On Wednesday it was time for the dev stream preparing us for the launch of 7.8! And it was... okay? To be honest, there were less "news" than I'd hoped for, though to some degree that was to be expected considering we already knew a fair bit about the patch's major features. Still, I was a bit surprised that they spent a full fifteen minutes recapping the Legacy of the Sith story so far - I have no doubt that's useful to some, but probably not to that many among your hardcore audience that come to watch you live on Twitch. Whenever I glanced at chat people were complaining that they were bored and wanted to move on, and I had to remind myself to not pay too much attention to chat because it always contains a lot of negativity, sadly.
Also, what they told us about Galactic Threads was pretty much the exact same stuff they told us about it a year ago when it was originally meant to launch. Really, the most exciting bit of genuinely new news to me was a note that they'd be updating the large ceiling hooks in strongholds to work like large floor hooks, meaning they'll also contain a three by three grid of small hooks. This is great because the devs gave us a bunch of new small ceiling decorations this season which look really nice but were hard to place in an effective way due to the lack of small ceiling hooks available in almost every single stronghold. Either way, 7.8 still looks exciting to me and I'm very much looking forward to it.
One slightly sad piece of news for German and French players was that the voice work for 7.8 still isn't ready in their languages, so they'll only be able to play the new story in English for now. I wonder why Broadsword couldn't get that done while the English voice actors were on strike - the German and French ones weren't, as far as I'm aware! Either way, I guess it's up to these players to choose whether they want to play through the new story in English or wait even longer.
(Side note: Sometimes I wonder whether a future project for me, once I hit legacy level 50 on all servers, could be something like playing through the class stories in German. I bet hearing all those player characters and familiar NPCs talk in completely different voices would be wild. But I'm getting sidetracked...)
Anyway, I'm not going to go into much more detail about things like new Cartel Market additions and so on. As usual you can find full, detailed breakdowns elsewhere:
I just want to talk about two more things that came out of this livestream. First off, Papa Keith appeared at the end of the stream, and as he so often does, casually dropped a hint about something yet to come that would hype players up. Specifically he said that in his producer's letter for the game's anniversary, he'll be giving us a roadmap for what's coming in the new year, and that they'll be "wrapping up Legacy of the Sith". That's kind of like... telling us that they are working on an 8.0 expansion without telling us that they're working on an 8.0 expansion.
EEEEEEEEEE!
I'm trying not to get too excited just yet because it hasn't been stated explicitly at this point, but it's hard not to be hopeful. I still stand by everything I wrote about the prospect of a new expansion a little over a year ago, but man, would I ever love for a good expac launch to make the game feel exciting and invigorated again. I'm looking forward to learning more.
Now, for something completely different - at the end of the stream, they showed off the new loading screen/key art for patch 7.8:
My reaction to that was something along the lines of "Nice. Lana and Shae's faces look a bit weird though. Oh well." And then I moved on, because a character looking slightly off on the art of a SWTOR loading screen is a pretty common thing. I remember when Legacy of the Sith's key art was first revealed, one of my guildies commented that Lana looked like Adam Driver in a wig, which made me go "oof" but I could also see what he meant!
When I checked social media for chatter about Wednesday's dev stream, I was surprised to see a few comments that accused the new key art of being AI generated. Warning: I'm going to go into way more detail than this probably deserves, but it's a subject that interests me and I kind of fell down a rabbit hole, so you've been warned.
As a certified AI hater my reaction was of course to be shocked and want to find out more. From what I've seen of the SWTOR devs they've always struck me as very pro human creativity and anti AI art, but who knows what's going on behind closed doors when there's money to be saved? I tried to look for more information on where this whole idea came from, but couldn't find more than the same three people or so repeating that it was "obviously" AI, no explanation needed, which was of course not helpful.
Finally Tacogoats, a SWTOR player I follow on Bluesky, shared a link to a tumbler post of his in which he attempted to break down the evidence he saw for the art piece being artificially generated. And for all the details he circled... I did not find it convincing at all. The comments seemed to either be flat out wrong (no, Anri is not missing her index finger; it's right there), issues that I couldn't make out at all, or even if I could kind of see what it was that was bothering him, I still didn't see how these things were in any way evidence of AI involvement. So what if the one knuckle on Malgus' hand has a slightly odd edge? Cause human artists never create the slightest imperfections when drawing hands? I was expecting some kind of "gotcha" like a line from something in the foreground merging into a different line in the background, like you can often get with AI art since it doesn't actually have any concept of three dimensionality. However, there was nothing of the sort.
