I've Created a WC4 HD Video Pack

I took a look at the "source" videos that I could pull off the server and It looks like the original capture/conversion was just kind of bad. I did come up with some decent bandaids for it and hopefully we can improve them at least a little. It's just unfortunate that the original interlacing -at least in half of the videos - was baked into the progressive transfer,
Has the video that was posted by Steve Hilliker back in May been considered? That was supposed to be using Beta SP tapes. https://www.wcnews.com/news/2020/05/20
It would be great except that he didn't include everything. It would be great if we could get everything at that quality for sure. There's also some other edits and transitions that dont make it impossible to use, just quite incomplete
 
It's an option, isn't it? (Also not sure if it's worth contacting Mr Hilliker directly about this.) I remember there were a few odd pieces missing from the WC4 and WCP DVD-based movies - granted there may be more than a 'few' missing in this collection but it makes sense to use the best starting point, isn't it? (I honestly don't remember what was missing, I thought it was a pretty good set of the movies.)

To answer @ODVS' question about whether or not to pursue a P2 upscaling project, I would say if we can only work with the original game material it's probably fine to leave it for now.
 
It's an option, isn't it? (Also not sure if it's worth contacting Mr Hilliker directly about this.) I remember there were a few odd pieces missing from the WC4 and WCP DVD-based movies - granted there may be more than a 'few' missing in this collection but it makes sense to use the best starting point, isn't it? (I honestly don't remember what was missing, I thought it was a pretty good set of the movies.)

To answer @ODVS' question about whether or not to pursue a P2 upscaling project, I would say if we can only work with the original game material it's probably fine to leave it for now.

There's pretty much none of the alternate choices included in the set so pretty much every chunk of conversation is going to be missing something... I'd definitely like to compare and see just how much is missing overall. Between the actual good quality footage and the "not very good" VHS sourced stuff we probably do have a lot of it covered. It can be kind of jarring though if you have too many disparate quality sources. Ideally you want to keep it somewhat consistent across a project like this.

EDIT: I just put every video in the actual game into a playlist (including landings and transitions) and there's aprox 2hrs and 16 min of video in Privateer 2. There is about 1 hr total in Steve Hilliker's youtube compilation once you remove the 5 or 6 minutes of intro and end credits
 
Last edited:
Hi @Ashurek69,

Sorry to hear you're having issues. You're right, this hasn't come up before on the WCIV videos.

Just as an experiment, could you please set your Windows display reslution to 1920x1080 and turn off any Windows DPI scaling before launching the game just to see if that has any effect?

GOG did patch the game recently which, as far as I can tell, added hardware accelleration to the video playback. This is a guess on my part, because I noticed that the video was no longer pixelated on my display. I'm using a 2160p display, so the HD video is always displayed at 2x for me. Prior to the update, I had minor "nearest neighbor" pixelation on video playback, whereas now it uses bicubpic scaling, which would suggest HWACCEL has been added. So just to check, do you have the game set up to "keep this game up to date" in GOG Galaxy? If you're not using GOG Galaxy, it might be worth downloading and installing a fresh copy of the game.

If none of the above helps, let me know and I'll have some more thinks about it :^)
 
I think the second Privateer 2 test is actually more than a minor improvement. For the first, I think it's kind of a toss up. There are improvements, but some facial detail kind of gets blurred. For the second video, there may be some artifacting, but a lot of the faces still look decent. Also, the overall contrast and color balance is a lot better.

Thanks, Chris :)

I continue to tinker and experiment...
 
Hi @Lohrien and @Maze59077 - I had already compiled them on the assumption that you were right, and @KrisV just kindly overwrote the originals on the server for me :D

So, same links as before should work for you now!

https://download.wcnews.com/files/wing4/wc4_hdpack_5.0_DE_01.zip
https://download.wcnews.com/files/wing4/wc4_hdpack_5.0_DE_02.zip

Enjoy!
Thanks for reuploading the German HD video pack for WC4 .. it works perfectly 👍 👍 .. i started a new Borderworld War against Tolwyn right now🤪 first Time on a big 65" TV ..it looks amazing👌👌
 
@ODVS
Thanks for the tips, I believe it was the app scaling in windows, once I disabled that, the movies now work perfectly (albeit a little quietly).
Setting the resolution to 1080p and trying to run the game broke everything, my centre screen turned off and the two side screens were corrupting everything on display, but that aside, works like a charm now and love the work.
It would be fantastic if better source material could be found for WC3 and Privateer 2, even though the work you've done on them already is amazing considering what you had to work with.
Thanks again for the work you've done and the advice on getting it to run.
 
Very glad it helped, @Ashurek69 :)

I believe if you track down your WCIV executable, right-click it, go to properties and hit the "Compatability" tab, you can tinker around with the settings and get it to ignore Windows DPI scaling so you don't have to globally disable it every time you launch the game.

Microsoft tend to alter the layout of these settings with almost every Windows Update, but you might be able to figure it out from there :D
 
Okay, because everyone kept asking, I did an experiment remastering Steve Hilliker's footage from his Beta SP Master Tapes.


While it has an undeniably better frame rate (I don't need to interpolate up from 12fps) and colour depth, this really highlights my concerns about attempting to remaster footage from analogue sources. Sources like VHS/Beta tapes soften the edges around objects, and the AI routines maintain that softness during upscale - so we lose a lot of image definition. Ideally, it's best to work from a source that was digitally captured or scanned directly from film.
 
