@dstillman Thank you very much for this new addition. I have been testing it and it is working very well. I especially appreciate the possibility to adjust or create one's own theme.

For example, I liked the sepia theme, but I was able to darken the foreground a bit for more contrast. Thanks!

Customizing themes with just foreground and background colors is simple, and the darkening of images in dark mode is a useful improvement. Looking forward to (https://www.concretesrichmondva.com/) how the full-page image theming develops!

Thank you ! It works great. A minor issue : highlighted text seems to keep the light theme. For instance in papers, the citations appear very strongly in dark mode (example here: dx.doi.org/10.1016/j.jclepro.2016.05.077).

In the latest beta, we've added syncing of custom themes between devices.

Thanks a lot for this new feature!

That said, would it be possible to keep the option for the old dark mode settings? I’ve gotten used to them over the years, and personally, the inverted colors worked really well for me. Would love to have the choice to stick with the old version!

+1 for keeping an option for old dark mode settings. Inverted colors for images is much preferred for scanned pdfs.

Hi there. This seems to be working as expected with PDFs, but with snapshots it only shows the "Split View" option and nothing by way of themes. Is it yet to be implemented for snapshots?

No, no, NOOOO! (insert Michael Scott screaming GIF)

I'm sorry for the strongly-worded comment, but it is what it is: this new functionality is ABYSMAL for scanned book/document PDFs. It completely borked the dark mode content view for anything but clean, all-digital-workflow PDFs. I've tried all darker themes (and even a custom #000000 / #FFFFFF one) on some post-processed documents with transparent backgrounds and they looked like, well, outright illegible crap (the text is much too dark to even register as such), and in other vanilla, untreated documents with straight JPG/PNG facsimile content, it plainly won't work, at all, it will remain mostly light or become slightly darker.

I can show you screenshots if you wish, but unless you figure out some sort of AI-level wizardry to separate text and images from backgrounds and process them accordingly, this feature is just unacceptable as-is for those, it's fundamentally incompatible with a large portion of content that worked just fine-ish before (yes, I know, colour images and even some B&W stuff looked weird inverted, but we could live with that and turn it off if needed).

Please, please, PLEASE bring back that old dark mode engine/setting, as a basic “invert content” mode (maybe even on a sticky, per-item basis, as that would be great for all those users who have a varied mix of documents). You can please everyone – yes, I understand that other people may wish to theme their digital PDFs to their hearts' content, and that's fine, I'm not asking you to revert the whole thing to what it looked and worked like and likely *will* make use of this themeability myself on the documents that are indeed compatible with it – by just adding it to the list as an uneditable “theme” of sorts, if you must.

I love the new annotation display and search functionality, but I'm now basically forced to choose between having it and burning my retinas with the normal light content mode, or eschewing it altogether and being able to read old stuff – of which I'm pulling quite a lot of citations lately. And in the long run, I may have to keep the current release version of Zotero running on my main machine, with the big dual screens where I read stuff, or to resort to looking for plugins or other shenanigans to revert this behaviour… This is an egregious regression if I've ever seen one, and that's how dependent I became on the old dark content mode (for context, on top of extremely aggressive f.lux settings which I still use to this day much to the chagrin of bystanders, I used to invert my entire Macs' screens with a keyboard shortcut before system-level dark mode was even a thing, and it included, back then, Mendeley and Microsoft Word, which made a big difference on how long I could work on my dissertation each day)… I don't even have any serious eyesight disability, just a recurring but relatively mild one that does trigger photosensitivity when it flares up, but please consider the users who have it even worse.

Okay, so here are a couple of examples:

https://s3.amazonaws.com/zotero.org/images/forums/u6505448/n1n5mjmej3ouwx7hhpp1.jpg
This is a scanned document, with the default light theme.


https://s3.amazonaws.com/zotero.org/images/forums/u6505448/efosux3dcz8mgog0pj7c.jpg
This is the same document, with my custom #000000/#FFFFFF theme (but all other dark themes will produce the same slightly darker result).


https://s3.amazonaws.com/zotero.org/images/forums/u6505448/ciuyecpxl0znmc0n5qcs.jpg
This is another document, with the default light theme (it was highly processed, but doesn't have a transparent background).


https://s3.amazonaws.com/zotero.org/images/forums/u6505448/nhmkc1xnl4fgffdq4rdc.jpg
This is that second document, with my custom theme.


https://s3.amazonaws.com/zotero.org/images/forums/u6505448/ohcusqh6ql8ux4vklv4n.jpg
This is a third document, with the default light theme; it doesn't have a white/transparent background, but the text layer does seem to have an alpha channel.


https://s3.amazonaws.com/zotero.org/images/forums/u6505448/9wdw2spu9vl9j0k07ol2.jpg
This is the same document, with the new default dark theme (as you can see, it is, indeed, an unreadable mess).


https://s3.amazonaws.com/zotero.org/images/forums/u6505448/2qz5x78ltgmemb76nx8c.jpg
And finally, this is the same document, with my custom theme (yes, legibility improved a bit, but it seems the text stays the same gray colour, and a full-on inversion would work great).


For sure, some of these documents may be a bit non-standard or whatever, and I can send you them for analysis, but they exist in people's libraries, have good contrast on light mode and had the same decent contrast when inverted, and now they range from unpleasant to borderline unreadable.

IMHO, the easiest way to accommodate these edge cases is to just bring back basic inversion as an option instead of trying to parse their contents/layers, as I believe the PDF standard is just that messy and complex to begin with.