PurpleSun | [TAS] SMG2: Flip-Swap GS1 in 5.96s @PurpleSunTAS | Uploaded 7 months ago | Updated 2 minutes ago
Luigi hyperjumps eastward into a Green star in 5.96 seconds [IGT, 358 frames].
-2 over my 2022 TAS: youtu.be/RikJY-wwPDk
@Joselle_Astrid really had everything figured out long ago, and her research pops back up every now and then, and something within it is then realised for optimisation.
This time it's that aerial movement is ever so slightly faster when going east or west (side-facing the camera) compared to north/south (away from or towards the camera respectively). docs.google.com/document/d/14DAcbHw0OPSL2psmxpvAPtpAVdF_AYkYJyeE8GNZKjM/edit
This helps both in getting a slightly further long jump and single jump so they can be done earlier, and in a faster flip panel long jump, overall saving 6 frames (note: a smaller bit of improvement may have also been from a better understanding of gaining height from flip panel jumps).
As the camera is so free in this level, it's much easier to get the potential out of the east/west gain, but a lot of levels would have trouble due to either limited movement or forcing itself back to the over-the-shoulder perspective - we don't have SM64's luxury of knowing if we're keeping straight while the camera moves, so it's harder to get the benefits.
#supermariogalaxy2 #toolassistedspeedrun
Luigi hyperjumps eastward into a Green star in 5.96 seconds [IGT, 358 frames].
-2 over my 2022 TAS: youtu.be/RikJY-wwPDk
@Joselle_Astrid really had everything figured out long ago, and her research pops back up every now and then, and something within it is then realised for optimisation.
This time it's that aerial movement is ever so slightly faster when going east or west (side-facing the camera) compared to north/south (away from or towards the camera respectively). docs.google.com/document/d/14DAcbHw0OPSL2psmxpvAPtpAVdF_AYkYJyeE8GNZKjM/edit
This helps both in getting a slightly further long jump and single jump so they can be done earlier, and in a faster flip panel long jump, overall saving 6 frames (note: a smaller bit of improvement may have also been from a better understanding of gaining height from flip panel jumps).
As the camera is so free in this level, it's much easier to get the potential out of the east/west gain, but a lot of levels would have trouble due to either limited movement or forcing itself back to the over-the-shoulder perspective - we don't have SM64's luxury of knowing if we're keeping straight while the camera moves, so it's harder to get the benefits.
#supermariogalaxy2 #toolassistedspeedrun
![[TAS] SMG2: Flip-Swap GS1 in 5.96s](https://i.ytimg.com/vi/fnTWLzyNF_M/hqdefault.jpg "[TAS] SMG2: Flip-Swap GS1 in 5.96s
Luigi hyperjumps eastward into a Green star in 5.96 seconds [IGT, 358 frames].
-2 over my 2022 TAS: https://youtu.be/RikJY-wwPDk
@Joselle_Astrid really had everything figured out long ago, and her research pops back up every now and then, and something within it is then realised for optimisation.
This time its that aerial movement is ever so slightly faster when going east or west (side-facing the camera) compared to north/south (away from or towards the camera respectively). https://docs.google.com/document/d/14DAcbHw0OPSL2psmxpvAPtpAVdF_AYkYJyeE8GNZKjM/edit
This helps both in getting a slightly further long jump and single jump so they can be done earlier, and in a faster flip panel long jump, overall saving 6 frames (note: a smaller bit of improvement may have also been from a better understanding of gaining height from flip panel jumps).
As the camera is so free in this level, its much easier to get the potential out of the east/west gain, but a lot of levels would have trouble due to either limited movement or forcing itself back to the over-the-shoulder perspective - we dont have SM64s luxury of knowing if were keeping straight while the camera moves, so its harder to get the benefits.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] SMG2: Sweet Mystery Green Stars](https://i.ytimg.com/vi/giU2MNLEkAE/hqdefault.jpg "[TAS] SMG2: Sweet Mystery Green Stars
Green Stars TAS WIP
Star order matters a very little bit due to the Choppah cycles explained more below. GS1 first is 1 frame faster.
