Uploaded June 2020 | Updated September 2026, 2 weeks ago
Mario cataquacks to the skies in 54.75 seconds [3285 frames].
This emulates a full-game run by including cutscenes, which gives a deficit of ~13s compared to IL runs.
The first section was fairly simple, freezing a boulder so it doesn't get in the way, and using a 2P jump from a wallkick to get up to the launch star.
The next section has the most room for improvement - long jumping to the grass cube is the fastest method, but there could be better angles to land earlier and set up the triple jump earlier.
A spin was better than a 2P jump following the triple jump, as it didn't stop to change gravity and meant I could fall further backwards, allowing me to pull off a fast swinging route around the planet until the penultimate Cataquack flung Mario up the last point.
As a backflip 2P jump to the launch star wasn't possible, the fastest method was to wait for the last Cataquack to get around the edge to launch up - a tilted triple jump + 2P jump was possible but slower - and jumping before the Cataquack arrived meant it could launch faster than just standing there.
Finally was the coolest strat of the level. A backflip 2P jump into a glitchy wallkick off the bottom of the flower, into a spin knockback to quickly land on the flower and ground 2P jump off the flower into the star.
#SuperMarioGalaxy #ToolAssistedSpeedrun
Mario cataquacks to the skies in 54.75 seconds [3285 frames].
This emulates a full-game run by including cutscenes, which gives a deficit of ~13s compared to IL runs.
The first section was fairly simple, freezing a boulder so it doesn't get in the way, and using a 2P jump from a wallkick to get up to the launch star.
The next section has the most room for improvement - long jumping to the grass cube is the fastest method, but there could be better angles to land earlier and set up the triple jump earlier.
A spin was better than a 2P jump following the triple jump, as it didn't stop to change gravity and meant I could fall further backwards, allowing me to pull off a fast swinging route around the planet until the penultimate Cataquack flung Mario up the last point.
As a backflip 2P jump to the launch star wasn't possible, the fastest method was to wait for the last Cataquack to get around the edge to launch up - a tilted triple jump + 2P jump was possible but slower - and jumping before the Cataquack arrived meant it could launch faster than just standing there.
Finally was the coolest strat of the level. A backflip 2P jump into a glitchy wallkick off the bottom of the flower, into a spin knockback to quickly land on the flower and ground 2P jump off the flower into the star.
#SuperMarioGalaxy #ToolAssistedSpeedrun
![[TAS] SMG2: Beat Block GS2 in 0xA Presses [Luigi]
Triple bouncing off a Goombeetle again helps with a Green Star.
This branched off the GS1 run, with new input starting when I land next to the Goombeetle used in GS1.
Bloopiero also had already done this as Mario a few years ago, but we like to have videos as both characters.
I managed to get a specific spinbonk that happens instead of a ledgegrab, which gives a lot of height, and it bumped into the ? block, slowing the backwards recoil so I could land on the ledge.
I also used the ground pound spin reset tech to get onto the Goombeetle, which wasnt necessary but looks cool.
#SuperMarioGalaxy2 #AButtonChallenge [TAS] SMG2: Beat Block GS2 in 0xA Presses [Luigi]](https://i.ytimg.com/vi/vUIo2HBB80Q/mqdefault.jpg)

![[TAS] SMG2: Hightail Falls without Yoshi
Hightail Falls Hot-Stepping Dash Pepper is possible without Yoshis help.
I had started this in March of last year, but only reached some way up the huge wall on the second planet.
I came back to this recently - with everything Ive learned since then - and slowly figured out how to scale the wall and beat the rest of the level, thanks to these wallkick techniques.
ANGLED WALLKICKS
When moving toward a wall in the air, if you strain to the left or right of the wall, your next wallkick will be angled that way. This angle can be made sharper if you approach the wall with a spin, even getting up to 90°.
DOUBLE WALLKICKS
If you grab a wall in the corner of two adjacent walls and kick off it, you can get a second wallkick off that same wall.
This cornered wallkick can either be the first (0:56) or the second (0:50 & 2:24).
It can even allow for a wallkick off a wall opposite your facing angle (0:57 would not have worked from the green slope without the cornered wallkick at 0:56).
OUTER CORNER WALLKICKS
Inner corner wallkicks are a long-known thing in Galaxy, where you use the sharpest angled wallkicks to climb up two adjacent walls from inside their joining.
Outer corner wallkicks are also possible in this way, but they would not be effective if not for another key technique...
