Zombeaver

I've had decent success with the upscaling efforts so far. You can find some examples below. You can click the drop-down to select different images to compare. https://imgsli.com/MjYxMzcx/0/1

Thank you for the comparison, that's good to see as an example. I don't think either one looks particularly great honestly. Both of them look what I would describe as "weird" on quick glance. It would be pretty obvious to me that it had gone through some kind of digital tampering in both cases. The other two less so. I wish there was a model that was specifically trained for text, but to my knowledge there isn't one.

This looks similar to the results from Nomos8kSC in Upscayl and demonstrates some of the same font weirdness. This is particularly evident in section 4 of that page. Nomos8kSC for comparison: I almost wonder if it wouldn't be beneficial to run some of these through multiple models in sequence with each at smaller integer scales.

Yeah that's fine and to be expected. I'm used to that kind of thing already. There have been some cases where something was so bad that I just completely recreated it from scratch using closely approximated fonts/colors, etc. Really good tip here! Doing some quick testing 4xHFA2k and 4xNomos8kSC seem to be yielding the best results so far. 4xHFA2k has been consistently a bit better between the two though, with less weirdness in font interpolation. That's good to know! I'm not totally opposed to dropping a one time amount of $100 on something like this if I need to. In no dimension would I ever pay $10-$30 a month until the end of time for a finite number of images per month though. That's an insane pricing structure. I appreciate the tips here from everybody! This definitely points me in the right direction. I'm liking Upscayl quite a bit so far. It's really easy to use, has a nice UI, is fast, yields pretty good results so far, and is free. Some examples of some lower quality material: 4xHFA2k: The second one here seems to be the most problematic, with a fair number of oddities in the text. A few things here and there in the Lunar Leeper one too, but overall not too bad. I probably would use the first and third one here as replacements.

Someone on my Discord just recommended Upscayle which is apparently free and open source. I'll be giving that one a shot tonight. https://www.upscayl.org/

I have Photoshop. I'd like to avoid another monthly subscription just for this so imgupscaler is probably out (and as a side note I seriously hate this pricing structure that a lot of these sites have of pay per number of images processed, that entire model would be unusable for my needs). I've heard of topaz, and looking into it a bit it looks like you can pay once and then use local software, which is much more appealing to me. I'm a little concerned about the fact that it says includes 12 months of updates though - does the software outright stop working after 12 months if you don't pay again or does it just not receive updates anymore (this is implied but I want to be sure)? I'd be leery of the former but if it's the latter that could work. Definitely! I'll send some examples in a couple hours. Thanks for the offer!

Thanks @C-Beats! Hopefully one of them will have some ideas here. I would think there's probably some option out there that would work well here. I've messed around with a few different free options online and didn't have great results, but admittedly I don't have a ton of experience with this.

So I asked this over on my Discord but I'll mention it here as well in case anyone has any ideas: Does anyone know of a good option for AI upscaling that would yield good results for manual scans (primarily text, but some art as well)? Part of the process for each update is upgrading manual scans wherever possible. A higher quality scan is obviously the preferred method for this, but because of the obscure nature of this stuff that won't always be an option. It seems to me that AI upscaling could be a good option for this but my experiments with it thus far have yielded mixed results. In some cases it's a bit of an improvement, in others it looks a bit like using one of those awful "Eagle" shaders on a SNES game (a smeary mess). It seems like with as much as this tech has advanced in the last few years that surely there has to be a decent option for this kind of thing by now. If it's free that would be preferred so that I could work on them myself, or if not and someone has access to a relevant platform and would be willing to volunteer some time to work on it I can supply the media in question. I'm open to any ideas here. I realize it's probably not going to work in every scenario, but if this is an avenue to improve some of them I'd like to do it.

This is how it works currently, yes. Each subsequent release is complete. There is no update process per-se at the moment as there's too much in flux currently from one version to the next and I don't expect this to change until at least v1.0. It requires a complete delete and replace. There have been multiple instances of significant backend changes over the course of the project (and this will be true in the next release as well) so it would be impractical to do it any other way until that's no longer the case. The relevant folders for the purposes of importing would remain the same as those I listed above though. It would be those plus the main C64 Dreams subfolder of course. If you have any questions or need help with anything please feel free to hit me up on my Discord. That's where I typically post about ongoing progress and collaborate with other developers like StatMat of the OL64 project.

