Since the publishing of the movie is fairly soon, I decided to work some more on the Atlas encode. This time with a dynamic camera that needs a bunch of fixes to make transitions not as jarring.
Link to video
I didn't particularly like the submission because I had no idea what in the blazes of hell was going on, but this on the other hand, is so much better and easier to grasp, even if it is still crazy as hell. Nice job.
Is this what you're referring to around 0:35-0:38, where the movement seems slightly jerky as Sonic moves away from, then back towards, a stationary Tails? Cuz I was noticing that today on my one millionth rewatch of this video and don't really know how to describe it.
If so, I wondered if it would be smoothed out by having the camera position adjusted each frame based on having its own momentum which would accelerate the camera position towards that midpoint, rather than just following the midpoint exactly each frame. (Or maybe some combination of the two, to avoid the camera overshooting its intended target when there's a lot of momentum.)
<Swordless> Go hug a tree, you vegetarian (I bet you really are one)
I wonder if you can increase the camera zoom between both players are far away from each other, this can fix some of the camera bugs?
Not sure honestly, but in LRZ2 there is a level warp, how the camera behaves in this particular spot?
I wonder if you can increase the camera zoom between both players are far away from each other, this can fix some of the camera bugs?
Unless this can be done smoothly, it will be really jarring. It might look weird even if done smoothly. Zooming on pixelated stuff can look odd sometimes if it's not part of the original game, though I can't really articulate why. Maybe it's kinda disorienting? I'm thinking about the way that the original arcade Samurai Showdown dynamically zooms in and out during the battle as an example.
In LRZ2 there is a level warp, how the camera behaves in this particular spot?
Right now it would probably snap to the Sonic-Tails midpoint like any other place in the run, which I guess would be the middle of the level until Tails warped to Sonic's level-wrapped location, where it would snap again. If the camera had its own momentum that would accelerate/decelerate toward here the camera "should" be on that frame, maybe it could swing to the right with considerable speed until it "catches up" to Sonic?
In any case, it looks like some spots will need manual handling anyway - for example at 1:49, I dunno what the game does to handle that spinning orange pipe section, but it messes with the camera in an unexpected way. There's probably more spots like that throughout the run.
<Swordless> Go hug a tree, you vegetarian (I bet you really are one)
Joined: 4/17/2010
Posts: 11495
Location: Lake Chargoggagoggmanchauggagoggchaubunagungamaugg
Selicre wrote:
Since the publishing of the movie is fairly soon, I decided to work some more on the Atlas encode. This time with a dynamic camera that needs a bunch of fixes to make transitions not as jarring.
Link to video
The work you're doing is absolutely incredible! My dream is having this kind of encode for the whole movie, available when it's published! I know it's super hard and takes lots of time, but we could also delay the publication a little bit. I just really think that such a video is the only way to fully understand and enjoy a movie like this. And if we publish as is, and add the atlas encode later, it might not help the publication rating, leading to lower rating than it deserves overall.
As a solution for off-screen chars (sorry I haven't been following the talk), maybe just make the tool write down the frames when they're off-screen, then attach the camera to Sonic for one vid, to Tails for another vid, and make a split-screen encode for those frames? Whether it's vertical or horizontal could be determined by which direction they go off-screen.
Warning: When making decisions, I try to collect as much data as possible before actually deciding. I try to abstract away and see the principles behind real world events and people's opinions. I try to generalize them and turn into something clear and reusable. I hate depending on unpredictable and having to make lottery guesses. Any problem can be solved by systems thinking and acting.
There are so many cool things that have been implemented here. You can see how every magnetized ring eventually gets to Sonic or Tails. This will look real cool on some of the places where this is used heavily, like Hydro City 1 or Ice Cap 2. The view window I think is the perfect compromise between being zoomed out far enough, but not too far. The water effects and sprite layering is handled well. The important objects like rings are always shown. I think the camera's focal point between Sonic and Tails does a fantastic job of not making either of them look "more important" and really complements the cooperative dynamic in the gameplay.
I realize I'm giving a lot of opinions on this tool and I don't want to dominate the discussion, nor do I want to discourage any progress in this endeavor, regardless of the direction. I think this is such an exciting development and the idea it could also be used for other games is crazy to me. With all that being said, here are a couple more thoughts I have:
I worry about adding too many bells to this - things like dynamic zooming and dynamic split screen. Maybe they would look better than I realize, but part of what I think makes the atlas encodes so cool, especially for this game, is to be able to understand the path to the overall level much better, even not having memorized the map. You can see where paths double back on themselves and why. My worry here might also be bogged down in "this sounds hard and would delay the encode if efforts were put toward it".
This example looks significantly better than Hydro City 2 and Angel Island, partly because Carnival Night has a lot of decorative elements in the foreground, and the background of the stage is quite dark already, so the absence of the background layer is not as jarring. r57shell's atlas work includes some of this, though I don't know how it looks for every level, and also isn't zoomed out as far as Selicre's atlas encodes are. This is one of those things that seems like it would look amazing, but the practicalities of it may make it impossible to do in a visually satisfying way (especially in the general case).