NEUTRAL FRAME WALLKICK STRAIN
If you go neutral (bring the analog stick to centre) for at least 1 frame after wallkicking, you are allowed much harder strain than normal, making it much easier to control your direction.
This is used everywhere in this video, and is particularly helpful for outer corner wallkicks - to get a wallkick off the other wall, youd need to angle the kick away from it, but you can strain backwards with this technique to avoid getting too far away.
And one more thing:
SLOPE JUMP
Galaxy 2 is known for having patched the first games slope jump exploit, however, you are still able to spin after jumping off a slope if the jump is off the very edge of the slope. This was done at 1:21 to be able to bounce off the Goomba high enough to grab the wall above.
#SuperMarioGalaxy2 #MarioChallenge [TAS] SMG2: Hightail Falls without Yoshi](https://i.ytimg.com/vi/w8HgazzIMb8/mqdefault.jpg)
![[TAS] SMG2: Throwback Green Star 3 in 16.18s
Luigi cloud jumps to a Green Star in 16.18 seconds [IGT, 971 frames].
This Blast Away The Wall route was made by Pottoww, using the slope jump spin bug by jumping off the very edge of the slope: https://youtu.be/MzC6UY2QRL0
The long jump walljump at the start took the same amount of time as the triple jump insta-climb.
My original TAS route (a 16.48) was to backflip and walljump off the thin corner of the wall, hoping that the subsequent angle would lead to a slope jump that went backwards up towards the Cloud flower. That didnt happen, slope jumps only ever went sideways or forwards.
But in the process, I found that it is possible to land on a sloped lip at the edge of the main slope that causes a very speedy jump.
So I went with a backflip from further right to set up an earlier and higher slope landing to make use of this speed.
The backflip was skewed to the right to have room to drift to the left during the spin. Drifting sideways before walljumping affects the angle of the jump.
Neutral walljumping for a stronger backwards strain to be closer to the slope.
For the cloud jumping, I struggled to get fast cloud jump speed when jumping straight from the wooden platform, so the fastest route I found was to long jump out to the row of clouds and then get a fast cloud jump from there, followed by a mid-air cloud jump.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun [TAS] SMG2: Throwback Green Star 3 in 16.18s](https://i.ytimg.com/vi/xOL5sW1U2sw/mqdefault.jpg)
![[TAS] SMG2: Starshine Beach Green Stars
Green Stars TAS WIP
Star order doesnt really matter here - the Lakitu for GS1 is the same with or without cycles.
I showed the movement on the world map just because its one of the faster cursor-click movements (11 frames over joystick).
Green Star 2 - 23.65s IGT - 1:23:09.11 real-time split (1:20.00 segment)
While carrying the land cancel out to the water enters water faster, its facing the wrong way so the movement in water is much slower than for long jump ground pound.
Long jump on the slope works by landing on the very edge.
While the trajectory of the long jump looks off, this was the perfect setup for a zip jump that went off a small part of slope angled towards the Yoshi egg at 60+spd.
This zip jump (plus conserving speed for the single jump by landing in neutral) saved 3 seconds over long jumping across the gap.
With backwards Yoshi flutters, you get higher acceleration but your speed is fixed once you reach a certain height above ground.
By keeping low to ground, acceleration is even higher and can get above the usual cap of 32spd up to 37.
Eventually it decelerates back down to 32. There is also a short burst of speed after dismounting.
Green Star 1 - 24.51s IGT - 1:24:16.96 real-time split (1:07.85 segment)
After cutscene skips, Luigi is spawned slightly down-left of centre which is why the start goes in a different direction.
I tested three routes for this star: #1: Same as GS2, including varied flutter routes trying to bounce off the Pianta and instant triple bounce off the fruit. #2: jda7s Lakitu bounce ( https://youtu.be/2uiiaGeGeWY ). #3: An alternative Lakitu bounce or straight flutter going under the tunnel.
The results were: #1 24.80, #3 24.61, #2 24.51.
Green Star 3 - 43.43s IGT - 1:25:55.58 real-time split (1:38.62 segment)
This is an improvement over my 2021 TAS demo: https://youtu.be/hg8Tyo10qE8
UPDATE: This has been improved, a 42.28s just about managing to get a boosted leaf jump to work instead of waiting for another cloud. https://youtu.be/1fOPdsSK5ic
A Blooper launch is possible (seen here: https://youtu.be/PTWmPQlxNIw) but it takes too long to lure the Blooper to the surface.
Other launching ideas couldnt get the balance of horizontal and vertical speed necessary for where they are.
Zip jump on the outskirts of the starting area was 0.6s faster.
