Translators receive source text in every condition imaginable: copied from PDFs with broken line breaks, riddled with smart quotes and non-breaking spaces, and salted with invisible characters that confuse CAT-tool segmentation. Cleaning the source before it enters the workflow saves hours of fighting the tool later.
This workflow prepares source text for translation: fix line breaks and whitespace, normalize quotes and Unicode, and remove invisible characters so segmentation and QA behave.
Translators & localizers get a cleaner result from “Text Cleanup for Translators” when each step stays separate. Finish the current step and check the output before you open the next tool. The utilities named for this workflow (remove line breaks, smart quote converter, unicode text normalizer) are linked further down so you can jump to the one that matches the file you have, instead of guessing from the full directory.
Keep the original text or file until the last step looks right. These tools run in your browser, so you can retry a step on the same copy without uploading the list. If the result is shorter than you expected, look back at the input and confirm the addresses, lines, or columns are still there, then rerun only the step that dropped them.
A useful check for translators & localizers before you export: Clean source before importing into your CAT tool - fixing segmentation after the fact is far slower. After that, make sure the output is the shape the next tool expects, usually one item per line or a single column, and only then import it into a spreadsheet, CRM, or sending platform.
Worked example: copy three rows that look like the real file and run only “Repair line breaks”. Compare that small output with what you expected before you process the full list. A common pitfall is accepting the first clean-looking result when a later step still has to remove duplicates, fix separators, or drop blank lines. If the sample is wrong, correct the input and rerun that step. Checklist: keep the original file, confirm the sample output, finish every later step on the sample, then run the same order on the full file and spot-check the first and last rows.