1 2
8 9 10
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
FractalFusion wrote:
Instead of bit.band, maybe use math.floor(iRNG/0x10000)%0x100. This gets the third byte as an actual byte, and not as an 8-bit number times 0x10000. Then you write c=0x84 instead of c=0x840000.
Thanks for that. :) My problem with bitwise operators is that I require a separate file for my Lua to use them. Lua 5.1 doesn’t support bitwise operators, and I don’t know how to upgrade to a newer version.
Joined: 6/27/2008
Posts: 49
ALAKTORN wrote:
FractalFusion wrote:
Instead of bit.band, maybe use math.floor(iRNG/0x10000)%0x100. This gets the third byte as an actual byte, and not as an 8-bit number times 0x10000. Then you write c=0x84 instead of c=0x840000.
Thanks for that. :) My problem with bitwise operators is that I require a separate file for my Lua to use them. Lua 5.1 doesn’t support bitwise operators, and I don’t know how to upgrade to a newer version.
I've tried to upgrade it by recompiling DeSmuME with newer lua (5.3) headers & libs before, but got some compile error. It seems that lua engine coders are using some functions that specified to lua 5.1 which prevents an easy migration. EDIT: Tried to compile DeSmuME with luajit which is said to be fully upwards-compatible with Lua 5.1, but also failed with #include "lstate.h"
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Testing more Aspis kills. Previous tests showed that late spit start is best. Late spit + late dive = terrible, too much time offscreen, you lose 10 frames of charge (which would still be faster than my vid so far) but by doing so you prolong Aspis’s invincibility a lot, losing further frames on the last hit, and he can’t be killed on the ground. Late spit + early dive = bad, you can get a perfect fight (saving 11f over my vid) but you don’t get the grounded kill, and lose too much time (would also lose time from healing 3HP). Considering a spit after another spit has always the same timing (as far as I can tell), the 11f lost vid I made is a perfect fight. Late spit + spit + early dive is the best possible combination, considering the 2nd spit’s timing can’t be made any quicker. Now I need to actually TAS the fight. Hmm… HHS, do you know anything about how Giga Aspis (the snake boss) determines what attack to do? It actually looks really simple. There’s only 1 step in its decision (2 total). At least for the first attack. If it decides to do “nothing”, it will do 3 steps to determine what to do as its late attack. Edit: Just had 3 steps and “nothing”. Not quite sure what makes it 2/3 steps. Edit: HHS has said before that only bits 16–30 are used in general. Well I’m already stumped. Tried checking out 2 different RNG values that give the same behaviour (late spit start Edit: so this means they give “nothing”): values: 1083742552 1287626704 steps (unfrozen): 1083742552 1819325074 (2) 1287626704 1250995402 (2) conclusion: 1819325074 110110001110000 0x6C70 1250995402 100101010010000 0x4A90 Not very many similarities. Won’t be as simple as I thought? Edit: Well I guess it would make sense for it to be a bigger range rather than a single value like a Life Up drop. Edit: Testing: 888243357 1209073663 (2) 100100000010000 0x4810 gives “nothing” 1609663645 1259405311 (2) 100101100010000 0x4B10 another “nothing” 157385885 1073807359 (2) 100000000000000 0x4000 again 1663599773 1073741823 (2) 011111111111111 0x3FFF turns around and goes offscreen :0 1348568221 1057030143 (2) 011111100000000 0x3F00 nothing again… Welllllllll I doubt I’ll be able to figure out anything manually.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Anty-Lemon wrote:
HHS wrote:
As for how I figured it out, I have an emulator with data breakpoints, but it hasn't been released to the public yet...
no$gba debugger has breakpoints too
Tried looking into this but on the website it says no$GBA doesn’t support savestates for DS games? How is a debugger even helpful in that case? lol (Also is the debugger built into the normal no$GBA version?)
