Online HDF5 viewer for .h5 and NetCDF-4 files

How to open an HDF5 file online

1. 1Open your .h5 or .nc file. Choose it, or drop it on the viewer at the top of this page. Its first group opens with its first dataset plotted against its time. No file to hand? Click “Try a 3.5-million-row HDF5 file”: two days of made-up readings, made on your device.
2. 2See what’s inside. The menu in the file’s line lists its tables: one per group, per 2-D dataset and per record array. “Variables” shows every group, dataset, link and attribute, with its type, its shape and its first values.
3. 3Plot, zoom and export. Tick datasets in the list or in Variables to plot them, zoom with the scroll wheel, a pinch or a drag, and read their statistics. “Download all columns (CSV)” under “Do more” saves the table shown, with its time column.

A worked example you can check

This example uses the smaller sample: “Sample HDF5 file”, under “Smaller samples”, opens a month of made-up UK weather, written with h5py: three groups at three rates, each with its own time, and a 2-D dataset (download the sample HDF5 file). It opens four tables:

• /hourly (744 rows, 5 columns): temperatures in London, Leeds and York, London’s humidity, and its pressure stored packed, every hour from 1 January 2026.
• /rain_gauge (4,464 rows, 2 columns): rain in London and Leeds every ten minutes, on a time of its own.
• /daily (31 rows, 4 columns): the day’s highest and lowest temperature, its rain and the weather in words.
• /stations/wind_speed (744 rows, 4 columns): one 2-D dataset of 744 hours × 4 stations: a column per station, named by its station dimension, against the hourly time.

It opens on /hourly, with London_Temperature plotted against the group’s time: a dimension scale whose units, “hours since 2026-01-01 00:00:00”, make it dates and times in UTC. In Variables, the dataset reads float64 (744,), with its first values 5.44, 4.99, 4.74, 4.7, … · degC, and its attributes long_name = Air temperature, London · units = degC.

London_Pressure is stored as packed integers, int16 (744,), with these attributes:

_FillValue = -32768 · add_offset = 1000 · long_name = Sea-level pressure, London · scale_factor = 0.1 · units = hPa

The viewer unpacks it (each value times scale_factor plus add_offset: 1013, 1011.8, 1012.6, 1012.3, … · hPa) and leaves empty the hours stored as the fill value (2 of them), so the line has a gap there instead of a drop to −32768.

The same file read with h5py, the Python library for HDF5 files, gives the same groups, values and times:

import h5py

with h5py.File("megarows-sample-uk-weather.h5", "r") as f:
    for name, group in f.items():
        print(name, list(group))
    t = f["hourly/London_Temperature"]
    print(t.dtype, t.shape, t.attrs["units"], t[:3])
    print(f["hourly/time"].attrs["units"], f["hourly/time"][:3])
    p = f["hourly/London_Pressure"]
    print(p[:3] * p.attrs["scale_factor"] + p.attrs["add_offset"])
    print(f["stations/wind_speed"].shape, f["stations/wind_speed"][0])

hourly ['time', 'London_Temperature', 'Leeds_Temperature', 'York_Temperature', 'London_Humidity', 'London_Pressure']
rain_gauge ['time', 'London_Rain', 'Leeds_Rain']
daily ['time', 'London_Max_Temperature', 'London_Min_Temperature', 'London_Rain_Total', 'London_Conditions']
stations ['station', 'wind_speed']
float64 (744,) degC [5.44 4.99 4.74]
hours since 2026-01-01 00:00:00 [0. 1. 2.]
[1013.  1011.8 1012.6]
(744, 4) [2.8 4.2 4.9 4.7]

And “Download all columns (CSV)” under “Do more” starts the CSV of /hourly like this:

time,London_Temperature,Leeds_Temperature,York_Temperature,London_Humidity,London_Pressure
2026-01-01 00:00,5.44,3.32,2.79,93.1,1013
2026-01-01 01:00,4.99,2.93,2.42,92.8,1011.8
2026-01-01 02:00,4.74,2.72,2.2,88,1012.6

Run on 2026-10-09 with h5py 3.16.0, HDF5 2.0.0, NumPy 2.5.3, Python 3.12.15. A test opens the sample on every change to the site and checks that the viewer shows these tables, times and values, and writes this CSV. Download the sample (.h5, 74 KB) to open it elsewhere too.

Groups, datasets and their time

An HDF5 file holds groups of datasets, each with attributes. The viewer turns each group into a table: its 1-D datasets of the same length become columns, numbers first, then true or false, then text, so the readings of one logger or one model output plot on one time axis. Datasets of another length in the same group get a table of their own, and datasets you tick in Variables from different tables open together, each on its own x axis.

The x axis is the dimension scale attached to the datasets, else a dataset of the group named like a time whose values increase. CF time units (“seconds since …”, “hours since …”, “days since …”) give dates and times in UTC. Units come from each dataset’s units attribute; packed values (scale_factor, add_offset) are unpacked, and _FillValue and missing_value are shown as empty, as the CF conventions ask.

A 2-D dataset becomes a table of its own: its samples run along the dimension with a time axis (else the longer one), and each step of the other dimension is a column, named by a dimension scale of text (as the sample’s station dimension names London, Manchester, Leeds and Glasgow) or by a columns attribute. A record array (a compound dataset) becomes a table of its members, a time member as its x axis.

NetCDF-4 files

A NetCDF-4 .nc file is an HDF5 file underneath, so it opens the same way: its variables of each dimension make a table, the coordinate variable of that dimension is the time axis, its CF times are dates and times, and a 2-D variable over a station dimension gets the station names as its column names. NetCDF dimensions without values of their own are listed in Variables. NetCDF 3 (classic) files are a different format and get a plain message.

Big files

HDF5 files are often large. The viewer reads the file on your device and loads the datasets its tables use, up to 2 GB of numbers; a dataset that would go past that is listed in Variables with why, and the rest of the file opens. Zooming reloads the full detail of the range in view, so a one-sample spike in millions stays visible.

To see it with a big file, “Try a 3.5-million-row HDF5 file” makes two days of made-up sonic anemometer readings on your device (the wind in three directions and the temperature, 20 a second) and writes them as an HDF5 file: /sonic with 3,456,000 rows against a CF time, /minute with each minute’s mean, lowest and highest value, and /daily, a summary of each day. Then it opens like any other file: nothing is downloaded or uploaded. Zoom into any ten seconds of /sonic and every reading is there. Phones get six hours, 432,000 rows, without /daily.