Green Star 1 - 7.30 IGT - 1:18:04.36 real-time split (1:02.70 segment)
Even with cutscenes, this is a 1f IGT improvement over my 2021 TAS (uses Mario as his acceleration is better): https://youtu.be/xrkFOBb8Qe4
2P egg break is tied fastest with a backwards double jump that spins at the end.
I had to delay jumping and fluttering on Yoshi because of a cooldown on spinning. Backwards fluttering stays at the same speed after a short time (the third flutter in this TAS).
The flutter acceleration - and subsequent speed (about 29 horizontal speed) - gained from the spin makes this strat a few frames faster than fluttering as early as possible.
Bouncing off the berry (possible by having Yoshi stick its tongue out with B) is a strat I lifted from Jeromes GS2 IL: https://youtu.be/Yz8QEhgTgY8
Bouncing gets a burst of speed and brings flutters straight to the cap of 32hspd.
It caps at 32hspd when ascending (y speed above 0) but can go a little higher when descending. The burst of speed is maximised by keeping a flutter through bouncing, as this makes Yoshi fall earlier, being able to get up to 38hspd rather than the 34 when actually bouncing.
The bounce saved 0.2s.
An extra frame was saved thanks to the Choppah, which is where star order came in. With cutscene cycles it ends up right next to the berry, and with a 2P freeze I brushed by it so it nudged Yoshis trajectory more towards the star. This saved having to strain harder during the flutters, which would have lowered the burst of speed.
Dismounting gets another burst of speed including a mini-warp of about 67spd.
Green Star 2 - 11.20 IGT - 1:18:58.55 real-time split (54.19s segment)
Backwards jump with a spin at the end was fastest egg break by 9 frames.
Same things apply as above with the spin-start flutter and berry flutter-bouncing.
Eventually the burst of speed runs out and there isnt much bonus falling speed when going (mostly) backwards, so I go for height-optimal flutters near the end to be able to dismount as early as possible.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS demo] SMG2: Starshine Beach GS3 in 45.46s w/ cutscenes](https://i.ytimg.com/vi/hg8Tyo10qE8/hqdefault.jpg "[TAS demo] SMG2: Starshine Beach GS3 in 45.46s w/ cutscenes
Luigi does 4 cloud long jumps to a Green star.
Long jump ground pounding into the water at the start is ~0.25s faster than spinning backwards from the land cancel straight into the water.
Camera turns are done to keep the clouds off-screen so that they stop moving. This makes up for the bad cycle that is a result of cutscenes, and sets up a quick triple jump off the lowest cloud and wallkick up to the highest cloud to get up the tower ~2s faster than the Pianta Im a chuckster! strat.
It also makes the long jumps more of a nuisance to optimise.
I found the backflip by accident when holding Z to try to get an instant sideflip. I believe the backflip only worked because Luigis vertical speed was slow enough that he snapped onto the clouds floor instead of lifting over and landing down on it. Landing on a cloud doesnt let you backflip.
Backflipping straight to the Cloud Flower was 0.43s faster than sideflipping and spinning and then long jumping (the next fastest movement).
For the main trick, I originally tested how close you could get with a long jump and 3 cloud long jumps, testing from both the edge of the tower and the palm tree. The palm tree was a better starting point, but still only reached the star about two thirds of the way between it and the water below.
So instead I tried to cloud LJ from the tower and grab a respawned Cloud Flower at the same time to regain a third cloud for a third mid-air cloud LJ.
Initially, this also failed because the grounded cloud LJ didnt accelerate in mid-air like cloud LJs tend to do (they accelerate to 34hspd, which seems to be a general speed cap - infinite flutters also reach this speed).
As I learned from Shiverburn and Cloudy Court, sometimes the first cloud LJ of the bunch needs something special to get that fastest speed; Im not sure what, but there is always a way to eventually get it.
In this case, it seems to be that the LJ has to be near the edge of the tower (which Im beginning to think is the case for all of them aside from mid-air ones).
This meant that I had to time the cloud spin to long jump near the edge while also just about picking up the Flower from the back of Luigis hitbox. This part definitely had little effort in optimising, as I was just trying to get the speed in the first place.
