Convert CSV to a MATLAB .mat file

How to convert a CSV file to MAT

1. 1Open your file. Drop a CSV onto the converter at the top of this page, or click “Choose a file”. TSV, TXT, Excel, JSON and Parquet files work too. The separator, header row, units row and column types are found for you.
2. 2Choose what goes in. Every column is ticked; untick the ones you do not need. Pick the layout (a struct with a field per column is the default), how dates and times are written, and whether numbers stay double. The size before compression is shown as you choose.
3. 3Download the .mat file. Click “Download”. Every row is written to a MATLAB v7 file named after yours, and your browser saves it like any download. In MATLAB, load it and your columns are there; GNU Octave and Python read it too.

No CSV to hand? Try it with sample data: 1,000,000 hourly readings of made-up UK weather, made on your device.

A .mat data file, not .m code

Not every “CSV to MATLAB” tool gives you a .mat file. Some write a .m script: MATLAB code with your numbers typed out as an array, which you run to rebuild the data. That is fine for a small table but slow and unwieldy for a big one. And a converter labelled “CSV to MAT” may turn the table into an image rather than keep its columns. This converter writes a binary MAT-file, the format MATLAB’s own save writes: one load reads it, every number exactly as in your CSV, each column under its own name.

Which programs read it

• MATLAB: the file is a MAT-file version 7, which MathWorks lists as readable by MATLAB 7.0 (R14) and later. We opened the sample’s file in MATLAB R2026b and every value matched, line for line (9 Oct 2026).
• GNU Octave: load reads it. We check every layout in Octave, value by value.
• Python: scipy.io.loadmat reads it, with every value the same as in your file. We check every layout with it too.

If a file from here does not load as you expect, tell us.

What you get in MATLAB

One load and your columns are there, each a column vector with every row. With the default layout they are fields of a struct called data:

load('weather.mat')                 % data, names, units, source_file, created_by, notes
data.London_Temperature             % one column, every row
t = datetime(data.time, 'ConvertFrom', 'datenum');
plot(t, data.London_Temperature)
T = struct2table(data);             % the same columns as a MATLAB table
names                               % the original column names

Next to data, the file holds names (your column names as they were, in order), units (when the file has a units row or the column names carry units), source_file, created_by and notes, which say how dates were written and how to turn them back into datetime.

Three layouts

• A struct with a field per column (the default): data.London_Temperature. Every column keeps its type: numbers, text, true/false. struct2table(data) makes it a table.
• One variable per column: London_Temperature straight in your workspace, as if you had typed it. Each column can be up to 2 GB.
• A matrix of the number columns: data(:, 2), with names and units in the same order as the matrix columns. Text columns are left out and listed in notes.

Numbers are double by default, every digit exactly as in your file. Choose single for half the size, or keep integer types to save whole-number columns as int32 (int64 when they need it).

Dates and times

A date-time column is written as a MATLAB datenum by default: days from year 0, the number datetime(x, 'ConvertFrom', 'datenum') turns back into dates. Choose POSIX seconds (seconds since 1 January 1970 UTC) for 'ConvertFrom', 'posixtime', or text to get the dates as the table shows them. A file that names its time zone keeps its own clock, and notes says which zone. Durations become seconds.

Excel or Parquet to MAT

The same converter takes an Excel workbook (.xlsx, .xlsm: pick the sheet), a Parquet file, JSON or JSON Lines, and TXT or DAT logger exports. A .mat file opens too, so you can save part of it again. To go the other way, use MAT to CSV; to look inside a .mat file first, open it in the MAT file viewer.

Big files

There is no fixed row limit; the practical limit is your device. A .mat file (v7) holds at most 2 GB per variable before compression, about 268 million numbers. With the struct layout that is the whole struct, so for a very big file choose One variable per column: the converter tells you before writing anything if a variable would be too big. Bigger variables (v7.3 files) are coming.

The same in MATLAB

With MATLAB at hand and a file that fits in memory, readtable and save do it in three lines:

T = readtable("weather.csv");
data = table2struct(T, "ToScalar", true);
save("weather.mat", "data")

Prefer code? How to convert CSV to a .mat file in MATLAB, Python or Octave: tested code, and what to watch for with column names, text, dates and big files.

The converter here needs no MATLAB licence, reads files too big to open comfortably in a script, and works on any computer with a browser. See all tools for the other converters.

Private by design: nothing is uploaded

Your browser reads the file straight from your disk and writes the .mat file on your device: the file and its name never leave it, which you can check in the Network tab of your browser’s developer tools. Read the privacy statement for details.