Which image formats Pillow can read and write

The ones you will actually use

Format	Read	Write	Worth knowing
JPEG	Yes	Yes	No alpha. Loses a little on every save. `quality` and `optimize` are the options that matter
PNG	Yes	Yes	Lossless, supports alpha. `compress_level` trades time for size
WEBP	Yes	Yes	Alpha, animation, lossy and lossless. Usually smaller than both of the above
GIF	Yes	Yes	256 colours, animation. Use `save_all` and `append_images` for frames
TIFF	Yes	Yes	Many internal variants. Some compressions need libtiff built in
BMP	Yes	Yes	Simple and enormous. Fine as an intermediate
ICO	Yes	Yes	Holds several sizes in one file. Pass `sizes=` when saving
PDF	No	Yes	Write only. Pillow cannot open a PDF, only produce one from images
PPM, PGM, PBM	Yes	Yes	Trivially simple formats, useful for debugging

Common, with a condition attached

Format	Read	Write	Condition
HEIC, HEIF	With a plugin	With a plugin	`pip install pillow-heif`. This is what iPhones produce
AVIF	Recent versions	Recent versions	Needs a build with libavif. Check before relying on it
JPEG 2000	Yes	Yes	Needs openjpeg at build time
SVG	No	No	A vector text format. Rasterise it with cairosvg first
RAW, CR2, NEF	No	No	Camera raw. Use rawpy, then hand the array to Pillow
PSD	Flattened only	No	Reads the composite image, not the layers
EPS	With Ghostscript	Yes	Reading needs Ghostscript installed on the system

Asking Pillow itself

Builds differ, so the reliable answer is from the copy you have installed:

from PIL import Image, features

print(features.pilinfo())

Image.init()
print(sorted(Image.OPEN.keys()))
print(sorted(Image.SAVE.keys()))

pilinfo prints the version, the compiled in libraries and which codecs are present. It is the first thing to run when a format works on your machine and not on the server.

Checking one feature

from PIL import features

print(features.check("webp"))
print(features.check_codec("jpg_2000"))

The format is not the file extension

im = Image.open("photo.jpg")
print(im.format)

im.format is what the file really is, read from its header. Plenty of files are named .jpg and are actually PNG, which matters when something downstream believes the name. See cannot identify image file.

Saving animations

frames[0].save("out.gif", save_all=True, append_images=frames[1:],
               duration=80, loop=0)

The same pattern works for animated WEBP and for multi page TIFF and PDF.