From there, it was as simple as getting 2-frame cloud cancels and LJs to optimise height, finishing with a spin into the star.
This route is 6 or 7 seconds faster than an altered IL route, swapping out the backflips that speedrunners use for a cloud LJ and two cloud sideflips.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![Supermassive GS1 0.5xA strat using Luma and Goomba [Luigi]](https://i.ytimg.com/vi/iXSJYV5za-4/hqdefault.jpg "Supermassive GS1 0.5xA strat using Luma and Goomba [Luigi]
This is tool-assisted.
Havent been able to get the Goomba float without spinning into the floating animation (it gives more backwards speed), so getting onto the Luma without jumping was the next step.
P2 let go of the Luma by spinning.
This also saved 0.5x for Green Star 2, as the float can make it at least onto the pipes, or far enough to the other side that P2 can grab the checkpoint before death.
Havent been able to reach the launch star with this yet so the other stars are still 1x/1.5x with either this strat + jump into launch star, or doing the Goomba float from further along with a backflip which is enough to get into the launch star.")
![[TAS] SMG2: Sky Station GS2 full game route in 3xA](https://i.ytimg.com/vi/iakuD0IpylY/hqdefault.jpg "[TAS] SMG2: Sky Station GS2 full game route in 3xA
This focuses on overall A presses instead of just gameplay, using 2 jumps and a cutscene skip that are unnecessary to complete the level but are done to avoid dialogue.
The cutscene skip and backflip avoid the Luma at the start who has 3 textboxes, so this combination is a profit (or loss?) of 1xA.
I then perform an Odme textbox skip, where I land inside the second Lumas cutscene trigger and then get pushed out of it by the edge of the ledge while holding neutral, which sets up the cutscene but doesnt actually start it, and then I press A (as seen with the jump) which seems to press an invisible textbox that triggers the Lumas transformation. I guess this could count as a dialogue press? But its done in gameplay. This is also a profit of 1xA.
So overall, this goes down from 5xA (0x gameplay) to 3xA (2x gameplay).
As for the rest of it, I hold that last A press to see what kind of fun I can have with a bit of extra spin height.
To bounce on the Octogoomba, I use Annoyings tech of a spin knockback cancel - hugging the wall, holding Z and spinning which gets the benefit of a knockback without keeping the spin - so I get the height to bounce but dont hit it with a spin.
Running up the slope is something Nim found, as the cameras distance from a slope affects its slipperiness somehow: https://youtu.be/7gjMQlb7KZY
LUMA BOUNCE STAR GRAB POG
#SuperMarioGalaxy2 #AButtonChallenge")
![[TAS] Super Mario Galaxy 2: Melty Monster Green Star 2 in 36.46s](https://i.ytimg.com/vi/ig1Qc9T3mCE/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Melty Monster Green Star 2 in 36.46s
Luigi jumps the lava waves with P2 helping to collect a Green Star in 36.46 seconds [IGT, 2188 frames].
The pull star section has room for improvement as I only tested one other route - there could be better ways through that first half to lead into the final couple of pulls.
Its worth noting that you can skip the first launch star but its yet to be done faster: https://youtu.be/ESACxGIBFbE
The main show is in the second planet.
Nemi43 found that the very front of the lava waves make a short wall, which gives you a fast horizontal lava boost (a lava wall boost), and the speed cap doesnt reset for the following lava boosts despite them being normal vertical boosts: https://youtu.be/A2DOqAVK-uo
As P2 can grab and hold a Life Mushroom from earlier, I was able to get the lava wall boost much earlier without worrying about health. This saved ~2.5s over Nemis 1P method.
I had to delay the backflip a few frames as an instant backflip wouldnt have hit the wall. Despite this, backflipping was 8 frames faster than long jumping, single jumping, and instant backflip + spin.
Delaying the first cutscene skip for a few frames would have saved in-game time for this, but this TAS was part of real-time testing of cutscenes vs no cutscenes.
I did also try combining a jump cancel with something, but I found you can only jump cancel if you delay activating the previous launch star - this might be to do with tilt. Launch stars are weird.
