Determine if WebP is lossy or lossless

[avalonv]

I'm developing an application for optimizing comics and knowing the original format is somewhat important, since converting from lossless to lossy formats can often increase file size. Currently I try to convert both and then compare their sizes, but is there a built-in way to check?

[radarhere]

A built-in way? No.

However, I saw

Pillow/src/PIL/WebPImagePlugin.py

Lines 17 to 29 in aedbc77

	_VP8_MODES_BY_IDENTIFIER = {
	b"VP8 ": "RGB",
	b"VP8X": "RGBA",
	b"VP8L": "RGBA", # lossless
	}
	def _accept(prefix):
	is_riff_file_format = prefix[:4] == b"RIFF"
	is_webp_file = prefix[8:12] == b"WEBP"
	is_valid_vp8_mode = prefix[12:16] in _VP8_MODES_BY_IDENTIFIER
	if is_riff_file_format and is_webp_file and is_valid_vp8_mode:

and think that the following should be able to tell you if the WebP file is lossless or not.

with open("Tests/images/hopper.webp", "rb") as fp:
    lossless = fp.read(15)[-1:] == b"L"

[avalonv]

Would it help if I added a sample file for you to test on your machine?

[radarhere]

This is also mentioned in the specification, so I'm surprised that it isn't working for you.

If you could attach sample files, that would be great. Also if you could mention how you're figuring out whether they are lossy or lossless.

[avalonv]

Well, I'm fairly sure they're lossless because I'm doing batch image processing which checks the size of files at the end, and ones generated with the Image.save() lossless parameter set to true are clearly larger (and they don't really change size on repeated saves).

Here's a file https://www.dropbox.com/s/jx5ti5oavookqil/berk2.webp

Here's the relevant line I used to create it:

https://github.com/avalonv/reCBZ/blob/bee82c61c5864f8c753567c9722a3715fc93a696/src/reCBZ/formats.py#L45

[radarhere]

Oh. I think I just made a simple error in my first comment, writing 15 instead of 16.

Try

with open("berk2.webp", "rb") as fp:
    lossless = fp.read(16)[-1:] == b"L"

[avalonv]

c>>> with open('berk2.webp', 'rb') as fp:
c...    lossless = fp.read(16)[-1:] == b"L"
c... 
c>>> print(lossless)
True

You're a lifesaver, thank you!