It is possible to zip jump on the grassy slope above the tunnel, but the left side can only be done high up which takes a while to get up, and the right side only gets jumps going way off to the left. Ultimately the best was 10 frames slower than long jumping.
Its possible to shore launch up the tower, but it takes too long to set up, making it 1.4s slower.
Cloud cycles are manipulated by keeping them off-screen.
The walljump spin was able to snap directly onto the cloud, allowing for a backflip which saves 0.5s.
The issue with the cloud cycles is where the other cloud is when starting the cloud long jumps.
You get much more speed (34spd compared to 20) if you jump from a moving object over a void or very high above ground, so the ideal is to have no clouds under Luigi when starting the long jumps, which is difficult with the fastest cloud cycle for climbing the tower.
I managed this by slightly diverting where the first long jump starts but still having a trajectory towards the star, which edged around the cloud below.
#supermariogalaxy2 #toolassistedspeedrun [TAS] SMG2: Starshine Beach Green Stars](https://i.ytimg.com/vi/xmnvjrK1zgg/mqdefault.jpg)
![[TAS] SMG2: Sweet Mystery Green Star 1 in 7.31s
Mario betrays Yoshi to collect a Green Star in 7.31 seconds [IGT, 439 frames].
Mario can break Yoshis egg 3 frames faster than Luigi, thanks to the formers better acceleration.
I tried optimising the usual strat of fluttering sideways to the star, but then remembered the fast backwards flutter tech Nemi found for Grandmaster ILs: spinning close to ground before starting the flutter to somehow cause backwards flutters to have a higher top speed.
As this star is only a short distance away, the faster acceleration of this tech outweighed the slightly faster top speed of sideways fluttering (backwards fluttering reaches its top speed of 28 units per frame just after the first re-flutter, whereas sideways fluttering needs a couple more re-flutters to reach its top speed that moves between 32 and 34).
There may be a faster top speed to gain with this tech - I had previously jumped and waited until falling to spin, rather than spinning instantly, and this raised top speed to around 29.5. Something about how far inland you are when fluttering seems to change the top speed, which is something I had started noticing with The Perfect Run.
For the first re-flutter, I had to do it early as otherwise there would have been a huge downshift in speed until I eventually did. After this, I optimised flutters for height so I could dismount as early as possible, as dismounting has more speed. But I also found that delaying the inputs between cancelling one flutter and starting the next seemed to help a little in overall speed.
More testing should be done to find the exact right balance between height and speed.
While Mario is 3 frames faster at breaking the egg/mounting, he has a tiny speed disadvantage when dismounting, as - for a brief moment - his speed raises to 34.3 before coming back down to 32 whereas Luigi reaches 35.1. However, this is only for a few frames and they both return to 32, so it has no significant effect, 1 frame of difference at most.
#SuperMarioGalaxy2 #ToolAssistedSpeedrun [TAS] SMG2: Sweet Mystery Green Star 1 in 7.31s](https://i.ytimg.com/vi/xrkFOBb8Qe4/mqdefault.jpg)
![[TAS] SMG1: Gusty Garden Secret in 46.95s
Rainbow Mario destroys the Golden Chomp in 46.95 seconds (2817 frames).
In an effort to give every star in Super Mario Galaxy 1 a TAS, a bunch of TASers got together, spun a big wheel with the level names, and made some LOTADs (low-optimisation tool-assisted demonstration. Bars.). I bagged Gusty Garden Secret, spending only 3 and a half hours on it.
Check out the other TASes:
- Toogs: Gusty Garden 3 (https://youtu.be/HZbwMH7MKgY)
- BillyWAR: Melty Molten S (https://youtu.be/-lpnaCX_Rpw)
- Xander: Space Junk S (https://youtu.be/q5qmgP59UiE)
- Shadow64: Sand Spiral (https://youtu.be/VDaQDqlj46I)
And the TAS server: https://discord.gg/h2YSCZm
I missed out on the big VC, but the next day I had Billy and then SaladBowl for support while they researched bonkwarps and saved A presses respectively. Fan favourite Toogs made a cameo appearance.
With Toogs having done Gusty 3 with cutscenes, I skipped them as this star is in the same level, emulating a full-game run.
Buffered crouch accel (holding Z before you gain control and letting go soon after) allowed for a 1f earlier long jump.
I made sure to stay in the gust tunnel as much as possible while still grabbing the ? coins, using Toogs Gusty 3 and Tibs IL record (https://youtu.be/cRsxCKsn1qc) as blueprints.