What happened to the master tapes for Wing Commander 3? Did EA lose them?

It's funny how everything positive is attributed to Origin and anything negative is EA's fault. :) The tapes were misplaced long before EA assumed ownership of Origin's old stuff.
 
It's funny how everything positive is attributed to Origin and anything negative is EA's fault. :) The tapes were misplaced long before EA assumed ownership of Origin's old stuff.

By lost are we thinking truly lost or in the hands of an ex-developer/ collector?
If we did have them what do we know about the quality? Film? VHS? Beta? Would they be green screen, combined with the backgrounds, if the latter what resolution were the source backgrounds at etc?
 
By lost are we thinking truly lost or in the hands of an ex-developer/ collector?
If we did have them what do we know about the quality? Film? VHS? Beta? Would they be green screen, combined with the backgrounds, if the latter what resolution were the source backgrounds at etc?
WC3 was tape IIRC (probably a Beta format of some sort) though they experimented with film a bit IIRC. The way they filmed though, the SGI machines were pumping the backgrounds directly into the feed so the director and actors could look on the monitors and see them on the sets as they filmed. That doesn't mean they didn't capture the actors independantly of the backgrounds necessarily but it's why the majority of the WC3 shots had to be static camera... any movement had to be pre-planned to match the previously chosen set.
 
I did an experiment remastering Steve Hilliker's footage from his Beta SP Master Tapes... this really highlights my concerns about attempting to remaster footage from analogue sources. Sources like VHS/Beta tapes soften the edges around objects, and the AI routines maintain that softness during upscale - so we lose a lot of image definition. Ideally, it's best to work from a source that was digitally captured or scanned directly from film.
Is it the case that the particular NN models you have access to are optimised for digital video? I'm just thinking that at some point, everything digital was converted from an analogue source, it's just that in digital cameras the ADC happens at the sensor level (light capture). Does scanning from film being 'ideal' mean that even if one had access to the Beta SP tapes we wouldn't get anything better than the videos Mr Hilliker has already put up? (Other than the obvious benefit of digitising from the tapes directly means that one could choose the final resolution anyway, cutting out the need for upscaling.)
 
Is it the case that the particular NN models you have access to are optimised for digital video? I'm just thinking that at some point, everything digital was converted from an analogue source, it's just that in digital cameras the ADC happens at the sensor level (light capture). Does scanning from film being 'ideal' mean that even if one had access to the Beta SP tapes we wouldn't get anything better than the videos Mr Hilliker has already put up? (Other than the obvious benefit of digitising from the tapes directly means that one could choose the final resolution anyway, cutting out the need for upscaling.)

Really it's just a case that the available footage is too fuzzy. We don't have defined, to-the-pixel edges (between objects, for example). The edge between objects blurs over several pixels, so the NN sees it as a gradient and not an edge and maintains it.
 
Here's an example of what I mean:

zoom.jpg


See how the "edge" of Joe's shoulder is spread over a range of about 3-4 pixels? That's our problem.
 
When I say "analogue sources" I guess it would be better to clarify that I mean "magnetic tape sources intended for old CRT screens". I mean yes, film is an analogue source, but the actual "to-the-retina" resolution of 35mm film is roughly equal to 4-8K of image resolution.

VHS and Beta have a maximum resolution of around 520p-ish (later versions of Beta might be higher). You can't capture them at higher resolution than that, because the image info just isn't there. On top of that, the magnetic analogue storage method leads the the kind of pixel-blurring in the example above, and that's our major issue here. Ultimately, I don't need high resolution footage to work from - that's the whole point of the AI. What makes all the difference is sharp/high quality standard definition - and that's where VHS and Beta let us down.
 
By lost are we thinking truly lost or in the hands of an ex-developer/ collector?
If we did have them what do we know about the quality? Film? VHS? Beta? Would they be green screen, combined with the backgrounds, if the latter what resolution were the source backgrounds at etc?
There's a stillshot of the blair/paladin concordia shot from wc3 on imdb.com it's way higher resolution than anything in the PC/Console versions of the game. Makes me wonder what would be possible with an SGI machine (I have a working O2 here) and the original materials.
1602730891478.png
 
Last edited:
When I say "analogue sources" I guess it would be better to clarify that I mean "magnetic tape sources intended for old CRT screens". I mean yes, film is an analogue source, but the actual "to-the-retina" resolution of 35mm film is roughly equal to 4-8K of image resolution...
Thanks for the clarification, that certainly makes a lot more sense to me now. Definitely agree that the softness of the source material contributed the softness of the output.

I probably mentioned it already but in the late 2000s, before neural-networks and all that 'deep learning' stuff really took off in this decade (yes, they've been around for longer than that), I worked in an image processing R&D type of job and edge detection algorithms was one of the things I implemented (by 'hand', before we invested in image processing libraries), which was particularly useful and educational for me as we were experimenting with different algorithms and parameters. It's clear from the example you highlighted that there just isn't much contrast for a computer algorithm to recognise anything as an edge (even though we as humans can discern it - that's because we're actually 'intelligent', not just 'artificially' so).
 
Back
Top