L10N Diff

User guide

Find what changed in a revision and what needs another look. L10N Diff works best with earlier and revised versions in the same language. It does not judge translation accuracy between different languages.

  1. Basic comparison: add the earlier and revised versions
  2. Filters: decide what to ignore and what to review
  3. Fixed anchors: correct misaligned lines
  4. Copy results into a review sheet
  5. Files, drafts, and confidential text

A sample to compare

These fictional game strings were written for this guide. Copy each block into the original and revised inputs to inspect the number 10→15, variable {name}→{player}, and wording change.

Original · earlier version

Welcome, {name}!
Reward: 10 coins
Please open the door.

Revised · new version

Welcome, {player}!
Reward: 15 coins
Please unlock the door.
Back to comparison →

1. Basic comparison: add the earlier and revised versions

Put the earlier version in the left input and the revision in the right input, then compare. Ctrl/⌘ + Enter also runs comparison. For a first look, use the sample button in the app to open a ready-made result.

Side-by-side view helps you read matching lines together; stacked view suits narrow screens. Removal and addition highlights mark changes. Adjust their colors and the font size in the result tools.

2. Filters: decide what to ignore and what to review

Check the ignore options first. Speaker/ID differences are ignored by default; turn this off when reviewing IDs themselves. Choose whitespace, case, punctuation, or tag and variable ignoring to match your task. Ignoring a difference does not delete input data.

When checking tags or variables, begin with their ignore option disabled. For a wording review, ignoring whitespace can help remove formatting noise before inspecting substantive edits.

Show changed rows first, then use number-mismatch or tag/variable-mismatch filters to find candidates. A person must decide whether 10→15 is an intended reward change or whether a variable should be renamed. Combining filters can reduce visible rows; if nothing appears, disable them one at a time.

3. Fixed anchors: correct misaligned lines

Inserted or deleted lines and repeated dialogue can confuse automatic alignment. Use the review-needed filter and next-review navigation, then read the surrounding context to identify matching lines.

Open the anchor control in the result tools and connect the original and revised line numbers. If original line 12 and revised line 14 contain the same dialogue, enter that pair. The intervals between fixed anchors are realigned.

Line numbers refer to the original inputs. Anchors cannot cross in order; remove a conflicting anchor first and choose rows outside split or merged groups. A fresh comparison resets anchors, so finish editing the inputs before fixing alignment.

4. Copy results into a review sheet

Check your filters and view, then use formatted copy for the currently displayed comparison rows. Use original-only or revised-only copy when you need just one side.

Paste into Excel or Feishu Sheets and check rows, in-cell line breaks, and highlights. Formatting depends on the browser and receiving app. If copying fails, try Chrome or Edge over HTTPS.

Before delivery, confirm that all required rows are present and that filters have not hidden something you need. A diff points out edits; it does not guarantee translation quality or correct behavior in the game.

5. Files, drafts, and confidential text

Import text, Office documents, or ZIP files. For spreadsheets, select the worksheet you need. Extracted text may not preserve the visual layout of the source document, so inspect the input before comparing.

Comparison runs in your browser; inputs are not uploaded to L10N Diff servers. Temporary drafts may remain in this device’s local storage. If saving fails, keep a separate copy of your input. Clear inputs after using a shared computer.

Do not put confidential strings in the feedback form. Reproduce issues with fictional sentences like the sample above. See the privacy notice for information about site analytics and the external feedback service.