This could still be optimised, though, as I only really had a few glances at how much actual speed I was keeping.
My initial strat with the last two coins and Piranha Plant was to hop off the floaty fluff facing forward so that I entered the planet facing backwards, using that momentum to get the first coin and swing straight around either to the Piranha or the other coin.
The issue with that was I had to wait to hop off to be far enough along to swing right around.
My best effort with this was still 1s slower than the final strat.
The final strat landed forwards, landing much earlier, and gathered a lot of speed before landing so I could single jump with all that speed (about 23hspd, above the top speed of a long jump) and get all the way around the planet to get near the Piranha.
The Piranha kill could be improved by a few frames with a slightly earlier spin or a better jump line, however, the real bottleneck is vine RNG.
This RNG can be manipulated with different timed jumps and spins, so it could end up the same whether you kill the Piranha earlier or later. It could also end up much much faster. More playing around needed here.
The last planet was disappointing. I had hoped for some gravity shenanigans, maybe some momentum conservation to have super fast single jumps and ground 2P jumps, but long jumps won in the end by a handful of frames.
#supermariogalaxy #toolassistedspeedrun [TAS] SMG1: Gusty Garden Secret in 46.95s](https://i.ytimg.com/vi/yt_IGkf4QRI/mqdefault.jpg)
![[TAS] SMG2: Upside Dizzy Green Stars
Green Stars TAS WIP
Star order matters due to the cycles of the moving platforms and the starting Paragoombas. GS1 first is at least 1 second faster.
Green Star 1 - 13.18 IGT - 1:41:39.88 real-time split (1:08.98 segment)
See here for a 12.96 IL TAS with better cycles: https://youtu.be/rw64kFfrQNg
Bonking and going through a gravity flip causes Luigi to twirl around, which leads to a backwards wallkick.
The height of the bonk affects the facing angle by the time you need to wallkick, which slightly affects speed.
A starbit shot was used to bring the Paragoomba further left and bounce off it earlier. This saved 0.25s (15 frames) over 2P freezing despite the 3f hitlag.
Ground pounding at the end saved 0.32s (19f).
Green Star 2 - 16.38 IGT - 1:42:38.95 (59.07 segment)
Neutral wallkick strain (neutral on joystick after jumping to bypass the speed cap) and a slightly different bonk to change wallkick angle helped to bounce off the higher Paragoomba.
I initially had trouble with backwards falling, as it doesnt regain momentum after getting slowed down by gravity flipping.
Also, gravity flip vertical momentum is affected by holding A as well as how you move into the changing gravity (e.g. from the long red chute, the long jump bounce in GS1 that goes through the blue gravity flips higher than double jumping from red platform despite the latter falling for longer).
Mizu first suggested spinning once at the top. Spinning in the red gravity just before the very top kept enough speed and height through the final gravity flip to make it to the final platform.
The setup for backwards falling is scuffed, involving a frame-perfect jump from getting thrown off a ledge-grab, and getting stopped by leaving the red moving gravity through the side of it to then spin up.
But overall it saves 1.2 seconds.
#supermariogalaxy2 #toolassistedspeedrun [TAS] SMG2: Upside Dizzy Green Stars](https://i.ytimg.com/vi/zBRstp3u2Lo/mqdefault.jpg)
![[TAS] FAST Racing League: Breeze City in 14.56s
A few years ago, I decided to stop working on a TAS of FAST Racing League, after finishing the first 4 of 12 tracks: https://youtu.be/ItOBCvu-u0Y
But I really wanted to at least do this track as well.
BONK JUMPPAD
One of the reasons I stopped the TAS was that I stumbled upon a trick in Turbine Canyon to make the jumppad shorter, which would save quite a bit of time there, but I couldnt replicate it and so skipped over it but later didnt want a TAS that couldnt have that.
Here I found that the same sort of mechanics can work the opposite effect too, getting more height which was helpful to be able to turn in the air without landing too short of the checkpoint, and gained a bit more understanding.
It seems like you either need a certain angle against the wall while jumping, or need to jump a certain time after bonking, or both.
Angle seems more likely, because slightly adjusting control stick angle would change the height a little bit. But it might need the effect of a bonk too.
LAP SKIP
This game is a bit like Mario Kart Wii where there are checkpoint areas, mostly used for respawn points. But unlike MKW, you arent limited to the one before and after.
The checkpoint area for the line actually ends some way in front of it (the bottom of the closest yellow diagonal, so 2-3 cars long), so if youre able to jump and land behind this invisible line, youre counted as also being behind the finish line from a further checkpoint, so you can finish a lap.