In fact, the more I looked, the more I became convinced that this couldn't be AI. AI image generators are very good at conjuring up things for which they have lots of references. They can easily create a photorealistic image of a non-existing human because their training data contains millions of photos of humans! What they are not good at is creating very specific details for which they don't have (enough) references. And let's be real, random SWTOR NPCs are not something for which any AI is going to have lots to work with.
I remember when I encountered the very first image generator I'd ever seen - a few years ago now - I tried to get it to generate images of Darth Malgus and Satele Shan and it failed utterly. I think Malgus was a black blur with angry eyes so it clearly had some vague reference for him, but Satele pretty much just resulted in random noise. Obviously things have progressed somewhat since then, but it's not like the internet suddenly got flooded with millions of screenshots or art of all these SWTOR characters that wasn't there before.
I actually got curious enough to dust off my OpenAI login and ask ChatGPT to generate me a "painterly" image of Lana Beniko (since she's one of the characters on the key art). I'll be honest, the result was actually better than I expected. It clearly had enough material to deduce that Lana's a white, slightly angry-looking woman with mid-length blonde hair and a side parting. I also asked it to generate an action shot in which she'd have he lightsaber out and again - it wasn't totally terrible in my opinion.
But two things stood out to me: that it was actually very good at giving her an attractive face (even if it doesn't look exactly like Lana does in game) - which is something to think about when the first thing anyone mentioned being slightly "off" about the key art were the faces - and the fact that the AI had absolutely no clue what sort of armour she wears, dressing her in a generic black combat suit with matching pauldrons and a cloak.
Now, I'll be the first to admit that ChatGPT might not be the best image generator and that my prompting was extremely low effort - I'm sure someone could prod an AI into producing something better, but anywhere close to the sort of detail we actually see in the key art? That seems incredibly unlikely to me, not with where we're at right now. And all that's just talking about a single character, never mind five of them in a balanced composition. In a way, going down this rabbit hole has actually increased my appreciation for this new loading screen.
I can't blame people for being paranoid and suspicious. Corporations trying to cut costs by replacing real artists' jobs with an inferior AI-generated product is a real thing. But so is people being overly paranoid and thinking that everything is AI when it actually isn't. As a creative person myself (even if my hobby is writing, not drawing or painting), someone accusing my stuff of sounding/looking like AI would actually be more hurtful to me than just saying that they think it's terrible. We're not doing the art community a favour by immediately accusing things of being "AI slop" as soon as we don't like something.
Bonus AI laugh: I also asked ChatGPT to generate me an image of Shae Vizla and that one was a lot less charming. It clearly had enough of a clue to give me a red-headed Mandalorian, but what the fuck is that armour? Why is the "blaster" I asked for in the prompt just a plain old 20th century gun? And why is there another random Mandalorian's helmet attached to her thigh?!
"I wonder why Broadsword couldn't get that done while the English voice actors were on strike - the German and French ones weren't, as far as I'm aware!"
ReplyDeleteI was wondering that same thing. Anyway, the generated images of Lana look pretty good, IMO. Shae isn't terrible, I guess. Not old enough, but the mando thigh helmet is hilarious.
Shae's AI doppelgänger is fascinating. I'm willing to bet that it's assessed that Shae is a Mandalorian and from the Old Republic, so has created a blend of Cassus Fett (or other such Neo-Crusader Mandalorians) and 'modern-day' Mandalorians. I'm just curious what the heck she's looking at to the side while aiming elsewhere with her... concerningly accurate depiction of a modern pistol.
ReplyDeleteI don't really have a lot to say about the new key art - on the whole I think it's okay, and I really like the colours, they just seem to 'pop' nicely, but Arn's face... oh, Arn. What have they done to you?
This might not apply to Broadsword because of them not being exactly rolling in money but in general I would expect game companies to train their own generative AIs on their own in-house assets rather than farm them out to the big names. When that really gets going, at least everything will look consistent.
ReplyDeleteWhat's definitely going to happen for a while is lots and lots of mis-attribution as people who have absolutely no idea what they're talking about claim to be able to tell the provenance of any picture they do or don't like. I'm old enough to remember when the term of abuse thrown at artists was "airbrush". Anyone using the tool was automatically deemed a hack and people were always arguing over whether something had or had not been "airbrushed".
I did think about the possibilities of a custom AI model trained on the right imagery... but that's effort again, so does somewhat go against the idea of using AI for quick savings unless you're producing a sufficiently large volume of promotional art that it will pay off in the long run. For a small MMO that doesn't really do a lot in that area (getting a new loading screen once every six months is about exciting as it gets) that just seemed like an unlikely direction.
Delete