DiscoShark
Banned
Lol, double post. Anyone down for a set today? Im open for the duration.
-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
Skullgirls |OT| New age of Heroines
- Thread starter SolarPowered
- Start date
Master Milk
Member
never again lol
DiscoShark
Banned
D:
Your zoning with Peacock has actually really improved, you did a good job of keeping me out when I was in Peacock's optimal range. The problem I felt was when you were fighting an opponent that knew how to wait her out and knew at what ranges to chill at when Double would whiff. You stayed unpredictable to a degree with using those opportunities to throw your opponent but couldn't really transition back into a zoning based game as comfortably.
Here's some tips!
Peacock on point :
If your opponent is forced to block your assist at any point of the screen, you have free reign to throw out whatever projectile you want, typically Bombs are your best choice simply because they stay out for longer, dont have much recovery, and force your opponent to cover multiple angles. After you've thrown out a few bomb cancels I usually item drop.
If your opponent is staying around mid screen and waiting for your double assist to whiff then throw out a car and teleport, punish the passiveness and use it as an opportunity to avoid corner situations. You may get a crossup and full combo.
cMP - The unlikely antiair
This move is actually really great and hits characters like Filia and Fortune out of their airdash approaches. It can combo straight into LK Bomb and can be used to give you the necessary breathing room to continue zoning. It gets really scary when LK Bomb is combined with an assist input as well.
Example : c.MP -> QCF + LK+MP
In this setup you use cMP to antiair your opponent and combine an assist call with LK Bomb which, with Double, means you can confirm into Argus Agony or just keep throwing out more nonsense.
j.HK runaway.
Typically what you want to do is jHK once, airdash backwards, jHP, then call assist. While throwing out multiple jHKs can be convenient if you're trapped in the corner typically you want to get back into bomb / item drop based zoning asap.
Gimmicks :
After you throw your opponent with Peacock, toss out a bomb and force them to guess on a teleport. They guess wrong and you get a free combo.
Launcher -> jLP jLK jHK -> land -> Walk Forward
It crosses up
Double on point :
Abuse cat heads forever and continue to jHP
Edit : Oh and antiair assists help. Consider sticking Parasoul or Filia on that team and you'll find your matchups will get much easier.
shaowebb
Member
I got a developer question for you Ravidrath. Im doing some hand animated sprite work now for some small games in Construct 2 and I really dig how clean the characters look in Skullgirls.
What size are the images for the characters in photoshop as they sprite them out?
Its been well over a year since I had that art test you sent me and I can't remember how big they were. Im just trying to make certain my work doesn't get pixelated all to crap on a PC at different aspect ratios since Im not using vector art and Im hand animating the classic way.
What size are the images for the characters in photoshop as they sprite them out?
Its been well over a year since I had that art test you sent me and I can't remember how big they were. Im just trying to make certain my work doesn't get pixelated all to crap on a PC at different aspect ratios since Im not using vector art and Im hand animating the classic way.
Actual size sprites, composited:
There isn't a set sprite size - it's just about making them the size they need to be the correct resolution.
Like, I think Valentine's blade slash attack is probably at least 2000 pixels wide.
darkblade77
Member
It's probably worth mentioning that thanks to the split format/layer munge process/palette limitations/other wizardry used in SG, it has an insanely lower memory footprint compared to just trying to put vanilla high-res assets into a given game. Even most pre-optimized image import formats don't seriously come all that close, as far as I've seen.
Fuck yeah!
Back in first!
Yeah, Mike's sprite shit is pretty ingenious.
BlazBlue has like half the frames we do at a lower res, and they can just barely get two characters in memory.
Back in first!
Yeah, Mike's sprite shit is pretty ingenious.
BlazBlue has like half the frames we do at a lower res, and they can just barely get two characters in memory.
Basically...
- They went into Sony's system when the patch did
- But we can't release them until the 360 patch is out, because of MS's parity policy
- But Sony's system requires continual manual updates of the release date to keep pushing it back
- Sony Europe slipped up and forgot to do this, I guess?
- So they got out
- But we can't release them until the 360 patch is out, because of MS's parity policy
- But Sony's system requires continual manual updates of the release date to keep pushing it back
- Sony Europe slipped up and forgot to do this, I guess?
