>>23506 Apple *spits* is kinda pushing that direction with their ARM desktops that have unified memory. If you have 256gb of shared ram-vram, then it stops mattering as much. >syncing files between devices This should be built into the foundation for kinda competing with ehentai's hentai@home software. I don't see any point in letting my phone sync with my desktop, if it also isn't applied more generally. Internet censorship, boorus restricting access due to scraping tools, etc. >>23492 Just use VNC >>23517 Are you misusing a watcher, or maybe I'm confused? I use a gallery download page to get everything from someone, then once the download is done add them as a subscription. 1 subscription downloader per site, and each downloader filled with 1 tag subs for artist names.

>>23521 >>23523 okay, I ran the DB integrity check overnight and it says that there were 0 error found. I guess that means that my DB is actually fine. so hopefully that means that I'm safe, but then I don't know what that error is about or what's causing it.

>>23525 It's saying it ran into some search objects different from what it expected and they couldn't be loaded. You probably just lost a single page in the session, no actual data issues.

>>23521 >>23523 >>23525 Hey, I'm sorry for the error and related panic here. Your database is fine, this is all me. It looks like there is a dumb bug when loading a system:rating predicate that is inside an OR predicate too early in program boot. I'm not sure if this would be a database update step or some initial part of the Main GUI load. If any of your sessions or favourite page loadouts or anything hold something like an OR with a rating pred inside, that is probably what is being hit here. It may have deleted/ignored that page or that pred from the active search predicates on that page, but everything else around it probably loaded ok. I've fixed the bug for v628. I'm doing more ratings work tomorrow so I'll give it another look just to be sure. Please let me know how v628 goes.

Could it be possible to add export patterns for modified times and such? I can just write a little script to do it myself since the exported files have the proper times but it would be nice if Hydrus could do that.

I had a very good week. I cleaned a ton of code, fixed bugs, and improved some quality of life. There are several more rating UI features, too. The release should be as normal tomorrow.

>>23526 >>23527 thanks guys. it's good to hear that it's nothing serious. I'll update when v628 is released and see if that stops happening then

I don't know if it's just me, but it seems like in version 627 if you use the keyboard to switch tabs, the keyboard focus will sometimes not be put on the file thumbnails.

https://www.youtube.com/watch?v=ByMrACzUCzk windows zip: https://github.com/hydrusnetwork/hydrus/releases/download/v628/Hydrus.Network.628.-.Windows.-.Extract.only.zip exe: https://github.com/hydrusnetwork/hydrus/releases/download/v628/Hydrus.Network.628.-.Windows.-.Installer.exe macOS app: https://github.com/hydrusnetwork/hydrus/releases/download/v628/Hydrus.Network.628.-.macOS.-.App.zip linux tar.zst: https://github.com/hydrusnetwork/hydrus/releases/download/v628/Hydrus.Network.628.-.Linux.-.Executable.tar.zst I had a great week cleaning old code and fixing/improving small things. There are also more neat ratings UI options. This release folds in the library updates we tested last week. The test went well and there are no special install instructions, but let me know if you have any trouble. Full changelog: https://hydrusnetwork.github.io/hydrus/changelog.html ratings Thanks to a user, we have another round of ratings UI features. A new options->ratings page lets you edit the size of ratings in every location, including separate sizes for inc/dec ratings, and in the 'manage ratings' dialog, and you can now choose to collapse the thumbnail numerical ratings display to just '3/5 (one-star)'. Under services->manage services, you can now add this '3/5' text to the left or right of any normal numerical ratings display. Also, thanks to another user, we have a bunch of new default ratings svgs. I think they look great. Also, under options->media viewer, you can now choose to show the 'top-right' hover window in the preview viewer! So, you can now quick-set ratings with one click. Try this out--I like it so much that I think I'll soon make it default for new users. duplicates quality of life Duplicate pages now count up their searches' pairs with my new incremental duplicate fetch tech. Touching a duplicate page will no longer lock your database for a couple seconds, and there's pause and fast cancel-reset tech in there too. Should just work better all around, but let me know if it ever under-counts. Semi-automatic duplicate rules similarly commit large approve/deny commands in pieces, rather than blocking the database for the whole job, and they update the UI with a 'approving: 16/55' label so you can see how they are doing. There's also a 'select all' button. next week I've got a lot of ideas, but I would like to get the incremental duplicate fetch tech working in the main duplicate filter fetch routine and the 'show some random' button too. I'd love to reduce the fetch time when I only need a handful of pairs, and if I can get this all working on the same unified and sane pipeline, I can finally think about changing the order pairs arrive.