Yep, I can confirm both of these are true. Both are corrected in the current WIP. Good catch. It would probably make more sense to simply pull everything from Data\Platforms and Data\Playlists as it will always be everything in both folders. The same goes for Images\C64 Dreams, Images\Platforms, Images\Platform Categories\, Images\Playlists, Music\C64 Dreams and \Plugins. It will always be everything in those folders regardless of what's specified (or not) in Files to import.txt. I will probably end up changing Files to import.txt to eventually just state that anyway just to simplify matters. The Parents and Platforms xml inserts would of course remain as they are by necessity. I appreciate the work on your app. Importing into an existing library has always been an overly fiddly process that's been more difficult than I would like for people to deal with so if this helps streamline that that's great. The next question would be is it able to deal with subsequent releases when existing (and potentially outdated) data is already present from a previous one? The process has to be repeated when a new version is released. In the case of the individual xmls for the platforms and playlists it's a simple overwrite. With images it's not quite that simple because sometimes files end up getting deleted between releases, or change filetypes (so they wouldn't be overwritten and an unneeded image is left behind if the existing folder isn't deleted beforehand). The inserted portions will change as well, usually just with new data, but sometimes with altered existing data. I'm not sure how difficult that would be to deal with - I would imagine a differential of old to new image files wouldn't be too horrendous albeit time-consuming (and it could always just outright delete the relevant folders beforehand instead, since that's the current process when doing it manually), but the Parents.xml and Platforms.xml changes I would expect to be more difficult since those would contain a lot of non-C64 Dreams data when incorporated into an existing library. Maybe it could just do a diff between the old and new insert files, then compare against the user's existing xmls and then update/add based on that? Sorry, I know this is a bit in the weeds but I just want to make sure that it wouldn't implode when someone attempts to import a new version on top of an existing one.

So, looking at your video again, it looks like the icons for them aren't showing up which indicates a cache issue. I don't think I've ever seen that result in a completely ignored playlist before but admittedly I don't use BB a ton to begin with so maybe it's possible. You can refresh the BB cache manually via System Menu > Options > Image Cache; either that or you didn't pull the image files for them over. I assume it's the former though just based on the way it appears. This is why I said I would recommend testing in LB first. I wouldn't mess with BB at all until you've done that (note that LB needs to be set to Platform Category view to display correctly with C64 Dreams). If that doesn't fix it then I have to assume that something has gone wrong with the import process, either something wasn't pulled over or the sections that need to be added to the parents and platforms xmls weren't added completely/correctly. You'll also want to double check that all of the playlists have been pulled over - Zzap isn't technically a platform, it's a playlist. It's C64 Dreams (Platform Category) > Magazines (Platform) > Zzap!64 (Playlist). The implication if Zzap, specifically, is missing would be that either the Zzap playlist wasn't pulled over or that the parents xml wasn't updated correctly to include it under the Magazines platform.

It's also worth noting that based on your previous demonstration you're not actually using the XML that defaults to the local versions of the magazines because it's loading up in a browser. I'm not sure if this was intentional or not since you show the relevant folder where the local one is, but to be clear the Default Local one needs to be used if you want the default launch behavior to use the local files (which I assume you do since you downloaded them).

They definitely work when imported into an existing collection when following my instructions.

I'm not sure what you mean by them being in Big Box. The manuals aren't integrated into LB/BB in any way (unless BB is finding them and doing something with them in the background which I'm unaware of, its unintended if so). As I said, they're toggleable in-game via the hotkeys. Numpad minus + numpad enter on keyboard or back + right thumb stick button on controller. You can navigate them via arrow keys on keyboard or right thumb stick on controller. Pressing the keyboard or controller hotkey combo again brings you back to the game. Viewing them via LB/BB is not intended. The reason for this is because, functionally, everything is frontend agnostic/non-dependent. All functionality from manual toggling to custom music for text adventures, etc. is handled via external scripts and apps that use relative pathing. From Launchbox's perspective it's essentially just Windows shortcuts. You could remove LB/BB from the equation entirely and it would still be fully functional by just starting the .vbs files in each game folder, magazine folder, demo folder, etc. This is by design so that someone could theoretically import into any frontend and it would still be functional. I don't know what the issue is with the magazines beyond something has gone wrong with your importer since by your own demonstration they work in the normal version. I know that BB itself has its own separate cache, so that could be related. I would recommend checking in LB first.

I'm not exactly sure what you're asking. If you mean in general, yes. If you mean with your importer I don't know as I haven't tested it. Manuals are toggled in-game via the keyboard/controller hotkeys. Magazines default to the web version but this can be changed if you replace with the included xml in the "Default Local" folder in the magazines subfolder. I appreciate the contribution.