- So they got out
We're not calling attention to it because of MS's parity policy. And they'll be out everywhere for reals whenever we get the 360 patch out.
Any chance in the future we can get a blog post with the technical details on how you guys accomplished this wizardry?
darkblade77
Member
I think they already have some art blog posts and the art process video that touch on how this is done, but barring that I'll explain what I understand about why the size is so much smaller from a technical standpoint.
You know an image is made up of pixels, and said pixels can can take on thousands of color values and or/alpha channel modifications.
If you build such an image in data, it has to be designed so that any one pixel could hypothetically display any of these values. This is a 'bitmap'; a 'map' of 'bits' that represent color values.
A lot of image formats try and reduce this space footprint through various techniques that go up in complexity.
Mike's solution 'splits' an completed image into multiple images with extremely limited palettes(well, I say 'split' but really the art is just largely given to him in those divided parts at the beginning), removes most of the cost of empty/transparent space by just marking when it should be displayed and not actually reading it in, and then combines the limited-palette images together at runtime, likely via shaders.
In addition, the image is encoded in a block format(if a pixel is one unit, think of a block as literally a square of n^2 pixels covering a similar area, all with one value), likely with some small overhead to retain any important pixel distinctions that might need to be represented at decode time. This allows for really efficient storage when the displayed image is half the original resolution, but also allows as much quality as possible to be retained for other sizes as long as the ratio is the same.
Finally, there's some image quality filters and whatnot (probably identifies/focuses around the lineart portion) to smooth the final product out so it it retains sharpness close to pixel-perfect even when being resized, unlike the consistent blurriness that would come with something like a bilinear filter.
I'm sure there's even more secret sauce that goes into this, but that's my understanding from what I've observed/heard.
You know an image is made up of pixels, and said pixels can can take on thousands of color values and or/alpha channel modifications.
If you build such an image in data, it has to be designed so that any one pixel could hypothetically display any of these values. This is a 'bitmap'; a 'map' of 'bits' that represent color values.
A lot of image formats try and reduce this space footprint through various techniques that go up in complexity.
- Reducing palette size. If you're only using a set amount of colors and you won't deviate from that, then the image doesn't have to account for thousands of other values. (biggest factor)
- Compression; attempting to identify areas in an image that don't convey important/discernable information to the human eye and 'smudging' an image so that those extra portions are exactly the same as some other similar portion.
- Formatting; changing the way image data is interpreted by a system so that common/redundant values take up less space.
Mike's solution 'splits' an completed image into multiple images with extremely limited palettes(well, I say 'split' but really the art is just largely given to him in those divided parts at the beginning), removes most of the cost of empty/transparent space by just marking when it should be displayed and not actually reading it in, and then combines the limited-palette images together at runtime, likely via shaders.
In addition, the image is encoded in a block format(if a pixel is one unit, think of a block as literally a square of n^2 pixels covering a similar area, all with one value), likely with some small overhead to retain any important pixel distinctions that might need to be represented at decode time. This allows for really efficient storage when the displayed image is half the original resolution, but also allows as much quality as possible to be retained for other sizes as long as the ratio is the same.
Finally, there's some image quality filters and whatnot (probably identifies/focuses around the lineart portion) to smooth the final product out so it it retains sharpness close to pixel-perfect even when being resized, unlike the consistent blurriness that would come with something like a bilinear filter.
I'm sure there's even more secret sauce that goes into this, but that's my understanding from what I've observed/heard.
darkblade77
Member
They're not related; there are a lot of hit/hurtboxes in SG compared to other games because they wanted attacking and vulnerable areas to match the sprites-on-screen to a greater degree than other FGs do.
Also, when I mean 'split' I don't mean the image is literally cut into segments; I mean like something closer to Photoshop layers.
Also, when I mean 'split' I don't mean the image is literally cut into segments; I mean like something closer to Photoshop layers.
shaowebb
Member
Thanks for all of this you two. Thats around the size I was going to be working at after my rough mockups that I am putting through color tests right now so I should be good then. Im already planning to run in game filters to help sharpen things up as well. Im not a coder though which is why I'm using Construct 2. My brother in law IS a coder though so if I need to get some more stuff coded thats outside the gui and my skills I can try to contract him for a little cash to code it out in html5 for construct 2 to accept.
