Convert Excel to CSV

1. 1

   Add your workbook

   Drag and drop an XLSX, XLSM, XLS, XLSB or ODS file, or click to select one. It is read on your device and never uploaded.

2. 2

   Choose how values are written

   Pick full precision to keep every digit the cell holds, or the workbook's display format to match what Excel shows. Choose ISO 8601 dates for databases and pipelines, or the sheet's own date format.

3. 3

   Set the CSV format

   Choose the delimiter, quoting, line endings and output encoding. Trim blank rows and columns from either end, and skip any heading rows above the table.

4. 4

   Select sheets and download

   Tick the sheets you want, optionally combine them into a single CSV with a sheet-name column, then download one file or a ZIP of them all.

Everything below is available without paying: the value and date handling, every CSV format option, the cleanup controls, sheet selection and combining. The paid tier adds batches and lifts the file-size limit — it does not unlock the conversion itself.

• Full-precision numbers. The value the cell holds is what gets written, so a figure displayed as 3.14 still exports all fifteen of its digits.
• ISO 8601 dates. Unambiguous 2020-01-01 output that every database and BI tool accepts, instead of the workbook's local display format.
• Formula export. Write the formulas themselves rather than their cached results, for auditing or for rebuilding a model elsewhere.
• Five output encodings. UTF-8, UTF-8 with a byte-order mark, UTF-16 LE, Windows-1252 and ISO-8859-1, with a count of any character the target cannot hold.
• Leading and trailing cleanup. Remove blank rows and columns from the start of the sheet as well as the end, and skip heading rows above the table.
• Multi-sheet selection. Convert any combination of sheets — not just one, and not only all of them — including hidden sheets.
• Combine into one CSV. Merge the selected sheets into a single file, with the header written once and an optional column naming each row's source sheet.
• Live preview. A table view and the exact CSV text, both regenerated as you change settings, so you check the result before you download it.
• Copy to clipboard. Take the CSV straight into a script, a message or another spreadsheet without saving a file first.
• Batch and ZIP. Queue many workbooks and take everything away as a single archive.
• Nothing is uploaded. The workbook is read on your device, which is why a file of salaries or customer records is safe to convert here.

A spreadsheet cell holds one thing and shows another. Type 3.14159265358979 into a cell formatted to two decimal places and the cell still holds all fifteen digits; the two-decimal 3.14 is only how it is drawn on screen. Most converters export the drawing, and the difference is invisible until someone tries to use the numbers.

The same three cells, two ways

• 3.14159265358979 shown as 3.14. Exporting the display writes 3.14 and eleven significant digits are gone for good. Full precision writes the value the cell holds.
• 1234567.891 shown as 1,234,567.89. Exporting the display writes a thousands separator into a comma-separated file. Either it is quoted and every consumer reads it as text rather than a number, or it splits the row into extra columns. Full precision writes 1234567.891.
• A percentage shown as 7.5%. The cell holds 0.075. Exporting the display writes a percent sign into a numeric column, which nothing will do arithmetic on.

When to choose "As displayed" anyway

Use it when the formatting is the point — a report someone will read rather than compute with, a column of currency that must keep its symbol, or a file that has to match a printed sheet exactly. It is a deliberate choice here, not the default, and the preview shows you what it will produce before you download.

Spreadsheets have no date type. A date is a number counting days from an epoch, wearing a display format — which is why the same file can show 01/02/2024 to one person and 02/01/2024 to another, and why a date that survives an export can still be wrong on arrival.

The three options

• ISO 8601 (default). 2020-01-01, or 2020-01-01T09:30:00 with a time. Sorts correctly as text, is unambiguous in every locale, and is accepted by every database without a format string.
• As displayed. Whatever the workbook's own format produces. Right when you need the file to match the sheet exactly; risky when it crosses a border.
• Serial number. The raw day count. Useful when the receiving system will do the conversion itself, or when you need to preserve a value the calendar cannot express.

Two quirks worth knowing

• 1900 was not a leap year. Spreadsheets carry a decades-old compatibility bug that treats 29 February 1900 as a real day. Dates before 1 March 1900 are therefore offset by one unless a converter accounts for it — this one does, and reports that day as 1900-02-29 rather than shifting it silently.
• Some workbooks count from 1904. Files created on older Mac spreadsheets use a different epoch. Ignoring it puts every date four years and a day out, and nothing about the result looks wrong. The setting is read from the file and applied automatically.

The right settings depend entirely on what will read the file. These are the ones that matter and what each is for.

• Delimiter. Seven to choose from. Comma is standard; the others exist because of locales and pipelines, covered in the next section.
• Encoding. How characters become bytes. UTF-8 for anything modern; UTF-8 with a byte-order mark when Excel will open the file by double-click; Windows-1252 or ISO-8859-1 for older Windows and Unix software; UTF-16 LE for tools that expect what Excel calls Unicode Text.
• Line endings. LF for Unix, macOS, git and most databases; CRLF for Windows tools and some strict importers.
• Quoting. Minimal follows the CSV standard and quotes only what must be quoted. Quote-all wraps every field. None removes the characters that would otherwise break a row.
• Trims. Remove blank rows and columns from the start of the sheet, the end, or both.
• Skip rows. Drop a fixed number of rows above the table, for exports that begin with a title and a date stamp.

