XLSX to CSV

XLSX to CSV, one sheet or all of them

A workbook is a stack of sheets. A comma-separated file is one flat grid. Every conversion between the two has to answer the question of which sheet you meant, and the usual answer is to quietly take whichever one was in front. That is what Save As does, and it warns you about it in a dialog most people dismiss without reading. A quarterly file with a tab per month therefore costs twelve trips through the same menu.

Here the sheets are listed the moment the file is read, each with its row and column count, and you tick whichever ones you want. One ticked gives you a plain download. Several ticked gives you an archive holding a separate file per sheet, named after the workbook and the tab so you can tell them apart a week later. Nothing is chosen for you and nothing is hidden.

A newly collected inventory can begin with the barcode scanner to Excel; export XLSX there for review, then return here when a receiving system needs flat CSV instead.

A workbook and a CSV file, in brief

XLSX

Full name	Office Open XML Workbook
Extension	.xlsx
Format type	Spreadsheet — a ZIP archive of XML parts
MIME type	application/vnd.openxmlformats-officedocument.spreadsheetml.sheet

CSV

Full name	Comma-Separated Values
Extension	.csv
Format type	Plain text table, one record per line
MIME type	text/csv

How to change Excel to CSV without opening the file

Google asks this on your behalf, in the questions underneath the results, and it is a fair thing to want. Launching a spreadsheet application to export one table is slow, it needs a license you may not have on this machine, and on a locked-down laptop it may simply not be an option.

An .xlsx is a zip archive with XML inside it, which means reading one takes no spreadsheet at all — just something that can unzip and parse. Your browser can do both. Drop the file above and it is unpacked in the page, so the workbook is never opened in the sense the question means, and never handed to anyone either. The other suggestion you will see, importing the thing into Google Sheets, does work — at the cost of putting the file on someone else's servers, which is the exact trade this site exists to avoid.

The separator is not always a comma

Send one of these files to a colleague in Stockholm, Berlin or Madrid and there is a good chance every row arrives crammed into column A. Their copy of the spreadsheet reads the semicolon as the field separator, because in those languages the comma is the decimal point and cannot do both jobs at once. Nothing is broken; the two machines simply disagree about what the punctuation means.

So the separator is a choice here rather than a consequence of your system settings, with comma, semicolon, tab and pipe available and a line under each saying who wants which. A tab is the quiet favorite when you control both ends, since tabs almost never turn up inside a value and therefore never need escaping. Whatever you pick, values containing the separator are wrapped in quotes and internal quotes are doubled, so an address with a comma in it stays one field.

Accented characters, and the three bytes that save them

The second thing that goes wrong is names. Åsa, José and 张伟 survive the conversion perfectly well — the trouble comes later, when a spreadsheet on Windows guesses at the encoding and guesses wrong, turning them into Ã…sa and José. The fix is a byte order mark: three invisible bytes at the front declaring the file to be UTF-8, ticked by default above and explained at length in what a byte order mark is.

Untick it when the file is going somewhere else. Command-line tools, database loaders and a good number of scripts read those bytes as ordinary characters and end up with a first column whose name has invisible junk glued to the front — a bug that takes an afternoon to find because the name looks correct everywhere you inspect it.

Dates are numbers, and formulas are answers

Inside the archive a date is stored as a count of days since the beginning of 1900. What makes 46036 mean the fourteenth of January is a formatting rule kept in an entirely separate part of the file. Converters that skip that part hand you a column of five-digit numbers, which is the single most common way this job goes wrong. Those rules are read here, dates come out in ISO order by default, and you can switch to the American or European arrangement if the recipient expects one. Only the columns the workbook genuinely formats as dates are touched — a text cell that happens to read like one is left alone.

Formulas resolve to their results: a cell holding a sum arrives as the number. This is not a shortcut but the only honest option, since the format has no way to store a formula and recalculating a workbook would need the whole function library the spreadsheet has and we do not. The number written is the one the spreadsheet itself worked out and saved. Files generated by scripts sometimes carry formulas that were never calculated at all; those cells come out empty, and the receipt counts them for you rather than leaving you to wonder.

Where the column names actually are

Plenty of real workbooks do not begin with their headings. There is a banner across the top — a project name, a quarter, the word DRAFT — then a blank line, and the words region, units and closed only turn up on the third row. Read literally, that banner becomes the heading line of your CSV and the genuine headings are demoted to ordinary data. Every column then carries the wrong name, and the file looks fine until something downstream reads it.

So the heading row is worked out from the shape of the rows rather than assumed to be the first one, and — this is the part that matters — it is shown to you as a choice you can change before anything is downloaded. Underneath the menu sit the headings it settled on and three genuine rows of data, so a wrong answer is visible rather than discovered later. Anything above the heading row is quoted back to you, because a title stranded above a table is not data and writing it out would leave the file with a first line a different width from all the others.

The detection is deliberately cautious, and it does miss a case worth naming: a budget sheet whose columns are called 2025 and 2026. Bare numbers are treated as values rather than names, which is what stops a stray total from being mistaken for a heading, and the cost of that caution is this one. Pick the right row from the menu and it is corrected in a click. You can also say there is no heading row at all, in which case every line in the sheet is written out as data — a perfectly ordinary CSV, since whatever reads it can name the columns itself.

What this page will not do

It will not read .xls, the format from before 2007. That is a binary document with nothing structurally in common with the modern one, and supporting it would mean a second reader for a format Microsoft itself has been retiring for fifteen years. Open such a file anywhere and save it again in the current format first.

Colors, column widths, merged cells, charts and validation rules all disappear, because the target format has no concept of any of them — that loss is inherent to the destination, not a shortcoming of the conversion. Going back the other way, from a flat table to a real workbook, is a genuinely different job, and it has its own page: CSV to Excel writes proper .xlsx output rather than a renamed text file, and lets you fix the separator and the column types before the file exists instead of arguing with an import wizard afterwards. (SQL to Excel is the same destination from a query result.) If you only want to read what you were sent, the CSV viewer lays a file out as a sortable grid, and Excel to SQL takes the same workbook straight to insert statements when a database is where it is headed. Excel to Markdown keeps worksheet, range, formula, hidden-cell and merge decisions visible when the destination is a pipe table.

Nothing leaves this tab

Several of the results above this page charge for the privilege of receiving your file — one desktop product on the first page of Google asks $49.90 — and all of them need a copy of it to do the work. Here the unpacking, the parsing and the download are assembled by your own browser. Open the network panel before you hand the file over and you can watch it stay empty.

When the CSV is only a waypoint and the real destination is an API or a script, the sheet can carry on to CSV to JSON in the next tab, which decides one type per column and shows you each decision before writing anything.

That matters more than it might for other formats. A workbook heading for conversion is usually a customer list, a payroll run or a set of accounts, and posting it to a stranger in return for a text file is a poor bargain. Large files are not refused either: you are told what your own machine is in for and the work then goes ahead, because a ceiling on a job that runs on your hardware would be an arbitrary limit rather than a real one.

One workbook this page cannot read straight away is one whose sheets are locked. The conversion itself does not care about a protection flag, but a spreadsheet you cannot edit is usually one you cannot check either — so if that describes your file, removing the sheet protection first takes a moment and happens in the same tab, after which the sheets convert normally.