I had a couple people ask where to find the tools I’m using for the Arena Depixelization Project. So here are the links…
Occasionally, I like to test out games on my 4k monitor to see how they look at such a high resolution (3184×2160). For some games, that impact is less than you would expect while others look surprising good. For example, Bethesda Gamebyro engine-based games (Elder Scrolls and Fallout) look a little better but I’m not very wow’d by it. I think it boils down to the engine and lower resolution textures. Less aliasing (jagged edges around the 3D model) and more detail but not the punch you would expect.
On the other hand, Two Worlds looks fantastic. The textures are high resolution enough to make them really pop. Unfortunately, the interface doesn’t scale so it’s very “difficult” to play at such a high resolution. Visually, it’s very pleasing though.
So one day, I got a bug to try the old game Fate. It was the predecessor to Torchlight. A top-down action-RPG click fest. My boys played all but last one in the series so as usual, there is a touch of nostalgia for me. The only thing I had to figure out tech-wise was how to get the resolution desired in game. Some games work out of the box and some require manually editing “ini” text files. In this case, I had to edit the “ini” file AND disable scaling on high DPI settings. I did that by right clicking the program executable file and putting a check under that line in the compatibility tab. Viola…4K work for a decade+ old game.
Both ends of the spectrum
You can see in the screenshot (even if you don’t have a 4K monitor) that the character models are actually quite detailed. In 4K, they really stand out and don’t really need any work. Well, I could do a little but I digress…in general, they are perfectly nice. However, as you can see in the “snow/ice” texture, the non-character models are blurry low-resolution messes. Basically, they are tiled 128×128 textures. They are so dramatically worse than the character/object textures in 4K. The building are also very low resolution…which is even more boggling since they don’t blend into the background like the landscaping. That is when I got inspired…to fix the disparity.
The above (partially complete) picture is post landscape editing. For the record, I’m still not crazy happy about it but small steps.The landscapes textures have revealed the oddites of the engine. A few places where they don’t actually blend and such but not too much. The real work and most dramatic change was the building and landscape objects…but that is for next time.
Windows to the soul
Recently while playing around with “vanilla” (unaltered) Arena, I noticed that the windows turn on an off for day and night. I believe this is based on the window color being pure white. In quite a few of the SET files, I had made the windows yellow breaking the on/off effect. It’s an easy fix though.
Caveat: I’m a hopeless tech nerd.
I like trying out software and technology. Things like text-to-speech, software debugging, emulators, etc. are like candy to me. Why is this on an art blog….err i mean…collegiate journal? Because sometimes I mess stuff up. This time I was trying to get a Windows XP/7 program that doesn’t like Windows 10 to install on my computer. In doing so, I borked my login so that it wouldn’t accept my password. From that point started a chain of events (all my own doing) to fix it that made the situation worse and worse. Finally, I couldn’t even boot my machine. Seriously, I’m pretty good with tech but on this day I rolled a “1”.
Long story short, I recently had to wipe my boot drive (no important files were on there but the OS) and install a clean Windows 10. I tried to cold turkey eliminate using GIMP since I already have Photoshop. That didn’t work too well. Photoshop is different enough that some very common things I was used to doing all the time in GIMP didn’t work. I’ll probably start watching some intro videos to see how Photoshop works before I try that again.
All in all, not my proudest tech week (or two).
I did some research into the MAP files for Arena. However, I want to compile it in a cohesive manner so that post will wait till I feel it is ready enough. My hope is that the information that I have uncovered might help others if they decide to “crack” the format.
The straight and narrow
Some days I can blaze through images getting 5 or 6 done in a sitting, then other days just part of an image can take a couple days. When it comes to the Arena Depixelization Project (ADP), the easier ones are just patterns (especially ones with straight lines horizontal or vertical). The low pixel count doesn’t really muck up the design too much since there is no need for a “fine detailed” line in those cases.
It kind of looks like….
However, once lines start curving or the image becomes complex, the time required is increased, sometimes dramatically. It’s hard to get non-angular shapes to look good when you only have a 64×64 grid to work with. I find that it takes multiple revisions to get it looking just right. To make matters worse, the source material for ADP contains many pictures that are either so pixelated as to be barely discernible or they generally don’t look very good as you clean them up (such as dress thing for the “angel” in this picture).
Close enough for horseshoes and hand grenades
In that image, I had to make some artistic decisions on how “accurate” to the source material I would be. The image is generally too complex to leave unaltered for my “Depixelization” theme. At first, I didn’t even know how to tackle it. I completed all the background wall first and left the “Angel” and the “alcove” till last. The good thing about that is that it let me establish the colors for the image. Eventually, I had to do this one as I only have a few SET files left. I decided to break it up and focus on individual pieces of the angel. I really wished I had made a timelapse of this one to show how much back and forth I did on it.
Mr. Potato Head
I started with the more clearly defined shapes, to include the arms, head, and the key. Those came into place with minimal fuss. Then, I tackled the wings. It took me awhile as the initial version (closer to the source) just didn’t look very good. After a couple of iterations, I ultimately decide to make the wings bigger as if they were more full body sized. Next, I attempted to do the feet. However, they just did not look good, especially when I started working on the weird “billowed” dress/robe. I decided to put the feet off and do the clothing. I touched up the sleeves to improve the “hanging off the arms” appearance. I made my first rendition of the lower part of the robe very close the source version. However, it looked terrible because the source version is kind of ridiculous looking, as if someone tucked an oversized shirt into a skirt. I played with it a little before I decided to just alter the design. I revised it to look more like a regular robe. Unfortunately, the feet still looked awkward, so I removed them and lengthened and curved the robe as if they were hidden by it and it was floating.
