CSV to JSON Converter

How to use

1. Pick a direction (CSV → JSON or JSON → CSV), then paste your data into Input or use Open file to load a .csv, .tsv or .json file. The result appears in Output as you type.
2. Adjust the options: the delimiter can be detected or set by hand, "First row is header" treats the first CSV row as field names, and "Convert types" turns numbers and true/false into JSON numbers and booleans.
3. Use Copy or Download to get the result. Keep "Add BOM" on when the CSV will be opened in Excel. Swap moves the output back into the input so you can convert it the other way and check the round trip.

Worked example

CSV → JSON with header and type conversion on: the header id,name,price,active plus the rows 9007199254740993,Amy,1200.50,true and 007,Ben,300,false gives 2 rows × 4 columns. 300 becomes a number and true/false become booleans, while 9007199254740993 (beyond the safe integer range), 007 (leading zero) and 1200.50 (trailing zero) stay strings, so converting back to CSV reproduces them exactly.

JSON → CSV: [{"name":"Amy","tags":["a","b"]},{"name":"Ben","email":"ben@example.com"}] produces the columns name, tags and email (the union of both records), 2 rows × 3 columns. Amy has no email and Ben has no tags, so those cells are empty. The tags array becomes a JSON string, and because it contains commas and quotes it is written as "[""a"",""b""]".

Reading the results

• The rows × columns count in the status line covers data rows only, not the header. Blank lines in the CSV are skipped and not counted.
• A field-count warning means some rows have more or fewer fields than the header, usually because a value contains a comma but isn't quoted. Short rows are missing the trailing keys; extra values go into __parsed_extra.
• Without "Convert types", every JSON value is a string. With it, only numbers and true/false are converted; empty cells stay empty strings and dates are left as text.
• Duplicate header names are renamed automatically, so two columns called a become a and a_1 instead of one overwriting the other.

How it works

• CSV parsing and output use the papaparse library and follow RFC 4180 quoting: fields containing the delimiter, a quote or a line break are wrapped in double quotes, and quotes inside are doubled. Output lines end with CRLF.
• Delimiter detection tries comma, tab, pipe and semicolon and picks the one that gives the most consistent field count across rows. With a single column there is nothing to detect, so comma is used.
• Type conversion rule: a value becomes a number only if turning it into a number and back yields the identical text, and integers must stay within the safe range of ±(2^53−1).
• JSON → CSV: for an array of objects, the columns are the union of all keys in order of first appearance; a 2D array is written row by row; an array of plain values goes into a value column. Integers beyond the safe range are kept as their original text before parsing, so they are never rounded.
• Opened files are decoded as UTF-8 first; if the bytes aren't valid UTF-8, Big5 is used instead.

Limitations

• Only flat tables are supported. Nested JSON isn't flattened into columns like user.name; the whole object goes into one cell as a JSON string.
• Delimiter detection relies on rows having the same number of fields. With very little data or ragged rows it can guess wrong, so choose the delimiter manually.
• Only UTF-8 and Big5 files are decoded automatically. Save files in other encodings, such as Shift_JIS or GBK, as UTF-8 first.
• All data is processed in browser memory, so files of tens of megabytes or more may make the page slow.

FAQ

💬 Can't find an answer? Report a problem with this tool or send a suggestion