>>23537 I'm glad you added that fraction beside ratings option, but in darkmode it's unreadable. it shows up for me as black text on a dark grey background. Is there a way to change the color of the fraction?

>>23538 I figured it out. for anyone else having this issue, it looks like the "border for liked" color is reused for the color of the text. it's a bit confusing because I'd usually want that to be a dark color, but I guess I gotta make it light if I want to see the fraction. not too big of a deal, just a bit strange.

Hell yeah rule34.xxx downloader is working again. The site has horrible tagging, but hosts stuff that gets struck down from other boorus. >>23537 TY dev. Quick question: how does Hydrus currently handle duplicates? If I subscribe to the same artist on derpibooru and e621, how is that handled? Or is there automatic md5 deduplication, and the duplicate filter is only for duplicates that don't share a md5 hash?

I found a bug with version 628. it looks like in the archive delete filter, the top right media viewer hover isn't allowed to pop up, so you can see it and click on stuff like links and ratings. it's only in the archive-delete filter, and it happens even when you have to check box to stop the top-right from popping up disabled. notes do still pop up though

>>23540 yes. if files are hash-identical (i.e. all same bytes and size), only one copy is stored. otherwise you have to run a job to find files that are visually identical/similar (duplicates processing) and then decide what to do with them (duplicates filter)

Maybe of interest to image-file-format enthusiasts: People are sharing images with a special embedded 'chara' exif metadata field. These can be loaded into many AI/LLM programs to use that personality. There are many, many image sites hosting these things. All of them are flakey and it's a good idea to start archiving cards you use. Hydrus currently can filter for files with 'embedded metadata', but there's no way to find all files with the 'chara' embedded text. The data is encoded too, so viewing it requires manual external action anyway. Maybe important to note is that this data most often includes NSFW text even if the image itself is tame. I've also seen embedded links to nsfw images, with some UIs autoloading these. Of course there's also ones with legally questionable text. specifications for the exif data: https://github.com/malfoyslastname/character-card-spec-v2/blob/main/spec_v1.md (including description of embedding method) https://github.com/malfoyslastname/character-card-spec-v2/blob/main/spec_v2.md 4 example images, hopefully

>>23542 TY >>23543 Interesting. ComfyUI and other AI generation programs also embed AI workflow data into .png files. I feel like this is a bit out of scope for hydrus directly though. It's built for booru scarping & management. I wanted to use it for doujins or movies, but it isn't built for it. Ideally if you wanted to use hydrus for AI template management, it would need a specific downloader for that site, and you would run it in a separate instance IMO.

>>23544 >out of scope >It's built for booru scarping >The hydrus network client is a file-management application written for internet-fluent media nerds who have large file collections. Sorry for sharing the downloader I made I guess. I didn't know I was using hydrus in a way some anon didn't like.

>>23543 we've been over this last thread, including Stable Diffusion and ComfyUI embeddings.

Hydev, would it be possible to run a parser/downloader import from within hydrus? Like I have this pic anon posted just now and it would be pretty neat to be able to actually import it from a right click menu or something.

>>23543 >>23545 As someone who has recently started collecting character cards from Chub and JannyAI, I find this extremely useful. I've been adding mine to hydrus with local tags by hand. Making downloaders it beyond my knowledge, so I for one am glad you shared it. This will save me so much time.

