Convert MAT to CSV

A .mat file is a saved MATLAB workspace: one or more named variables in a single binary file. One fact decides which tools work on yours: files saved with MATLAB’s -v7.3 flag are secretly HDF5 files and need different tooling than v7.2 and older — if a converter fails with an “unable to read” error, that’s almost always why (see the FAQ). There’s no in-browser converter for this format here yet, so this guide covers the free routes. Fastest with no software: Octave Online below. Python users can jump ahead.

No installation: convert in the browser

Option 1: Octave Online

Octave Online (octave-online.net) is a free browser version of GNU Octave, the MATLAB-compatible environment.

1. Open it — Go to octave-online.net and sign in (free — required to upload files).
2. Upload your file — Use the file panel on the left to upload your .mat file.
3. Run the commands — Paste these into the prompt, one at a time:

load yourfile.mat            % replace with your uploaded filename
whos                         % lists the variables inside
csvwrite("output.csv", A)    % replace A with a variable name from whos

4. Download the result — output.csv appears in the file panel; download it from there.

Two honest caveats: csvwrite handles numeric matrices only (for cell arrays or structs, use the Colab option), and Octave cannot load -v7.3 files — if load errors, use the Colab h5py snippet in the Python section below.

Option 2: Google Colab with scipy

1. Open Colab — Go to colab.research.google.com, sign in with any Google account, and click New Notebook.
2. Upload your file — Click the folder icon in the left sidebar, then the upload button, and select your .mat file.
3. Run the code — Paste the snippet below into the cell and press the run button.
4. Download the result — Refresh the file sidebar, right-click output.csv, and choose Download.

from scipy.io import loadmat
import pandas as pd

data = loadmat("yourfile.mat")  # replace with your uploaded filename
print([k for k in data if not k.startswith("__")])  # your variable names
df = pd.DataFrame(data["A"])  # replace A with one of the printed names
df.to_csv("output.csv", index=False)

Your uploaded file lives in a temporary private session that disappears when you close the tab — but it is on Google’s servers, so for confidential data use the desktop option below.

Free desktop app: GNU Octave

GNU Octave is a free MATLAB-compatible environment (Windows/macOS/Linux, from octave.org). Like the R case for .rds files, this is an app plus two pasted commands, not a tool you need to learn: install it, then run the same load / whos / csvwrite commands from the Octave Online section. Everything stays on your machine.

Python

scipy.io.loadmat() returns a dictionary of variable names to arrays (plus a few __header__-style keys you can ignore):

# pip install scipy pandas
from scipy.io import loadmat
import pandas as pd

data = loadmat("data.mat")
print([k for k in data if not k.startswith("__")])  # see the variable names
pd.DataFrame(data["A"]).to_csv("output.csv", index=False)  # replace A

For -v7.3 files, loadmat raises an error — read them as HDF5 with h5py instead (this is the same task as the HDF5 guide):

# pip install h5py pandas
import h5py
import pandas as pd

with h5py.File("data.mat", "r") as f:
    print(list(f.keys()))  # variable names
    data = f["A"][()]      # replace A with one of the printed names

pd.DataFrame(data).to_csv("output.csv", index=False)

One gotcha: MATLAB stores arrays column-major, so v7.3 arrays arrive transposed through h5py — if rows and columns look swapped, use data.T.

R

For v7.2 and older files only, R.matlab::readMat("data.mat") returns a list of the file’s variables — extract the one you want and write.csv() it. The package refuses -v7.3 files; use the h5py route for those.

Which method should I use?