How to Merge Word and PowerPoint Files in the Browser with Open XML

DOCX and PPTX files are ZIP packages containing XML parts, media, and relationship graphs. That makes browser merging possible without Microsoft Office, but it also means concatenating document.xml or copying slide XML alone is not enough.

A correct merger must unzip each package, preserve order, allocate new relationship IDs, recursively copy images and dependent parts, rewrite relative targets, resolve filename collisions, update content types, and zip the result with the original Office MIME type.

Read Office files as package graphs

Open XML parts are connected by .rels files. A Word body can reference images, hyperlinks, numbering, headers, or embedded objects. A PowerPoint slide can reference a layout, master, theme, chart, notes, and media. Relationship targets are relative to the part that owns them.

The package root also contains [Content_Types].xml, which tells Office how to interpret extensions and individual parts. Every copied destination part must retain an appropriate default or override content type.

Build a reusable recursive part copier

Given a source part and destination path, copy its bytes, copy its content type, parse its relationship file, and recursively copy every internal target. External links stay external. If the desired path already exists, allocate a merged suffix and rewrite the relationship target relative to the new owner part.

Cache source-to-destination mappings during one import so shared layouts or media are copied once. A used-path set prevents two imports from claiming the same destination.

function copyPart(sourcePath: string, destinationPath: string) {
  destinationZip[destinationPath] = sourceZip[sourcePath];
  copyContentType(sourcePath, destinationPath);

  for (const relationship of readRelationships(sourcePath)) {
    if (relationship.targetMode === 'External') continue;

    const sourceTarget = resolveTarget(sourcePath, relationship.target);
    const destinationTarget = uniquePartPath(sourceTarget);
    copyPart(sourceTarget, destinationTarget);
    relationship.target = relativePath(destinationPath, destinationTarget);
  }

  writeRelationships(destinationPath, relationships);
}

Merge Word bodies and rewrite relationship IDs

Use the first DOCX as the base package. Split its w:body from the surrounding document XML, remove section properties from intermediate bodies, and insert an explicit page break before each imported document. The final base section properties remain at the end.

For every source document relationship, allocate a new rId in the base list, copy the target graph, and replace r:id, r:embed, r:link, or o:relid references inside the imported body. Without this rewrite, an image can point to another document’s unrelated relationship.

const base = unzipSync(await bytes(files[0]));
const bodyParts = [withoutSectionProperties(readWordBody(base))];

for (const file of files.slice(1)) {
  const source = unzipSync(await bytes(file));
  let body = withoutSectionProperties(readWordBody(source));

  for (const relationship of readDocumentRelationships(source)) {
    const nextId = getNextRelationshipId(baseRelationships);
    copyRelatedPartRecursively(relationship);
    body = replaceRelationshipReferences(body, relationship.id, nextId);
  }

  bodyParts.push(pageBreakXml(), body);
}

writeWordBody(base, bodyParts.join(''));
return zipSync(base);

Merge PowerPoint by slide order, not filenames

Slide filenames are not the authoritative order. Read p:sldId entries from presentation.xml, resolve each r:id through presentation relationships, and only fall back to numeric slide filenames when the order list is unavailable.

Assign a new slide part index, recursively copy the slide and dependencies, create a new presentation relationship, and append a p:sldId with an unused numeric ID. This preserves source order while avoiding collisions with existing slides.

• Copy layouts, masters, themes, charts, notes, and media through relationships.
• Preserve external hyperlinks without trying to package their targets.
• Update [Content_Types].xml after all copied parts.
• Count slides again from the rebuilt package before reporting success.

Why regex is acceptable only at controlled boundaries

This implementation uses targeted XML extraction and attribute rewriting for known Open XML structures. It is compact and browser-friendly, but regex is not a general XML parser. Namespace variations, unusual formatting, malformed packages, macros, signatures, and advanced Office features can exceed those assumptions.

Validate expected main parts before modifying a package, keep transformations narrow, and open generated fixtures in multiple Office viewers. For high-fidelity enterprise merging, use a complete Open XML library or a server environment with Office-grade tooling.

Security, limits, and output validation

ZIP packages can expand far beyond their compressed size. Limit file count, compressed bytes, and extracted entry count, and be alert to zip bombs. Do not execute macros or fetch external relationships while merging.

A successful zip operation does not prove Office will accept the file. Verify required parts, relationship targets, content types, slide or body counts, and then open representative output in Word, PowerPoint, LibreOffice, and a web viewer.

Key takeaways

Browser Office merging works because DOCX and PPTX are package graphs, not opaque binaries. The reusable core is recursive relationship copying with collision-safe paths and content types; Word then merges bodies and rIds, while PowerPoint appends ordered slide relationships. Treat output validation as part of the algorithm.