>>23551 Oh, turns out you can drag and drop them onto lain and it works. I thought that the importer menu was one of the menus that blocks all other inputs.

>>23545 Sorry if I wasn't clear. It's cool, but I'd only run the AI card downloader in a dedicated instance, because it would be too easy to get it mixed up in random booru images. Also it wasn't clear you made a downloader, I thought you just posted the metadata spec. Sorry if I'm retarded!

Is there a way to blacklist combinations of two or more tags? Example, I want to blacklist files that contain both: 'creator:1' AND 'badtag' together, but I want to let files with 'creator:1' OR 'badtag' through. Thank you! But

>>23508 Yeah, on the main github repo here: https://github.com/hydrusnetwork/hydrus/pkgs/container/hydrus It builds every week with the other builds. I don't know much about Docker though. There's a guy who uses it for loads of stuff who figured out the Dockerfile build scripts here: https://github.com/hydrusnetwork/hydrus/tree/master/static/build_files/docker >>23513 Yes, thank you. Something is messed up when re-loading a fresh mpv window from stasis in that mode. I'll give it another go. >>23514 Are you completely sure the PNG here doesn't have an ICC Profile too? I did a bunch of work to get gamma&chromaticity pngs to render with correct colours about a month ago, and while a lot of this code is related, I don't think the actual file test overlaps. My 'has icc profile' flag is calculated basically by loading the png with Pillow and then seeing if Pillow provides an icc profile in the 'info' of the loaded object. If a png only has the gamma&chromaticity stuff, it doesn't have the 'icc_profile' info. If a png has ICC Profile bytes, I apply it if the user is set to currently apply icc profiles. If a png doesn't have an icc profile, or if the user has set not to apply icc profiles, I then see if it has an sRGB chunk, in which case I do nothing since hydrus is sRGB, and then I see if it has gamma&chromaticity, at which point I apply that. I'm thinking of adding flags like 'has sRGB chunk' and 'has gamma & chromaticity chunks', so we can search them, but I don't think they overlap with the old ICC Profile stuff. Maybe they overlapped in the past? What happens if you find some of these pngs and do manage->maintenance->determine if file has an icc profile? Do they fix themselves? In any case, if you have weird files, I'm interested in seeing them. If you can point me to some, or post a catbox zip or email me or whatever, I can check them on my end. That said, much of this code is a mess with old debug hooks in it and I'd like to rework it. If a user has 'apply icc profiles' off, then any thumbnails are generated with wrong colours, which I don't want. I want all those options to happen higher up, closer to the user flicking back and forth different render modes. I've also never been completely fluent in how these systems are 'supposed' to work, either, so feedback from users like you is great. >>23530 You have to do it with sidecars at the moment. Add a sidecar to your file export and then select 'time' as the source for the sidecar. It is awkward like the rest of sidecars, but you can do file or domain modified time. I do want to completely overhaul export filename phrases to handle stuff like times (and more gracefully handling non-present data). If I screwed my head on straight, it'd probably inherit most of the sidecar toolkit.

>>23534 I don't think I changed anything. There's a setting under options->gui pages->navigation drag and drop that lets you always move the focus to the primary 'text input' (like the autocomplete dropdown) of a page on page change, but I don't think I've ever had it for the media side of things. Afaik Qt just remembers whatever the previous focus was, although if you are using Ctrl+Tab there might be situations where the focus moves to the page tab itself. I added some shortcut actions to the 'the main window' shortcut set for 'move page selection left/right', that you might like to try mapping to something else, if there's Qt focus remapping going on in your OS. I think Ctrl+M by default moves the focus to the thumbnail grid, if that's helpful. You can remap that too, also under 'the main window'. I often do Ctrl+M->Ctrl+A for stuff. >>23538 >>23539 Thanks, I'll look into it. It should render the same as other text there, I bet some pen colour isn't being reset amidst the drawing routine. >>23540 Here's some formal help on this: https://hydrusnetwork.github.io/hydrus/duplicates.html And this is the thing I've been working on a lot this year. It isn't ready for normal users yet, but we are getting there: https://hydrusnetwork.github.io/hydrus/advanced_duplicates_auto_resolution.html That first help page is pretty old and patchwork now. Since you are new to the system, please let me know if anything is confusing or if any of the screenshots are so crazy old they don't make sense etc... >>23541 Yep, sorry--this is driving me nuts too. Something went wrong with the preview viewer top-right hover integration. I will absolutely fix this for next week. >>23543 Thank you, very useful information. I've got these images now. I plan to expand the 'human-readable embedded metadata' recognition to have a db cache, so you'll eventually be able to search for common stuff (and the same system will probably let you search for EXIF tags too), so these are great to work with so I know what I'm aiming at.