For the lava boosts, I started to strain to the left near the end to try to cancel or at least shorten the running animation, as you can only single jump or long jump out of this. I tried cancelling by running off the edge, but it was too slow getting back on far enough to do anything useful.
I first thought I had shortened the running animation, but on review, it didnt actually do anything. But with the leeway of where I needed to be for the final jump, it didnt lose any time.
Single jumping out of the start of the animation only lets you backflip 4 frames later than letting the animation play out itself.
Backflip spin saved ~0.85s over triple jump spin. The leeway mentioned above is that the lava wall Luigi backflips towards is very close, so going straight backwards wouldve got burned too soon.
I had managed to get the spin directly into the tornado to avoid a lava boost, but this loses time as the resulting instant tornado spin is too low to grab the star.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy 2: Honeybloom Secret in 14.93s](https://i.ytimg.com/vi/isFyMAXZtao/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Honeybloom Secret in 14.93s
Luigi and P2 find the Secret Wall Jump secretly, thereby secretly adding another secret star to their secret collection of secret stars, secretly. In 14.93 seconds [IGT, 896 frames].
The fly-in cutscene is 11 frames faster for in-game time, as the timer starts ~30 units further into the level, and Luigi grabs the vine further on, having flying speed for a few frames into the timer.
I had not-so-secretly done this star a few months ago, looking to improve on Sonicpackers 15.16 (https://youtu.be/_kzoAt4X5Uc), but could only manage to tie this time with a different wall jump route (https://youtu.be/c8orefszdIU). But with better understanding of walljumps and techniques like neutral walljump strain (going neutral on the joystick when walljumping to allow for earlier strain), I went back to it.
2 frames were saved in the first half leading to the first walljump after the swing.
This took a lot of experimenting, but the eventual result was using neutral walljump strain from the second walljump to hold back before the spin onto the Goomba, and using P2 to freeze the Goomba into a certain mid-air position from its jump. And also letting go of A a couple of frames before the peak of the spin to bounce off a specific part of the Goomba.
This combination grabbed the swing 1 frame earlier, and at a specific height (1931.436157 Y position) necessary for the part below.
The height and speed from the following swing jump had enough height to allow me to let go of A at the peak of the following spin, which gives a significant speed boost (14.49 horizontal speed compared to 8.74 holding A). This grabbed the wall 1 frame earlier.
Then came learning that holding different directions in walljumps can have a very slight effect on vertical speed, slight meaning from tenths to thousandths of units. Holding the direction away from the wall (the walljumps direction) will lower the vertical speed while having faster horizontal speed, and holding towards the wall (against the walljumps direction) will raise the vertical speed.
Neutral walljump strain pushes the differences further.
While that wallgrab had been 2 frames faster, it was also 12 units lower than the previous TAS, so I used the higher walljumps holding back towards the wall - as well as a better spin from the first walljump to the second - to try to claw back some of the lost height. This resulted in just 1 frame saved overall (after 6 hours of work).
I then went on Sonicpackers route with the knowledge of slightly lower walljumps.
The first walljump and spin had to be higher than what was possible as otherwise the third walljump wouldnt make it to the wall above the crystal, so a Wiimote spin holding A for 2 frames was used, followed by higher walljumps to get high enough to barely grab the bottom of that wall.
Lower walljumps were then used, however, these werent necessary as they didnt lead to an earlier landing.
But for the final walljump, neutral walljump strain was useful for straining up towards the star and also getting further left to set up a straighter line to the star, saving 3 frames.
Holding neutral just before landing kept momentum to allow for an earlier long jump, followed by another long jump to the star.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
![[TAS] Super Mario Galaxy: Snow Cap in 1:07.52 [Mario]](https://i.ytimg.com/vi/j0Uo4HQZ3-c/hqdefault.jpg "[TAS] Super Mario Galaxy: Snow Cap in 1:07.52 [Mario]
TAS by Xander and PurpleSun.
Xanders channel: https://www.youtube.com/channel/UCL8JNpJkmcX43bxcWhsXfAA
Mario catches bunnies in snow in 1 minute 7.52 seconds [4051 frames].
