Skip to main content

binding for the libvips image processing library, API mode

Project description

Build Status

PyPI package:

This module wraps the libvips image processing library.

If you have the development headers for libvips installed and have a working C compiler, this module will use cffi API mode to try to build a libvips binary extension for your Python.

If it is unable to build a binary extension, it will use cffi ABI mode instead and only needs the libvips shared library. This takes longer to start up and is typically ~20% slower in execution. You can find out how pyvips installed with pip show pyvips.

This binding passes the vips test suite cleanly and with no leaks under python2.7 - python3.6, pypy and pypy3 on Windows, macOS and Linux.

We have formatted docs online here:

How it works

Programs that use pyvips don’t manipulate images directly, instead they create pipelines of image processing operations building on a source image. When the end of the pipe is connected to a destination, the whole pipeline executes at once, streaming the image in parallel from source to destination a section at a time.

Because pyvips is parallel, it’s quick, and because it doesn’t need to keep entire images in memory, it’s light. For example, the libvips speed and memory use benchmark:

Loads a large tiff image, shrinks by 10%, sharpens, and saves again. On this test pyvips is typically 3x faster than ImageMagick and needs 5x less memory.

There’s a handy chapter in the docs explaining how libvips opens files, which gives some more background.

Install libvips

You need the libvips shared library on your library search path, version 8.2 or later. On Linux and macOS, you can just install via your package manager; on Windows you can download a pre-compiled binary from the libvips website.

Install pyvips

Next, install this package, perhaps:

$ pip install --user pyvips

On Windows, you’ll need a 64-bit Python. The official one works well. Anaconda probably won’t work without some effort – they have their own packaging system.

Extra notes for Windows

On Windows, you will need to add vips-dev-x.ybin to your PATH so that pyvips can find all the DLLs it needs. You can either do this in the Advanced System Settings control panel, or you can just change PATH for your pyvips program.

If you set the PATH environment variable in the control panel, you can use the vips command-line tools, which I find useful. However, this will add a lot of extra DLLs to your search path and they might conflict with other programs, so it’s usually safer just to set PATH in your program.

To set PATH from within Python, you need something like this at the start of your program:

import os
vipshome = 'c:\\vips-dev-8.7\\bin'
os.environ['PATH'] = vipshome + ';' + os.environ['PATH']

Now when you import pyvips, it should be able to find the DLLs.

Test your install

Try this test program:

import logging
logging.basicConfig(level = logging.DEBUG)
import pyvips

If pyvips was able to build and use a binary module on your computer (API mode) you should see:

$ ./
DEBUG:pyvips:Loaded binary module _libvips
DEBUG:pyvips:Inited libvips

If the build failed (fallback to ABI mode), or there was a header or version mismatch, you might see:

$ ./
DEBUG:pyvips:Loaded binary module _libvips
DEBUG:pyvips:Binary module load failed: not all arguments converted during string formatting
DEBUG:pyvips:Falling back to ABI mode
DEBUG:pyvips:Loaded lib < object at 0x7f29fa015190>
DEBUG:pyvips:Loaded lib < object at 0x7f29fa015110>
DEBUG:pyvips:Inited libvips

pyvips will work fine in this fallback mode, it’s just a bit slower.

If API mode stops working, you can fix it by reinstalling pyvips. You should make sure pip is not reusing a cached wheel, e.g. by using pip install --no-cache-dir pyvips.


This sample program loads a JPG image, doubles the value of every green pixel, sharpens, and then writes the image back to the filesystem again:

import pyvips

image = pyvips.Image.new_from_file('some-image.jpg', access='sequential')
image *= [1, 2, 1]
mask = pyvips.Image.new_from_array([[-1, -1, -1],
                                    [-1, 16, -1],
                                    [-1, -1, -1]
                                   ], scale=8)
image = image.conv(mask, precision='integer')

Converting old code

To convert old code, replace the lines:

import gi
gi.require_version('Vips', '8.0')
from gi.repository import Vips


import pyvips
Vips = pyvips

Instead of the pyvips = Vips, you can of course also swap all Vips for pyvips with eg.:



The Python binding included in libvips works, but porting and installation are more difficult than they should be.

This new binding is:

  • compatible with the current Python binding (it runs the same test suite, unmodified)
  • easier to install, since the stack is much smaller, and there are no issues with the overrides directory
  • faster, since we implement Buffer and save some copies
  • faster, since it is “thinner”. The ffi Ruby binding is about twice as fast as the gobject-introspection one, when running the test suite
  • portable across CPython, PyPy and others
  • more simply portable to Windows
  • easy to package for pip


Local user install:

$ pip install --user -e .
$ pip3 install --user -e .
$ pypy -m pip --user -e .

Run all tests:

$ tox

Run test suite:

$ tox test

Run a specific test:

$ pytest tests/


$ tox qa

Generate HTML docs in doc/build/html:

$ cd doc; sphinx-build -bhtml . build/html

Regenerate autodocs:

$ cd doc; \
  python -c "import pyvips; pyvips.Operation.generate_sphinx_all()" > x

And copy-paste x into the obvious place in doc/vimage.rst.

Update version number:

$ vi pyvips/
$ vi doc/

Update pypi package:

$ python sdist
$ twine upload dist/*

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
pyvips-2.1.5.tar.gz (30.2 kB) Copy SHA256 hash SHA256 Source None

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page