed by an operator:

<code>BT</code> <code> /Helvetica 12 Tf</code> <code> 100 700 Td</code> <code> (Hello, World!) Tj</code> <code>ET</code>

In this example: <code>BT</code> (Begin Text) starts a text block; <code>/Helvetica 12 Tf</code> selects the font named "Helvetica" from the resource dictionary at 12 units; <code>100 700 Td</code> moves the text position to coordinates (100, 700) in the page's default coordinate system; <code>(Hello, World!) Tj</code> shows the string; <code>ET</code> (End Text) ends the text block.

Coordinates are in the page's user space, which by default has the origin at the bottom-left corner of the page, with x increasing to the right and y increasing upward. This is opposite to screen coordinate systems (where y increases downward) — a source of confusion when converting between PDF page coordinates and CSS/screen coordinates in web-based PDF editors.

<strong>Token types in content streams:</strong> - <strong>Integer and real numbers</strong>: <code>100</code>, <code>0.5</code>, <code>-3.14</code> - <strong>Strings</strong>: <code>(literal string)</code> or <code>&lt;hex string&gt;</code> - <strong>Names</strong>: <code>/Helvetica</code>, <code>/DeviceRGB</code> - <strong>Operators</strong>: <code>Tj</code>, <code>Td</code>, <code>re</code>, <code>f</code>, <code>q</code>, <code>Q</code> - <strong>Arrays</strong>: <code>[200 0 0 200 0 0]</code> (used for the TJ operator's glyph array)

Inline images (using the <code>BI/ID/EI</code> operators) embed image data directly in the content stream rather than using a separate XObject. This is appropriate for small images; larger images should use Image XObjects referenced by the <code>Do</code> operator.

PDF content stream operators are grouped into functional categories:

<strong>Graphics state operators</strong> — Modify the current graphics state: <code>q</code> (save state), <code>Q</code> (restore state), <code>cm</code> (concatenate matrix — multiply the current transformation matrix), <code>w</code> (line width), <code>J</code> (line cap), <code>j</code> (line join), <code>M</code> (miter limit), <code>d</code> (dash pattern), <code>ri</code> (rendering intent), <code>i</code> (flatness), <code>gs</code> (set extended graphics state from ExtGState resource).

<strong>Path construction operators</strong> — Build paths before painting: <code>m</code> (move to), <code>l</code> (line to), <code>c</code> (Bézier curve to), <code>v</code> / <code>y</code> (partial Bézier curve), <code>h</code> (close path), <code>re</code> (rectangle).

<strong>Path painting operators</strong> — Stroke, fill, or clip: <code>S</code> (stroke), <code>s</code> (close and stroke), <code>f</code> / <code>F</code> (fill, nonzero winding), <code>f*</code> (fill, even-odd), <code>B</code> (fill and stroke), <code>b</code> (close, fill, stroke), <code>n</code> (end path without painting — used for clip).

<strong>Clipping path operators</strong> — <code>W</code> (set clipping path, nonzero), <code>W*</code> (set clipping path, even-odd). These must be followed by a path-painting operator or <code>n</code>.

<strong>Text operators</strong> — <code>BT</code> / <code>ET</code> (begin/end text block), <code>Tf</code> (font and size), <code>Td</code> / <code>TD</code> (move text position), <code>Tm</code> (set text matrix), <code>T*</code> (next line), <code>Tj</code> (show string), <code>TJ</code> (show string with individual glyph positioning), <code>'</code> (move to next line and show), <code>"</code> (move, set spacing, show), <code>Tc</code> / <code>Tw</code> / <code>Tz</code> / <code>TL</code> / <code>Tr</code> / <code>Ts</code> (character/word spacing, horizontal scaling, leading, rendering mode, rise).

<strong>Colour operators</strong> — <code>CS</code> / <code>cs</code> (set colour space for stroke/fill), <code>SC</code> / <code>sc</code> (set colour for stroke/fill), <code>SCN</code> / <code>scn</code> (set colour with additional parameters), <code>G</code> / <code>g</code> (DeviceGray stroke/fill), <code>RG</code> / <code>rg</code> (DeviceRGB stroke/fill), <code>K</code> / <code>k</code> (DeviceCMYK stroke/fill).

<strong>XObject and inline image operators</strong> — <code>Do</code> (invoke named XObject — image, form, or PostScript XObject), <code>BI</code> / <code>ID</code> / <code>EI</code> (inline image begin/image data/end).

<strong>Marked content operators</strong> — <code>BMC</code>, <code>BDC</code>, <code>EMC</code>, <code>MP</code>, <code>DP</code> — for accessibility tagging and optional content. See <a href="https://pdftheguide.com/glossary/pdf-marked-content" target="_blank" rel="noopener">PDF Marked Content</a> for details.

A page's content is specified by its /Contents entry, which can be:

- A single stream object reference - An array of stream object references (concatenated in order to form a single logical content stream)

Multiple content streams in the /Contents array are logically treated as a single stream — they share the graphics state. This design allows PDF generators to produce separate content streams for different parts of a page (text layer, annotations layer, background graphics) and assemble them into a single page without full reprocessing.

<strong>Form XObjects as reusable content streams</strong> — A Form XObject (distinct from form fields) is a named content stream referenced in the page's /XObject resource dictionary. The <code>Do</code> operator invokes it, embedding its content at the current transformation matrix position. Form XObjects encapsulate a self-contained piece of graphics with its own resources and transformation. They are used for: repeated elements (logos, backgrounds), reference XObjects (imported from other PDFs), and as the container for Type 3 font glyphs.

<strong>Content stream isolation</strong> — The graphics state is shared across content streams in a page's /Contents array, but Form XObjects have their own graphics state context. The <code>q</code>/<code>Q</code> operators save and restore the graphics state stack — a <code>Do</code> invocation of a Form XObject effectively runs the form's stream within a <code>q</code>/<code>Q</code> bracket.

<strong>Resource dictionary</strong> — Content streams reference resources (fonts, images, colour spaces, extended graphics states) by name. The resource dictionary (/Resources) on the page (or Form XObject) maps those names to the actual resource objects. This indirection allows the same resource to be shared across multiple pages and allows content streams to be self-describing.