Snow Cap was the setting for the first SMG TAS competition; you can see the results video here: https://www.twitch.tv/videos/852960154
Xander started out by picking up the shell with a spin cancel skid, 6 frames faster than backwards bouncing off the shell in mid-air.
He then held neutral just before throwing the shell so that Mario would tilt down, throwing the shell into the bottom of the chest to open it 1f earlier.
NOTE: We did try using mid-air shell bounces - as well as weird long jump momentum - to skip the chest, but its not close enough, and by the time youd reach the planet, the sling star would already have reached it.
For the flipswitch route, going left ended up slightly faster than going right (I dont remember the exact numbers, but no more than 1s). Xander used P2s IR cursor to clear the snow in front of Mario so it wouldnt slow him down - long jumping is slower for that second flipswitch.
While long jumping after the second switch, Xander uses the IR cursor to clear the snow on a rabbit-hole to set the first bunny up. It jumps away from Mario to the Fire Flower side of the planet, ready for later.
I took the reins on the fourth switch, doing a long jump around the edge and holding back for 3 frames before going over the edge so that Mario would stay close to the ground as he curved over. This saved 10 frames over running around.
I do a single jump so that I can land on the Fire Flower and run earlier than walking into the pit, which is helpful for the fireball strategy.
While Xanders strat of using P2 to move the bouncy bulb to hit the bunny can catch that bunny earlier, a better fireball strategy can come from delaying the bunny catch, hitting it with a precise starbit to send it on a more diagonal trajectory.
I throw a fireball out of the hole before catching the bunny (this was why I needed to start running earlier, to get to that point near the edge), and then jump and fireball out of the hole again to have two quick fireballs melting the next bunnys snowman disguise.
However there may be a couple frames of improvement here, as I had to hold left while jumping to avoid an instant ledgegrab - holding right the whole time would mean more speed overall.
Then two spin cancel jumps send fireballs to the shells snowman while keeping momentum, with the second jump curving around the edge.
While long jumping, I shoot the snowman bunny with a starbit so that it goes backwards into Marios immediate direction, spin cancel long jump the shell (delaying the long jump for a frame to pick the shell up earlier), and throw the shell just before catching the bunny - the shell self-destructs if you catch a bunny while holding it/while its spinning to you.
This suspends the shell in mid-air during the cutscene, and it then continues into the chest once the dialogue is over.
I long jump again, holding backwards for 2 frames before getting over the chest so Mario would stay closer to ground and grab the bunny as soon as it is grabbable, wrapping up the chase with 134 on the countdown clock.
Finally, Xanders buffered crouch acceleration (hold Z before gaining control, let go after first frame of control) was used to long jump 1f earlier, and spinning into the star is a couple of frames faster.
#SuperMarioGalaxy #ToolAssistedSpeedrun")
![[TAS] SMG2: Beat Block Green Stars](https://i.ytimg.com/vi/jX5mFAbKoA8/hqdefault.jpg "[TAS] SMG2: Beat Block Green Stars
Green Stars TAS WIP
Star order matters here due to the beat block cycle. Since GS1 mostly doesnt touch them, its GS2s route that matters, and the no-cutscene cycle has slower blocks midway and near the end, making it 0.5s faster to do GS2 first with cutscenes.
Green Star 2 - 13.58s IGT - 1:07:56.55 real-time split (1:09.44 segment)
Due to how close it is to the edge, the crystal at the start has no way to setup an extremely fast zipped slope jump, so it is faster to be left alone.
In the long jump sequence, running speed can be kept to speed up the single jump-double jump up the block, and then keep speed for the crystal.
This first crystal slope jump is a little slower than the others due to the limits of a quick setup, plus aiming the right trajectory with a good facing angle, which would minimise turning time and speed loss to jump straight to the next crystal.
With the next crystal jump, I get 60hspd and a great trajectory to land safely on the yellow block. Landing safely is important so that landing with neutral joystick can conserve speed into the following single jump (capped at 25spd). Landing on the edge in neutral causes 0spd.