Im setting up my basic stuff right now and once I get a run cycle and jump cycle I may just push this character through a Jetpack Joyride game first to feel out the engine before jumping into the action platformer Im intending to use him for which is a much bigger project. He's going to get some redesigns and such on the shoulders since from the side they dont look so great during run cycles and of course these aren't even fully colored or trimmed yet. Still here's a small version of where Im at in the rough tests on the design to give you an idea of it.
SolarPowered
Member
Even though Ravi mentioned it...
http://www.jp.playstation.com/psn/store/magazine/
Skullgirls... the animal crossing of fighting games? It's more likely than you think.
http://www.jp.playstation.com/psn/store/magazine/
Skullgirls... the animal crossing of fighting games? It's more likely than you think.
This is the kind of thing PC gamers would LOVE to hear about when you guys start talking more about the PC version. The tech behind a lot of Skullgirls seems to be one of a kind and PC gamers love learning how they can fiddle with the system for prettier visuals, more data, etc. Which leads me to...
This very post right here. Hopefully a while from now because I'm sure everyone is busy as hell right now.Any chance in the future we can get a blog post with the technical details on how you guys accomplished this wizardry?
Why is that? It doesn't seem like they way they make their characters has evolved that much since the early GG days. I find it so strange that a small American upstart can be so ahead of the pack in this regard.
darkblade77
Member
Hopefully I mangled just enough of that tech breakdown that Mike starts posting on NeoGAF just to correct me.
Also, Arcsys' solution is the result of... I guess just having a set of palette-limited images at that resolution. Limiting image size via reduced palette is the oldest trick in the book(maybe even literally the oldest), so it isn't like what they're doing is *bad*... it's just that like with many software solutions, someone can always come up with a superior method. And Mike's method, as far as I can tell, is cutting edge--despite the fact that the fundamental parts to make it work have existed for years. People are sleeping on this stuff.
How should I put it... you know Durante's supersampling thread? SG sprites are essentially getting a lot of the benefits of supersampling with almost none of the cost/overhead that should be coming with it (relatively). I'm not sure whether to call Arcsys complacent or just call Mike a wizard.
Also, Arcsys' solution is the result of... I guess just having a set of palette-limited images at that resolution. Limiting image size via reduced palette is the oldest trick in the book(maybe even literally the oldest), so it isn't like what they're doing is *bad*... it's just that like with many software solutions, someone can always come up with a superior method. And Mike's method, as far as I can tell, is cutting edge--despite the fact that the fundamental parts to make it work have existed for years. People are sleeping on this stuff.
How should I put it... you know Durante's supersampling thread? SG sprites are essentially getting a lot of the benefits of supersampling with almost none of the cost/overhead that should be coming with it (relatively). I'm not sure whether to call Arcsys complacent or just call Mike a wizard.
This makes me very excited for my purported MikeZ Skype chat. I'm taking a 3-year Game Dev course, and while it's nothing like majoring in CompSci, getting gdlk at programming is now my current endeavor. I hope he doesn't mind if I farm him for protips.
Also, Ravid, hear any more elations from Cyberfront? Sitting comfy on the #1 spot on PSN must feel pretty good right now.
darkblade77
Member
Nothing's really stopping you from talking to him now, y'know. If anything, I'd try and get in touch earlier so when you get to the actual Skype session you'll have identified more topics you really want to speak face to e-face about.
The man answers questions if he has the time. Shoot him a PM or email!
Master Milk
Member
Fucking Painwheel. I can't hang against that with Peacock. Nope. Nope. Nope.
I did read this post, by the way. Doing anything on here is another story, but I did read it. >.>
I did read this post, by the way. Doing anything on here is another story, but I did read it. >.>
They seem to have big plans for us...
SolarPowered
Member
Just keep ya head up. This business is cutthroat, man.
Clue me in and I can start up a rumor thread full of ebonics. It'll be a smash hit!
DiscoShark
Banned
If you think Painwheel is bad wait until you come across a good Valentine/Dp Assist. Jesus christ if there is any combination in this game that gets me raging it's that one. I'm convinced Valentine absolutely destroys Peacock in a way no other character in this game does. Her airdash -> attack angle puts her in the blindspot in Peacock's zoning, she can get VERY high in the air and still be able to call assists, and her air normals are all obscenely good so it's more difficult to try to beat her air approach.