In any case, I can't wait to see where this goes next!
<Swordless> Go hug a tree, you vegetarian (I bet you really are one)
Here's a WIP of a Sonic alone any% run I made up to the end of Hydrocity 1. I've been thinking about making this run for some time now to see how fast Sonic could beat the game by himself, but decided to hold off on it until I got better at tasing.
Sonic any%
I'm having trouble doing the AIZ bridge transition skip. I've tried to do it myself and I've also tried with the notes posted some pages back but I just couldn't do it. There isn't a reference run I could look at either so I'm kinda stumped.
EDIT:43 frame improvement in HCZ1
It looks good, I tried improving some parts of it, but managed to save only 1 frame before AIZ1 boss. It doesn’t count, I guess, because I expected more. I don’t think I can save more; however, I may have a time-saving zip for HCZ1, let me check (I’ll post updated information soon).
(upd) I’ve just checked it out, and surprisingly, it is slower than what you did.
Also, as far as I know, bridge skip requires some Tails :)
(upd) finally found a place where I can save more than 1 frame: around 0:29 in HCZ1, you can abuse the enemy to jump higher and reach the top floor faster than you did (I saved 4 frames in my test attempt): start rolling on the last possible frame before hitting the enemy (a fly) and jump immediately. If you jump on the same frame when you turn rolling and hit the enemy, you’ll get extra boost, just enough to land on the top. Of course, you’ll need to use insta-shield to regain horizontal control, but it isn’t a problem there. If you want, here is an example gmv.
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
I can probably resync AIZ2 to save that frame.
That bridge skip must've somehow been done in the past with Sonic alone because otherwise those notes Marzo posted some pages back would've outright said it's impossible. So I'm just gonna wait until he responds.
I have tried that zip myself. The route change is where most of those 43 frames came from.
I did not know you could use a badnik to boost your jump like that. Nice find. You already fully optimized it too :)
So were those really the only 2 places you could save time in?
Joined: 10/6/2011
Posts: 1697
Location: RU · ID · AM
kaan55 wrote:
So were those really the only 2 places you could save time in?
Definitely yes, I tried my best at many other places and couldn’t improve anything. I’m really impressed, the overall quality of your run is outstanding, you should continue working on it — and I will be keeping my eye on the progress to see if I can find any improvements in the future.
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
Definitely yes, I tried my best at many other places and couldn’t improve anything. I’m really impressed, the overall quality of your run is outstanding, you should continue working on it — and I will be keeping my eye on the progress to see if I can find any improvements in the future.
Well, I spent a lot of time making sure I saved every frame possible. I'm glad to hear it's working. It's also glad that you're going to keep an eye on the run because I sometimes overlook stuff :p
improvements
Anyways I managed to save 10 frames in AIZ1 after finding out that the collapsing bridge hitbox extends during a specific frame of its animation. Luckily I was able to resync all of Act 2 by manipulating my position.
I've pretty much finished HCZ1 as well. I do see potential to save more frames during the boss but I'm having trouble doing so because of the thrusters.
Definitely yes, I tried my best at many other places and couldn’t improve anything. I’m really impressed, the overall quality of your run is outstanding, you should continue working on it — and I will be keeping my eye on the progress to see if I can find any improvements in the future.
Well, I spent a lot of time optimizing the run and testing pretty much everything that could save time. I'm glad that you're going to keep an eye on the run because I sometimes overlook stuff :p
improvements
Anyways I managed to save 10 frames in AIZ1 after finding out that the collapsing bridge hitbox extends during a specific frame of its animation. Luckily I was able to resync all of Act 2 by manipulating my position.
I've pretty much finished HCZ1 as well. I do see potential to save more frames during the boss but I'm having trouble doing so because of the thrusters.
Joined: 10/6/2011
Posts: 1697
Location: RU · ID · AM
Yeah, this boss is annoying, especially with a huge character like Amy (though the DEZ1 boss is even worse). I can’t say for sure what exactly is the best pattern of moving to defeat it, it needs making different attempts and taking the fastest of them.
Also, only 14014 frames for now, but already almost 40000 rerecords, I am deeply impressed by the amount of effort you put into this, it is crazy!
Good thing is that HCZ2 is probably going to be one of the “easy” levels, and you’ll be able to take some rest while TASing it :)
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
I thought of a possible way to save those frames. What's really worrying me is the quick screenlock death I'm gonna have to do.
Oh yeah HCZ2 and MGZ will most definitely be a nice break. Actually the rest of the S3 half doesn't look too hard either. Most of the levels in the S&K half will require quite a bit of testing though.
Speaking of testing I discovered that it's possible for Sonic to do the CNZ1 level wrap from Knuckles' area by himself.
I managed to save 2 more frames. It looks like this is as fast as I can kill it. Now to figure out how to do the quick death.
2 framesEDIT: I think I figured out bridge skip for Sonic alone. Give me some time to make an optimized gmv.
