CSV to JSON converter

What This Tool Does

The converter reads a CSV or TSV file and turns it into JSON — in three structural variants, depending on what you do with the result next. While reading, it auto-detects the delimiter, guesses column types and respects leading zeros so a German postal code like 07001 does not shrink to the number 7001.

Pure-client. Every input stays in your browser. No server, no cookie wall, no tracking. Large files beyond five megabytes still process locally — through the NDJSON output that avoids bloating into a single giant array.

Which CSV Dialects Does the Converter Support?

CSV is not a single standard. Four presets cover the dominant dialects that turn up in real-world workflows.

RFC 4180 strict

RFC 4180 has been the de-facto international standard since 2005. Comma as the delimiter, period as the decimal separator, double-quote as the escape — so "O""Connor" for a name with an apostrophe. English-language APIs, research exports and most US spreadsheets follow this profile.

German Excel

German Excel exports differently by default: semicolon as the delimiter, comma as the decimal separator, UTF-8 BOM at the start. Run a German Excel CSV through an RFC 4180 parser and you get a single fat column back — the parser cannot tell the thousand-period from the decimal-comma. The German Excel preset honours all three idiosyncrasies in one click.

Tab-separated (TSV)

TSV uses the tab character as the separator. Without a visible character it tends to be more robust against comma or semicolon collisions in the content — research exports from R, Pandas and SPSS often land as TSV. Header row and quoting follow the same rules as CSV, only the delimiter changes.

Pipe-delimited

The pipe character | is common in mainframe, SAP and ERP exports. The advantage: pipe rarely shows up in natural-language content, so collisions are unusual. The drawback: Excel does not recognise it as a default and prompts for the delimiter at import time.

Which JSON Structures Does the Converter Emit?

Three structures, three target systems.

Array of objects

[{"name": "Alice", "age": 30}, {"name": "Bob", "age": 25}] — the default for REST APIs. Each CSV row becomes one object; the header row supplies the keys. Human-readable, directly consumable by any JavaScript frontend that expects a records array.

Object of arrays

{"name": ["Alice", "Bob"], "age": [30, 25]} — the column-oriented form. Libraries like Pandas, R data frames and D3.js prefer this layout because column operations (mean, filter, plot) work more efficiently than converting from a records array first.

NDJSON

NDJSON — one JSON object per line, no enclosing array:

{"name":"Alice","age":30}
{"name":"Bob","age":25}

Streaming-friendly: an NDJSON reader processes one line at a time without loading the whole file into memory. For CSV files larger than five megabytes the converter recommends NDJSON automatically — a 100,000-record array reaches three-digit megabytes as a single JSON string instantly.

How Does the Converter Protect Leading Zeros?

Leading zeros are the most common CSV-to-JSON conversion bug. The moment a cell like 07001 (postal code Sondershausen) becomes a number, the zero disappears and the address database breaks. Three heuristics prevent that.

Consistent leading zeros

If the column analyzer sees that every non-empty value in a column starts with 0 and runs at least three digits long, the column is classified as string — even when the rest of the value is digits. A ZIP column with 07001, 80331, 10117 stays a string, not a number.

IBAN detection

IBAN follows a clear pattern: two letters (country code), two check digits, then 10–30 alphanumeric characters. If every row of a column matches the pattern, the column stays a string — even when German IBANs contain only digits beyond the country code.

Phone detection

Phone numbers often carry +, dashes, parentheses or spaces. When a column shows ten or more digits PLUS at least one non-digit character per value, the analyzer locks it to string. That keeps +49 30 1234567 intact as a phone string instead of turning it into a nonsensical number.

Manual override

The heuristic is conservative — it errs toward string rather than number. If you want a detected leading-zero column as a number anyway, change the type per column via the dropdown in the column panel. The change applies on the next render.

How Does Dot-Notation Produce Nested JSON?

A flat CSV header like name,email,plz yields a flat JSON object. When the source data is structured — say an address table — dot notation pays off:

user.name,user.email,user.address.plz
Ada,[email protected],07001

becomes

[
  {
    "user": {
      "name": "Ada",
      "email": "[email protected]",
      "address": { "plz": "07001" }
    }
  }
]

Multiple paths under the same prefix merge into one object automatically. The reverse direction (JSON-to-CSV tab) flattens nested trees back through the same notation — a round-trip-capable format.

When Does the Bidirectional Mode Help?

Some workflows need a round trip: export CSV, adjust the data in a JSON editor, convert back to CSV. The second tab JSON-to-CSV applies the same preset and delimiter choice as the forward path. Paste an array of objects as JSON and you get CSV with a header row and RFC 4180 quoting; nested fields flatten to dot notation.

Which File-Size Limits Apply in Practice?

Pure-client does not mean unlimited. Three thresholds shape how the converter behaves at different file sizes.

Below 5 MB — standard mode

Up to five megabytes of CSV — typically a few thousand rows — the converter handles directly. Array of objects or object of arrays render smoothly. The result panel shows the JSON immediately, copyable in one click.

5 to 50 MB — NDJSON recommended

Beyond five megabytes the converter suggests NDJSON via a banner. Background: a 100,000-record array becomes a 30- to 80-megabyte JSON string quickly, which slows the render path noticeably and freezes the copy-to-clipboard step for seconds. NDJSON sidesteps that because it streams line by line.

Beyond 50 MB — split first

Browser memory generally handles a hundred megabytes, but rendering a result field that size makes the UI sluggish. For files in the hundreds of megabytes range it is faster to split the CSV outside the browser — with a shell script for instance — and convert chunk by chunk.

Which Related Tools Are Available?

• CSV cleaner — for broken CSV files with mojibake umlauts, mixed delimiters or duplicate headers. The pre-step to this converter when the input is not clean.
• JSON to CSV — the specialised reverse path if you only need JSON-to-CSV without the CSV-input options.
• CSV to Markdown — when the target format is a Markdown table instead of JSON.
• JSON formatter — format and validate the generated JSON before you pipe it onward.