Pro tip : Air Alpha Blade -> Knives is a 1 meter middle finger to all of Peacock's zoning and there's nothing you can do about it unless you have an item drop primed and ready to release the moment after she ends the super ... so in this matchup it's actually way more useful to keep an item drop charged just to cover that option if you see that they're using it.
And when Squigly comes out?
Wrap that up.
That said, I've written a ton about Painwheel v. Peacock in this thread.
http://www.neogaf.com/forum/showpost.php?p=42928220&postcount=3451
http://www.neogaf.com/forum/showpost.php?p=43125949&postcount=3459
Hope it helps.
darkblade77
Member
Where do I preorder the LE
As the saying goes, "If you always do what you've always done, you'll always get what you always got". What usually happens is a developer will continue to use the same techniques they used at lower res because they are proven techniques but they will become increasingly wasteful of memory because 2d memory consumption increases are squared. A new developer, on the other hand, can try new things, even things that may be trivial.
Take the supersampling thing: in modern 2D it's almost trivial, you just use a texture with an alpha component, ie just painting you image in photoshop and exporting it without a background, and because the transparency at edges is smooth it'll work well at a number of resolutions. This , however, won't bring the same benefits to people who use the pixel art look simply because that look is inherently tied to resolution.
Or in some cases it could be as simple as cutting sprites in smaller bits to have less transparent spaces, since a given single sprite will always be inscribed in a rectangle and the empty space is still wasting memory. Take that sprite, all the space on the lower left and right is wasted, so just cutting it horizontally in the middle would allow you to resize the lower part and use 20% less space.
DiscoShark
Banned
Skullheart.com
Hit up their forums for one, probably the best place to start if you're looking to get into the nitty gritty of those characters. The team in and of itself is actually super effective though. Run Valentine on point and Parasoul with Napalm Pillar assist.
and uh ... stuff! Here's some video of a really good Valentine player.
http://www.youtube.com/watch?v=1wfS4Q73OSo
DiscoShark
Banned
Solo is probably the most rare team count but I'd say the majority of better players are shifting over to 3 man setups as they get better. 2 obviously makes for a good middle ground. I'd say go with the team count that you can make the most use of, theres no point going a 3 man squad if all of the members cant pull their own weight.
shaowebb
Member
Just wanted to share my animation stuff with Ravidrath again.
Looking around through those sprites and images posted before it seems Skullgirls sprites tend to stay within a range of 900-ish to 2000-ish pixels wide or tall depending on the attack or pose. The DPI is right at 72 dpi as well. I considered running a 150 dpi on my sprite, but I've seen weird stuff happen with resizes before between dpi sizes so why screw around here? 72 dpi is standard web dpi so why screw with it? Thats a screen resolution so why should I care about print dpi stuff? I shouldn't. 72 dpi it is.
For my run cycle I'm running with wider poses since he will be down in a runner's "dig" pose more than a sprinters gazelle or drive phase. In animation its about cool over looking real sometimes and I'm going to add a speed skater style pivot to his torso to imply so much speed in his run that he seems to have to stay low and streamlined and swing his arms to balance the force driven by each stride. I'll post as it comes along. I've been sick, but I'll get to work on this some today and tomorrow now that my color tests are done. I'll let you guys run me over on it as I progress with it since I admire your own team's animation so much. Here's an in progress pic.
Also congrats. It would appear that Big Band is a lock.
Looking around through those sprites and images posted before it seems Skullgirls sprites tend to stay within a range of 900-ish to 2000-ish pixels wide or tall depending on the attack or pose. The DPI is right at 72 dpi as well. I considered running a 150 dpi on my sprite, but I've seen weird stuff happen with resizes before between dpi sizes so why screw around here? 72 dpi is standard web dpi so why screw with it? Thats a screen resolution so why should I care about print dpi stuff? I shouldn't. 72 dpi it is.