The tool that saved the Arena Depixelization Project (ADP)
Last post, I covered the Arena Font Editor, 1 of 3 tools that I use to edit TES:Arena graphics (and fonts). The font editor is part of the Arena Modding Suite by Hallfiry. The other part of that suite is the 2nd (set) of the three tools I’m going to write about.
Prior to the Arena Modding Suite, I used the method detailed in a previous post that was laborious and unpractical. Fortunately, this thread popped up on the Bethesda forums. And instantly my little “experiment” became a project and grew in scope. Were it not for Hallfiry, I would have surely abandoned it ADP before it every took off.
What’s it do?
The main functions of the Arena Modding Suite come in the form of the ArenaPacker and ArenaUnpacker. Rather than being a program that you work in, they are utilities that enable you to easily manipulate the game assets directly in Windows. Both programs directly work with the GLOBAL.BSA. Appropriately enough, one unpacks the entire BSA into a set of folders and the other takes that set of folders and packs it right back up although that is a bit simplistic view of what they do.
In reality, the programs not only work with the files but they also convert the files to the appropriate format. For ArenaUnpacker, when it extracts the files, it converts the non-standard IMG and SET files into easily edited PNG images. Additionally, it converts the INF files (map asset listings) to a text editor friendly format and SND files to WAVs (although I don’t have any interest in that part). ArenaPacker reverses the process and creates a packed GLOBAL.BSA based on files in the unpacked directory but doesn’t alter those files that were already extracted. This means I can have a working directory of all the files and my changes then “pack” my work-in-progress easily at any time to test in game.
Some notes about the Arena Modding Suite:
1. Quite a few of the images are compressed in a bizarro compression routine used by Bethesda and this software doesn’t have the functionality to uncompress them. No one had cracked that compression in all the years since the game was released (that is until very recently but more on that next time).
2. ArenaPacker is designed to compensate for using colors outside those available in palette file by converting non-palette colors to the nearest equivalent color in the palette. While it’s a handy feature, the images should be checked in game to make sure the colors aren’t changed to something wonky (as happened before I learned to use the palette tool in GIMP. If you stick to the exact palette (either by using a palette file or be just using colors in the actual images being edited), this isn’t a problem.
3. ArenaPacker is a little sensitive to what files are being repacked. When files are extracted, ArenaUnpacker creates a file list of all the files in each directory. This file list is used for when the files are repacked by ArenaPacker. So, if a file is missing or added that isn’t on the file listing, it will crash the program. So if I plan on “trimming” out the IMG/SET files not actually used, I’ll have to edit the file listing. However, it is very easy to fix so this isn’t that big of a deal.
I haven’t had a lot of time this summer but I have made tweaks here and there. There are a few textures I’ve done on another computer that I have yet to transfer over to my master files. Things have settled a little and I should be able to refocus my efforts on finishing the SET files as soon as possible.
For me, one benefit of extended “down time” on a project is that when I finally am able to return, I have regained a lot of my motivation. Additionally, I come back with “fresh eyes” and see ways to do things differently or better. While it’s a bummer to have to delay working on a project, I know I will always return because I love doing it and to me it’s very relaxing (even when I’m replacing every @#$% pixel in 1994 game with only 256 colors, 64×64 image dimensions and cryptic/bizarre file formats)
Here are some photos showing some changes and things I am working or have noted.
As I have stated in the past, one of the most taxing aspects of editing the TES:Arena textures is coming up with a unique take on the image within the constraints of the limited pixel and color count. While I was thinking about my methods and ways to be more efficient, it occurred to me that I don’t appropriately leverage my motivation. For the last 1 1/2 I have been working one texture at a time. This would exhaust my creativity for the day too quickly. I would waste a bulk of my time “finishing” said texture after I decided how I wanted it to look. All of the “fresh mojo” that I had when I started for the day was worn away by the time I got through a few (sometimes one) SET files. This is especially true with the most of the remaining SET files that have no defined look or image (i.e. barely recognizable as any thing but random colored pixels). Not only was that detrimental to the quantity of work completed in any one sitting, but it also deterred me altogether. There were days that I just didn’t have the energy or initiative to try and figure out a new take on what looked like white noise so I just wouldn’t decide to spend my efforts elsewhere.
To better capitalize on that “creative fronted”, I tried an experiment. I loaded up 18 SET files in GIMP. Then one by one, I worked on just getting the look I wanted for a small portion, enough to establish the look. I didn’t waste time completing the whole texture as that is really the easy and mostly mundane part. It worked quite well. Using this method, I was able to solidify my design for 16 SET files in just one day (really just a couple of hours). As I said the design is the hardest part with this project, and I wiped out almost half of the remaining SET textures left. Now, if I don’t feel particularly creative but still want to make progress, I can just finish up some of the ones I have already started (template on itself) and if I do…I may just be able to finish the other 16 or so.