Some exceptions though: if you land before the actual line and cross back over it, you cant be facing forwards, even if youre moving backwards. This is why I landed sideways every time, facing ever so slightly backwards but turned around enough to start moving forward immediately after crossing the line.
You also cant be braking or drifting on the ground during this skip for some reason, which is why I stop drift turning as soon as I land and just go neutral.
BOOST
Theoretically the boost would be better to use going towards the jumppad for a skip, as it only saves a few frames right at the end. But I havent managed to get the higher jumppad trick while boosting into the wall. Kinda harder to know your angle or time the bonk right - whichever one actually matters.
#toolassistedspeedrun [TAS] FAST Racing League: Breeze City in 14.56s](https://i.ytimg.com/vi/zFp3KJOVlVY/mqdefault.jpg)
![[TAS] SMG: Good Egg 2 in 44.38s (Mario)
Mario lets a Hungry Luma starve in 44.38 seconds (2663 frames).
The triple jump bounce at the start was found by nemi43: https://youtu.be/wM5Zpot5h7g?feature=shared
While it might be possible to float directly from the bounce, the angle is all wrong for the float needed to get the skip.
The skip had also been theorised by nemi, and, with some help from Dendonodon and BillyWAR, I was able to get it years after the theory: https://youtu.be/sk68AKjWnyg
But the specific angle and movement used to get the fastest possible skip (i.e. moving far enough to ground pound as early as possible into the final planets gravity) was figured out by the RTA community (sorry I cant nail down the names; Marioandtrafficlightfan, Subro, Cubic3107, saburoumaru and Nanase all have videos on the same day with the same technique).
#supermariogalaxy #toolassistedspeedrun [TAS] SMG: Good Egg 2 in 44.38s (Mario)](https://i.ytimg.com/vi/zRh6gq-d0e0/mqdefault.jpg)
![[TAS] SMG2: Slipsand Green Stars
Green Stars TAS WIP
Star order matters due to the cycles of the moving blocks on the 2nd planet, making an affect on GS2 & 3 but not 1.
Ive potentially lost some time here; my 2020 tests found the blocks to be slightly better with cutscenes so GS3 goes first, but during this TAS I improved that section to then mean the no-cutscene cycles ended up much better than before, but I didnt want to redo everything up to that point so Ive left it as future improvement.
Green Star 3 - 1:04.65 IGT - 1:44:39.70 real-time split (2:00.75 segment)
An improvement of 4.31s (+ cutscenes) over my 2020 TAS: https://youtu.be/xsd_7sIqS90
Its actually possible to long jump all the way over the first stone ring, but the route that follows is 0.2s slower.
Like with star order, theres a little bit of future improvement in the last segment of sandcurrents that was found in the latter stars.
And another case in the 2nd planet, where later stars get a very precise edge long jump after going over the big curve to avoid losing speed from the sand. I really did not have patience with this galaxy.
The Romp launch works because the ends of the Romp dont immediately crush you but instead try to push you to the side (see nemis IL for Romp pushing strats: https://youtu.be/Wy2RYwIa-Ck?t=185 ).
This means you can jump in front of the Romp and clip inside enough to cause a big push in 1 frame, and if you can get a slope jump during that, the clip speed is conserved and fixed to the whole jump.
I had to go to the second row of Romps because there was no good angle jumping from the first row that would go straight to the star.
Green Star 2 - 49.35 IGT - 1:46:12.38 real-time split (1:32.68 segment)
2020 tests had a 51.20 here.
The improved moving block cycle is obvious here; in the previous tests, the blocks started moving left before long jumping.
Getting the star, a straight beeline to the star was 1f slower than a little bit of extra drifting to the left.
Also, the starbit collection with IR was necessary. If you collect a starbit and youre still in collecting state (glowing white) when grabbing a star, that starbit will be duplicated. Picking it up on the way, I would grab the star just barely within this window, so instead I use IR to pick it up earlier and shoot it out, avoiding the starbit textbox after the level.
Green Star 1 - 37.43 IGT - 1:47:33.15 real-time split (1:20.77 segment)
2020 tests had a 38.48 here.
Much the same as GS2. For the star grab, I was able to get another edge long jump avoiding the sandy speed loss, saving a couple of frames by getting enough speed to grab the star on top of its hitbox rather than the side.
#supermariogalaxy2 #toolassistedspeedrun [TAS] SMG2: Slipsand Green Stars](https://i.ytimg.com/vi/zhA03CAA7UU/mqdefault.jpg)