For my run cycle I'm running with wider poses since he will be down in a runner's "dig" pose more than a sprinters gazelle or drive phase. In animation its about cool over looking real sometimes and I'm going to add a speed skater style pivot to his torso to imply so much speed in his run that he seems to have to stay low and streamlined and swing his arms to balance the force driven by each stride. I'll post as it comes along. I've been sick, but I'll get to work on this some today and tomorrow now that my color tests are done. I'll let you guys run me over on it as I progress with it since I admire your own team's animation so much. Here's an in progress pic.
Thanks again for all the help. Your a helluva game studio and lord knows you deserve some backers after the turn out you got in Japan for this game and after how your crowd sourcing exploded.shaowebb WIP said:
Also congrats. It would appear that Big Band is a lock.
CrazyTank940
Member
2-3 is basically the best. Period. 3 is probably better if you're good at 3 characters, or at least 2 characters and that assist.
1 character is essentially a gimmick. That I play.
Fellow member of the 1 character club reporting in.
Master Milk
Member
Master Milk
Member
i can't anything
DiscoShark
Banned
Play mee Beats! <3
----
So the weekly Skullgirls online tournament Skullbats are totally still going on and you guys should absolutely enter. It's setup in a format that, win or lose, every person gets the opportunity to play 5 rounds of matches. I entered week #22 and totally won! Some of the people I faced ended up recording footage of the matches.
DiscoShark v. Taluda
http://www.youtube.com/watch?v=nLsXA5_YoBg
^
Facing the guy with crazy innovative Fortune combos (starting from the third match) and pressure strings.
I'm on now if you want to play. I'm pretty tired, but I'll try to play well.
DiscoShark
Banned
Give a guy a year and an 8 character roster and he'll find his way around the basics of the majority cast.
----
Regarding Painwheel ...
The trick actually is to not bother teching the grabs at all. Painwheel is a character that wants to dominate air space, the response to her pressure shouldn't be putting yourself in a situation where she has the inherent advantage. Everytime you jump after one of my pressure strings, whether you tech the grab or not is an example of you picking the wrong decision in that situation.
So what does Filia do instead? She DPs. If I choose to have a flight based offensive than Filia has more than enough tools to deny that, don't fall for the shenanigans, end them with your invincible move designed specifically to take out aerial based approaches.
That's what to do when I've already begun creating pressure, in a neutral situation though you want to do everything in your power to keep me from getting comfortable in flight. You did a great job of using airdash j.HPs in neutral situations when the momentum of the match shifted in your favor but you were ignoring things like straight Fenrir Drive super (when I'm flying), dragon punch into air gregor samson, and LK Air Hairball.
Trust me when I say when you get a handle on this matchup, Filia should have 0 problems dealing with Painwheel.
--
General Filia Tips :
It's great that you have an airdash based offensive and your pressure is solid but after a certain point I simply shifted to just blocking high against all of your pressure, mix it up in that regard and throw in some dash up crLKs after a pushblock and you'll be much scarier.
Raw DP into Air Gregor Samson. It's VERY difficult to punish unless you pushblock guard cancel the initial DP and land right next to your opponent. In any other situation you land in a completely neutral situation after the fact and have successfully disrupted your opponents pressure.
GGs for sure though.
Yeah, I sort of felt like I should have used more DPs. I kept trying to use st.HP as an anti-air. Bad idea on my part lol. I'll definitely try changing up my play/mix ups/reset attempts now though.
Captain_Spanky
Member
Thanks to PS+ I got this updated finally. And now I can play online properly. I'm terrible, in the UK and utterly love the game. So feel free to take me to school!
SolarPowered
Member
Standing on Japanese sales charts for week #4. Doing about as well as it should do against Hatsune Miku and a surge in sales possibly driven by the Jojos anime. Still sitting ahead of Jet Set Radio and Journey. It looks like Skullgirls will stick around the top ten for a good while.
Your explanation makes a lot of sense and definitely sounds high end as hell. This sort of ingenuity sounds like the proper thing we should have seen out of the 2D field a long time ago instead of waiting for console memory to catch up to the demands of the style like Japan has. I doubt tech like this would have convinced Capcom to consider 2D again, but LZG should certainly make the best of next gen consoles compared to the other sprite based fighting game companies.
It looks like someone knows the proper running form. Also, yeah Big Band is totally making it onto the roster barring some crazy calamity.
SG needs a Hatsune Miku pallete.