This time I've really figured out the bridge skip. It turns out the issue I was having is that I needed to keep the object slot the switch would normally go to (which was B128) filled in order to force it to load in B1BC. I had to sacrifice 7 ingame frames which I am more than okay with that considering this saved 335 frames overall.
I also figured out how to do the quick death in the Hydrocity Act transition. long story short actually pretty simple. I had to sacrifice 11 ingame frames which is again no big deal. Especially since this one saves ingame time.
Sonic alone any%EDIT: Here's HCZ2. I'm not quite done optimizing the boss yet.
Hydrocity Act 2
WIP up till the end of MGZ. I also saved 2 more frames on the HCZ2 boss which unexpectedly ended up being more annoying than the Act 1 boss lol
Marble but with a Garden
Joined: 10/6/2011
Posts: 1697
Location: RU · ID · AM
kaan55 wrote:
WIP up till the end of MGZ. I also saved 2 more frames on the HCZ2 boss which unexpectedly ended up being more annoying than the Act 1 boss lol
Marble but with a Garden
Good job! With Amy, this boss is easy as hell, nearly impossible to NOT make it frame-perfect… :)
Congrats on having your run using death to save time now.
0:36.02 is so close to 0:35 — usually when I get such kind of result, I try my best to make it an in-game second faster — because it looks visibilly cooler.
Also, your method to beat MGZ boss is 202 frames faster than mine, I need to redo this boss fight in my Amy@S3K WIP now :)
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
I usually try to save the second too. Luckily there was an optimization I overlooked which saved 4 frames. Also saved 22 frames from lag for no seemingly no reason lol
MGZ 0:35EDIT: I managed to get a 34. I'll share how when I finish CNZ1.
Yeah I didn't really think about that improvement with the MGZ boss until I noticed how high you are compared to the boss. I also did the same thing in the ring attack. You basically just hit the boss the frame it disappears.
Joined: 10/6/2011
Posts: 1697
Location: RU · ID · AM
Let me guess: you performed the horizontal underflow a bit higher, just enough to land on that ground afterwards without the spindash and the big jump?
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
Good guess but no. There was a clip DMTM did last year in one of his videos immediately after the level wrap. It's actually about 5 seconds faster but since the screen took ages to reach the boss I could only save 62 and that's with camera manipulation. I also saved 9 more frames by doing quicker movement to the ring monitor.
Anyways here's CNZ1. I will improve the boss later after I take a break.
CNZ1 v1EDIT:25 frames saved on CNZ2 boss. Yeah I'm not sure when I'm gonna stop being able to improving it.
CNZ1 boss 25 frames
Joined: 10/6/2011
Posts: 1697
Location: RU · ID · AM
kaan55 wrote:
EDIT:25 frames saved on CNZ2 boss. Yeah I'm not sure when I'm gonna stop being able to improving it.
CNZ1 boss 25 frames
Haha, yeah, I also can improve my own runs many times again and again. Anyways, if you roll before the jump at frame ~3218, you’ll be able to keep the high speed for a few frames longer, but I am not sure if it saves at least one frame. Also, that rock, that you touch at frame 3460 — I know for sure that you can reach it’s top using 2-frame jump (it requires high Y subpixel position though). Even in Aglar’s run he reaches it with 3-frames jump, which, in theory, can mean that it’s improveable, but it’s not 100%. In fact, that particular place after the loop (and including the slope before) really sucks — it’s hard to select the best way to do it, some tricks may be possible with good positioning. For example, I am prettu much sure that in Aglar’s run it’s possible to reach the top ground by double-jump after the loop, because I nearly did it myself. Sadly, “nearly” is not “did”. If I succeeded, it would save 2-3 frames, as I remember, and *maybe* would give a better Y subpixel value, enough to jump to the rock using 2-frame jump…
Also, I love your new MGZ, very smart use of the monitor and very nice clip. Poor camhack though :'(
CNZ is made exactly as I expected :)
S3&A [Amy amy%] improvement (with Evil_3D & kaan55) — currently in SPZ2
my TAS channel · If I ever come into your dream, I’ll be riding an eggship :)
Frame 3218- I've tried a lot of things here and this turned out to be the fastest. I intentionally didn't do a roll jump because I wanted to hit the badnik with as less downward momentum as possible which allowed me to maintain a bit more speed and fall a bit faster.
Frame 3460- In this instance actually I could reach the rock with a 2 frame jump but a 3 frame jump gave me a bit more speed before doing the next jump off the rock. After you pointed out the improvement in this section I spent a lot of time optimizing it.
Yeah I like that monitor strat. Freaking camhack in MGZ though. I had to do that part of the level blind. I'm so glad that there's like 4 seconds to spare before the camera reaches the boss.
I've improved CNZ1 by 9 more frames. 1 frame from the main level and 8 frames on the boss which looks ridiculous now.
Carnival Day Zone Act 1EDIT: So uhh I managed to save 16 more frames. The boss fight currently only 130 frames slower than the one in the current 2-player any% tas. What's really absurd is that I see potential to save more.
CNZ1 boss v3?
