Developer tools
CSV Split
Choose rows, parts or column groups. Keep every field intact.
Processed in your browser
1. Input CSV
Paste CSV, or open a UTF-8 file.
Auto-detect tries comma, semicolon, tab and pipe. Choose explicitly if the preview looks wrong.
2. Split options
Without repetition, the header stays in the first part only. Header rows do not count toward the split size.
Preparing tool…
Parts
Selected part preview
Scroll within the table to see more columns. With the table focused, use the arrow keys.
Preview shows 25 rows and 8 columns at a time; downloads include every field.
How it works
- Add your CSV
Paste text or open a UTF-8 file. Check delimiter and header.
- Choose the split
Divide by row count, balanced parts or an exact column value.
- Save your parts
Check the preview, then download one part or all parts as a ZIP.
Your data stays on your device
Parsing, splitting and ZIP creation run in your browser. Input and output stay in memory and are cleared on reset or when leaving this page. Files are never uploaded.
Also try CSV Viewer, CSV Sort or CSV Dedupe Rows.
Details and frequently asked questions
How are rows divided?
Rows per part makes consecutive chunks; the final chunk can be shorter. Number of parts makes balanced consecutive chunks, with an extra row in earlier parts when needed. It creates fewer files if you request more parts than data rows. Empty or header-only input produces no files.
How do column groups work?
Use a column number, starting at 1. Exact field values define groups: case and whitespace matter. Groups follow first appearance; rows within each group keep their original order. A missing cell is separate from an empty field. Files use unique sequential names such as group-0001.csv; group values appear in the part list.
What happens to quoted newlines and blank rows?
The tool splits parsed CSV records, so quoted newlines stay in their field. Blank records and uneven row widths are preserved. A final line ending is treated as a record terminator, not an extra record. Duplicate headers, Unicode and leading zeros remain strings.
Are the output files byte-for-byte copies?
No. Parts are UTF-8 without a BOM, with every field quoted, CRLF record separators and the selected or detected delimiter. Field values remain unchanged, including formula-like text. Spreadsheet software may interpret formulas or numbers when opening CSV; import columns as text when necessary. Headers repeat by default; turning repetition off keeps the header in the first part only.
What are the practical limits?
There is no imposed input-size cap. Available browser memory determines what can finish. Cancel terminates processing. ZIP downloads use ZIP32 without compression: archives must be below 4 GiB and contain fewer than 65,535 parts. If ZIP creation fails, individual parts remain available. Preview windows affect display only.