The facing angle is also important as I can easily strain the single and double jump to set up the last crystal jump.
This last jump gets 73hspd, finishing with a single jump straight onto the green block and instant backflip spin to the star.
Green Star 1 - 9.51s IGT - 1:08:49.16 real-time split (0:52.51 segment)
My hope with this crystal slope jump was a facing angle and trajectory that could lead to a conserved speed single jump onto the side of the last crystal to conserved speed double jump into the star.
But I couldnt get that to line up. Good angles led to bad trajectories off the map, and good trajectories had not so good angles.
So I had to make do with a 71spd slope jump and quick turn to long jump onto the side of the crystal for a long jump spin to the star.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] SMG2: Freezy Flake GS2 in 37.95s](https://i.ytimg.com/vi/j_J_UWSEdpM/hqdefault.jpg "[TAS] SMG2: Freezy Flake GS2 in 37.95s
Green Stars TAS WIP
37.95s IGT - 60:59.31 real-time split (1:34.05 segment)
Freezy GS1/2 were two of the few stars I hadnt tested prior to starting this TAS, as star order didnt seem like itd matter (more about that later).
The first launch star is a balance of optimisation for the earliest statue break (not waiting so long for the launch star to activate) while still being able to get to the launch star on its first active frame.
While other routes looked cooler (e.g. getting turned backwards in mid-air to have more speed), the fastest by 29 frames was how RTA runners do it (used Pottowws IL as a blueprint: https://youtu.be/QWP7saYcBaw?t=10).
An instawalk from the launch star is common here, but interestingly, entering the launch star while turning ends up getting landstun, and entering while standing completely still gets faster and earlier acceleration.
1-flower cycle was attempted but failed (https://youtu.be/6GEkxIaaQ-k), missing out by around 0.6s.
With certain movement, conical slopes (e.g. the mountain, some crystals) can cause some very buggy fast slides for a frame or two, leading to glitchy slope jumps.
The consequence of slope jumps being patched in SMG2 is that they now stay at a constant speed and trajectory.
It is difficult to control the trajectory, and the jump in the 1-flower attempt has a much more wayward and slow setup after getting the 2nd flower.
As with the Tall Trunk slide, you dont lose any time with whatever movement you do provided you dont jump a lot.
However, exiting the tunnel without jumping is a little slower and doesnt let you spin/throw fireballs. Also, the gravity you leave the tunnel at will have an affect on your angle, sometimes changing angle mid-air.
Backwards exit with a mid-air fireball was 35 frames faster than the RTA method: forward exit, jumping on the lower brick block and fireballing around the corner.
Star order almost mattered at the end. The fastest regular route (by 6 frames) was to long jump onto the flying ? block, and its cycle changes with cutscene skips.
But another glitchy slope jump was found on the Goomba statue, getting a 67hspd jump that barely misses the lava geyser. Unfortunately this kind of jump couldnt make it onto safe ground, but the lava boost and movement on the sloped boundary was 20 frames faster than holding it back onto the ice platform and long jumping.
#supermariogalaxy2 #toolassistedspeedrun")
![[TAS] Super Mario Galaxy 2: Beat Block Green Star 2 in 22.16s](https://i.ytimg.com/vi/jkGB2xNf978/hqdefault.jpg "[TAS] Super Mario Galaxy 2: Beat Block Green Star 2 in 22.16s
Luigi eeeeeyahs to the beat to collect a Green Star in 22.16 seconds [IGT, 1330 frames]
I started on Green Star 1 as I was testing both stars for cutscene effects. Havent actually done a cutscene run with either yet, so Ill update this with results later.
This isnt fully optimised, but its certainly fast. At the start, theres a good use of a short spin that doesnt lose too much speed so that the next long jump can happen earlier than normal.
I did have to slow the penultimate long jump on the green block cycle at 00:25 to be able to land for the last long jump in that section.
The second long jump up the steps was a frame-perfect long jump on the very edge, which let me conserve running speed when I landed, which was useful in getting a faster line-up to long jump to the corner of the last green block.
I landed awkwardly on this, which caused Luigis backflip to be stuck in a walking pose.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun")
