Downloading Patreon images seems to yield a smaller file size than if I download it myself

[colbert2677]

These are from the same post; the top one is when I save the image myself directly from Patreon and the below is if I download it via gallery-dl. The resolution and file type are the same but gallery-dl seems to grab a file with reduced file size, so I'm not confident that I'm getting the exact same file from gallery-dl as if I were saving it manually from Patreon.

Are there configurations or otherwise that I'm missing out on here? Could not seem to find a relevant issue or discussion to this.

[mikf]

The extractor.patreon.files option might be relevant here.

[colbert2677]

I've tried every single option and mix of options in that list. I'm still seeing file size loss compared to if I save the image directly from Patreon myself; in another case, I got 4.36MB when saved myself vs 2.92MB when downloaded via gallery-dl.

[mikf]

I did a bit of investigating and found that Patreon image URLs contain a segment that has base64-encoded "quality" options.

For the big image files labeled by the API as "original" under "image_urls", it is
{"q":100,"webp":0} (/eyJxIjoxMDAsIndlYnAiOjB9/)

https://c10.patreonusercontent.com/4/patreon-media/p/post/12497641/3d99f5f5b635428ca237fedf0f223f1a/eyJxIjoxMDAsIndlYnAiOjB9/1.JPG?token-time=1734307200&token-hash=lEeSOpQzapNPL_1Iy0CQnUo1nRGmVigb62yyWxX898U%3D

For the smaller files downloaded by gallery-dl labeled "download_url", it is
{"a":1,"p":1} (/eyJhIjoxLCJwIjoxfQ%3D%3D/)

https://c10.patreonusercontent.com/4/patreon-media/p/post/12497641/3d99f5f5b635428ca237fedf0f223f1a/eyJhIjoxLCJwIjoxfQ%3D%3D/1.JPG?token-time=1733184000&token-hash=sC3NXeqoyGDwVO7JdT_o6bc8NGiIOPmwBK1NbJQ4Cd8%3D

The pixel value information of both versions is identical for PNG files, but there is a difference for JPEGs. I am not sure if the "q":100 versions are actually of higher quality or if they are just re-encoded with a JPEG quality level of 100. Lots of the extra file size is also used up by additional(?) image file metadata.

Changing these "quality" options is not possible, as that invalidates the token-hash value.

[colbert2677]

Hm... that's interesting. Appreciate you taking the time to dive deep here. It seems like a difference I'll have to live with if I'm trying to acquire a large amount of posts off Patreon via gallery-dl as far as I care about the actual files, but in terms of just viewing them, it should be good enough already that the resolution is retained. Besides Patreon, things like Pixiv yield the files I expect as though I downloaded them myself, so really solid tool here. Cheers.

[mikf]

Having gallery-dl download original images instead of the download_url version would be quite easy. I'm probably going to add an option for this.

I'd like someone with better vision than me (mine is terrible) to compare the two examples from #6569 (comment) in terms of quality, if that's possible.

[mikf]

a526a3d