The delimiter is the single most common cause of a CSV that opens as one long column, and the fix is almost always to match what the receiving program expects rather than what looks tidy.

• Comma. The default and the safest general choice. Use it for databases, APIs, scripts and anything that says "CSV" without qualification.
• Semicolon. Use it wherever the comma is the decimal separator — much of Europe and Latin America. A spreadsheet in those locales expects semicolons and will otherwise put every row in a single cell.
• Tab. Tab-separated values, and the format that pastes into a spreadsheet cleanly. Useful when the data itself is full of commas.
• Pipe. Common in ETL and data-warehouse loading, where it is chosen precisely because it almost never appears in real text.
• Colon, slash and hash. Rarer, and chosen for the same reason as the pipe: when the data itself is full of commas and semicolons, a delimiter that never appears in it is the one that survives.

Encoding decides how letters become bytes, and getting it wrong is why an exported file shows Café as Café. The trap is that a plain UTF-8 file is correct and still displays wrongly, because the program reading it assumed something else.

• UTF-8. The right answer for databases, scripts, git and any modern tool. Covers every language and every symbol.
• UTF-8 with BOM. The same bytes with a short marker at the front. Choose this when someone will open the file by double-clicking it in Excel, which otherwise assumes the system's local encoding and mangles every accent.
• UTF-16 LE. What Excel writes when you save as Unicode Text. Some Windows utilities and bulk-import tools require it.
• Windows-1252. The Western European legacy encoding still expected by older Windows software and many ERP imports. It covers accented Latin letters, curly quotes and dashes.
• ISO-8859-1. Latin-1, still required by some Unix pipelines and mainframe interfaces. Narrower than Windows-1252 — it has no curly quotes, dashes or euro sign.

The two legacy encodings cannot represent most of the world's characters. Rather than dropping them silently, the tool substitutes them and tells you how many were affected and which they were, so you can switch to UTF-8 knowing exactly why.

A CSV holds one table, and a workbook usually holds several. Every sheet is listed with its row and column count, hidden ones included, and you decide what happens to each.

• One sheet. Tick a single sheet and download a single CSV named after the workbook.
• Several sheets. Tick any combination. Each becomes its own CSV, delivered together as one ZIP, with names kept unique even when two sheet names reduce to the same filename.
• Combined into one file. Merge the selected sheets into a single CSV. The header row is written once rather than reappearing in the middle of the data, and an optional first column records which sheet each row came from so the file can be split apart again.
• Hidden sheets. Listed and marked, because a hidden sheet is often exactly the lookup table someone needs — and never converted unless you tick it.

The workbook is opened on your device and read once. Only the parts a CSV actually needs are expanded — the sheet data, the shared text and the number formats — so a file that is large because of its images or charts costs nothing extra to convert. Everything after that is a rewrite of what is already in memory, which is why changing a setting updates the preview instantly rather than starting again.

What happens to each cell

• Text is written as stored, with quoting applied only where the CSV rules require it.
• Numbers are written at full precision by default; the display format is applied only if you ask for it.
• Dates are recognised from the cell's number format and written as ISO 8601, because a spreadsheet stores them as plain numbers.
• Formulas are written as their last calculated result, or as the formula text if you choose that instead.
• Booleans and errors become TRUE, FALSE and the error text such as #DIV/0!, matching what the sheet shows.
• Empty cells become empty fields, so the column count stays the same on every row.

CSV is the lowest common denominator of tabular data. Where an XLSX is a compressed package of XML describing sheets, styles, formulas and charts, a CSV is a single table anything can read — which is exactly why it is what databases, APIs and analytics tools ask for.

• Data exchange. Every database import, BI tool and ETL pipeline accepts CSV; far fewer accept a workbook.
• Plain text. CSV is readable, diffable and easy to version in git, so a change to the data shows up as a change to the file.
• No spreadsheet required. The recipient can open it in any app, or read it from a script in three lines.
• Long-term storage. A plain text table will still be readable in twenty years, whatever happens to the application that produced it.

• Match the destination, not your own machine. Set the delimiter and encoding to what the receiving program expects. A file that opens perfectly for you can arrive unreadable.
• Keep full precision for anything computational. Use the display format only for files people will read rather than calculate with.
• Prefer ISO dates whenever the file leaves your team. It is the one format that cannot be misread.
• Check the preview before downloading. The row and column counts shown are the ones that will be written, so a trim that ate a column is visible immediately.
• Trim from the start for hand-made sheets. A title block above the table is the usual reason a CSV arrives with empty leading columns.
• Split very large workbooks first. If a file is beyond the size limit, splitting it into per-sheet workbooks converts each part comfortably.

The workbook is read once and everything after that is a rewrite of data already in memory, on a background thread so the page never freezes. That is what makes changing a setting feel instant even on a very large file.

• About one second to read a 30 MB workbook of 200,000 rows by 12 columns.
• Under a millisecond to rewrite the preview after any settings change, at any file size.
• Nothing is uploaded. There is no server step, so the size of your file is bounded by your device and not by anyone's bandwidth.

A 30 MB workbook of 200,000 rows is read in about a second, and changing any setting rewrites the preview in under a millisecond because the file is only ever read once.

• CSV to Excel
• Excel to JSON
• Excel Viewer
• Excel tools
• Split a large workbook

  Break a file past the free size limit into sheets you can convert one at a time

• Check the CSV afterwards

  Open the result and confirm the columns landed where you expected