#Editing Without Breaking Formatting

Translation review should improve the text without damaging the structure around it. This is especially important for EPUB, DOCX, PDF, and subtitle files where markup, styles, or timing are part of the content. This page explains what the CAT Editor preserves on export — and where to be careful.

#What The Editor Preserves (And What It Doesn't)

The CAT Editor is a rich-text editor, but export fidelity depends on the format:

For EPUB jobs the editor states this directly: "EPUB Translation Mode: only text edits are exported. Formatting changes are preserved from the original EPUB." For DOCX/PDF jobs, block-level formatting buttons are disabled because the export would drop them anyway — CatScribe disables what it cannot deliver rather than silently discarding your work.

Also note: tables and images are not rendered inside the editor panes. When a chunk belongs to a table or sits next to an image, check that region in the exported file rather than assuming the editor view is complete.

#Editing Principles

• Change the words, not the document structure, unless you mean to.
• Preserve placeholders, tags, links, and special punctuation.
• Keep paragraph boundaries — a chunk containing several paragraphs (the Source pane shows a paragraph-count badge) should keep the same number of paragraphs in the target.
• Review exported files in the app that will display them.
• Fix repeated formatting issues at the source (or with "Find and Replace") instead of chunk by chunk.

#Common Protected Elements

Watch for elements that should usually remain intact:

#Safe Editing Workflow

1. Review the Source and Target panes side by side; use "View in larger window" for long chunks.
2. Edit only the translated text; keep tags and placeholders in the same logical position.
3. Save, then check the "QA Flags" panel — a new tag or placeholder mismatch flag means your edit broke pairing.
4. If an "Improve with AI" pass damages formatting, don't repair it manually first — open "Version History" and "Revert" to the pre-AI version, then re-run with an explicit instruction (for example: "keep all tags and placeholders exactly as they are").
5. For engine output that consistently mangles markup, re-run the chunk with a different engine via the "Chunk Override" panel — see Chunk Overrides.
6. Export a small sample and inspect it before processing the whole project.

#Example: Preserve Tags

Source:

<i>Open the door.</i>

Good target:

<i>Abra a porta.</i>

Risky target:

<i>Abra a porta.

The risky version is missing the closing tag — everything after it may render italic. This is exactly what the "Tag count mismatch" QA flag exists to catch; never approve a chunk with that flag unresolved.

#Example: Avoid Merged Words

Merged words can appear when source text extraction loses spacing (most common in PDFs) or when a line break is removed incorrectly.

If you see text such as:

umabertura

Check whether it should be:

uma abertura

Then search nearby chunks for the same pattern — extraction problems cluster. If one page produced merged words, its neighbors probably did too. Use the job's chunk search ("Search source or target text...") to sweep for other instances before exporting.

#Pre-Export Formatting Checklist

• No unresolved tag or placeholder mismatch flags.
• Paragraph counts match between source and target on multi-paragraph chunks.
• Numbers, units, and dates verified on flagged chunks.
• Table and image regions checked in the exported file (not in the editor).
• Sample export opened in the target reading app before the final run.