>>23553 Haha, didn't know this worked! The copy/paste bitmap is also a good quick solution. I've thought about recognising these better. I may, one day, since there's a bunch of other gubbins you can export to pngs. My whole serialisation system works on it, and most multi-column lists with import/export/duplicate buttons should support a similar drag and drop import. It might be nice to have a generalised 'hey this thing contains a [tag filter]--want to import to [tag filter favourites]?' right-click menu or import panel. >>23556 No, sorry! The tag filter code needs an overhaul to improve performance and simplify some things. Once it is nicer, I may be able to add logical algebra like this, and wildcards are another highly requested feature, but I will be careful to say that this stuff is often a lot more tricky than you think, so I will not promise it. For now you pretty much have to have a separate favourite tag filter that applies to artists x y z and then sequester those guys in their own subscription with separate specific tag import options. Or, if we are talking only five files, set up a complicated file search that covers these bad files and then just ctrl+a->delete that search every month. Another planned overhaul is to separate the blacklist from the tag import options in every importer, which should allow for simpler import custom setup.

>>23541 This is now fixed in master. If you run from source, just git pull and you are fixed. Otherwise, it will be rolled into v629. Sorry for the trouble!

The derpibooru/ponerpics/twibooru parsers should probably replace "character:tank" → "character:tank (mlp)" "character:tofu" → "character:tofu (mlp)"

>>23561 >"character:tofu" → "character:tofu (mlp)" That's a rare one though. Tank is not.

>>23561 I use "character:(tortoise) tank" plus the MLP tag "show:(mlp) - my little pony"

>>23566 Oh, right. I thought I ought to be missing something.

v626. tag banners (on the top of the thumbnail) show both the ideal sibling and the original tag instead of the ideal sibling only. Is there a setting for that?

>>23560 yeah I do run from source and I pulled. Thanks for the fix!

>>23558 >although if you are using Ctrl+Tab there might be situations where the focus moves to the page tab itself. I added some shortcut actions to the 'the main window' shortcut set for 'move page selection left/right', that you might like to try mapping to something else, if there's Qt focus remapping going on in your OS. So I tried to do this, and I ran into a bug that's blocking me. For some reason, Hydrus refuses to recognize any shortcut that uses "Shift+Tab". Instead, it always just sees it as "Shift+Nothing". I checked with other software and it works fine there, so this isn't a problem with my keyboard. Hydrus also recognizes "Shift+Tab" when I'm using the default "Ctrl+Shift+Tab" to move left one page. (which btw, I can't find those default page-change shortcuts at all in the menu. I looked everywhere. I don't see them.) I don't know why Hydrus won't see the tab being pressed specifically when shift is pressed when I try to make a shortcut, but I can't change it due to that. The "focus to thumbnail grid" is helpful though, thanks! I didn't know about this shortcut.

>>23351 >>23352 turns out I was probably 15 or so versions out of date, which was causing the issue...

I had a great week mostly fixing bugs, particularly issues with the top-right ratings hover, export filenames, date parsing in the builds, and duplicates auto-resolution work timing. There's also improved AVIF support. The release should be as normal tomorrow.