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
If someone could do the disassembly and figure out how to manipulate the boss, that’d be fantastic. For now, I’ll try manipulating it myself with random attempts, streaming here: http://www.twitch.tv/alaktorn91 Edit: I feel like I’m figuring out some stuff. Anyway, I noticed that I was getting a lag frame right at the start of the boss fight, it happens when you try to move too quickly. Avoiding it doesn’t really save time (because you have to stand still for a frame), but it ties, and it can be used to manipulate the RNG, since during the lag frame it doesn’t advance, but if you manipulate the lag away it’ll advance and it won’t cost any frames, so it’ll be 1 advance ahead without wasting time. Edit: Maybe I could even avoid the lag and save a frame if I got into the battle without a full charge (more animation to process, which may create the lag). Trying every possible amount of +steps (late) and seeing what kinda behaviour I get (this is with the lag frame at the start): frame 16884: 0 back, 2 back, 3 dive, 4 dive, 5 back, 6 spit, 7 spit, 9 dive, 10 back, 12 dive, 13 spit, 14 back, 15 dive, 18 spit, 19 back, 20 spit, 23 spit, 24 spit, 25 dive, 26 spit, 27 back, 28 spit, 29 dive, And, advacing those amounts of steps early gives… frame 16824: 0 nothing, 1 dive, 2 back, 3 back, 4 back, 5 dive, 6 spit, 7 spit, 8 nothing, 9 nothing (+3, dive), 10 dive, 11 nothing (+2 back), 12 dive, 13 spit, 14 dive, 15 back, 16 spit, 17 nothing (+4 spit), 18 dive, 19 dive, 20 dive, 21 back, 22 spit, 23 dive, 24 dive, 25 dive, 26 nothing only advances 2 late so is −1 and gives dive instead of spit… 27 dive, 28 nothing +2 late so −1 so goes back… 29 nothing +2 late so +28 total = spit :) This works? 30 back, 31 spit, 32 dive, 33 spit, 34 dive, 35 dive, 36 nothing (+2 dive), 37 nothing (+2 back), 38 spit, 39 dive, 40 spit, 41 spit, 42 back, 43 spit, 44 nothing (+2 spit) I see a pattern??? I don’t get this. Edit: Pattern broken with 3 I guess… /// Yeah I guess it was just a coincidence… from 17, 16901: Edit: This number looks wrong (early). 0 back, 1 N/A, 2 nothing (+2 back), 3 back, 4 nothing (+2 dive), 5 nothing (+2 back), 6 dive, 7 dive, 8 dive, 9 back, 10 back, 11 back, 12 dive, 13 nothing (+2 dive), 14 back, 15 back, 16 dive, 17 back from 44, 17130: 0 nothing (+2 back), 1 N/A, 2 back, 3 back, 4 N/A, 5 nothing (+3 back), 6 dive, 7 dive, 8 dive, 9 nothing (+2 spit) Yeah that one worked. ^_^ I’m waiting 29 frames to manipulate the pattern atm, but I’m pretty sure I can achieve the same result by shooting walls and advancing the RNG 29 steps before the boss fight without losing any frames. Now to see if I can manipulate the last 2 attacks… Oh boy next behaviour is also an early–late. Wonder why I could never get a different spit timing. Edit: Seems like early spit is impossible, weird. (this with the 29 frames lost at the start) frame 17115: 0 back, 1 back, 2 dive, 3 nothing, 4 back, 5 back, 6 dive, 7 back, 8 back, 9 back, 10 dive, 11 dive, 12 back, 13 nothing, 14 back, 15 nothing, 17 back, 18 back, 19 back, 20 nothing, frame 17175: 3 +2 dive, 13 +2 back, 15 +2 back, 20 +2 back Holy crap I need to get 2 nothing+spit one after the other with the only manipulation possible being before the first nothing (before boss) and before the second nothing, but very limited (during boss shooting its body). 2nd manipulation seems to only be 1–9 steps possible. I think 8 steps cost 6 frames of time. Edit: Weird, the pink stuff of the big charge advances the RNG 2 steps normally, but I’m getting it to not advance the RNG at all in my TAS… Maybe it’s the way it’s hitting the boss… at an angle, between 2 of the body-spheres.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
I don’t even understand why, but it seems like after much trial & error I was able to get the Giga Aspis manipulation without losing any frames. I’m like 99% sure my manipulation took too long and I should be losing frames, but the lag + position update after the kill is on frame 18020, which considering the 44 wait frames I added before the boss fight, ties with my perfect (cheated) boss fight from before. It’s 5am here so I’ll look into it more later I guess. Gif courtesy of Zinfogel: http://i.imgur.com/uuQyM7R.gif But there’s a problem: I don’t think I can advance the 44 steps before the fight without losing any frames. That’d be really sad… either way, I’m far from done… Info on my old perfect (cheated) kill:
ALAKTORN wrote:
https://www.youtube.com/watch?v=qRV4I6D6PAM My fight gets a late spit at the start, so that I can do full attack combo loops without wasting time up until the second to last hit (3 frames lost to hit), Edit2: 8 frames lost on the last hit in my on-screen kill. So 3+8 = 11 frames away from perfect? 17406, 17576–17976 Boss 0 HP, explosion sprite appears – lag frame / position updated after kill
ALAKTORN wrote:
Trying every possible amount of +steps and seeing what kinda behaviour I get (this is with the lag frame at the start): frame 16824: 0 nothing, 1 dive, 2 back, 3 back, 4 back, 5 dive, 6 spit, 7 spit, 8 nothing, 9 nothing (+3, dive), 10 dive, 11 nothing (+2 back), 12 dive, 13 spit, 14 dive, 15 back, 16 spit, 17 nothing (+4 spit), 18 dive, 19 dive, 20 dive, 21 back, 22 spit, 23 dive, 24 dive, 25 dive, 26 nothing, 27 dive, 28 nothing, 29 nothing, 30 back, 31 spit, 32 dive, 33 spit, 34 dive, 35 dive, 36 nothing (+2 dive), 37 nothing (+2 back), 38 spit, 39 dive, 40 spit, 41 spit, 42 back, 43 spit, 44 nothing (+2 spit) from 44, 17130: 0 nothing (+2 back), 1 N/A, 2 back, 3 back, 4 N/A, 5 nothing (+3 back), 6 dive, 7 dive, 8 dive, 9 nothing (+2 spit)
As written above, I waited 44 frames before the boss fight, did a manipulation of 9 steps during the fight, and super luckily the final attack Giga Aspis did was an early dive, exactly what I need for the kill. I don’t even know if I could’ve manipulated anything at that point…
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Let’s try to get those 44 steps. I’m going to be streaming it: http://www.twitch.tv/alaktorn91
Editor, Experienced player (847)
Joined: 5/2/2015
Posts: 696
Location: France
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
^lol
ALAKTORN wrote:
Let’s try to get those 44 steps. I’m going to be streaming it: http://www.twitch.tv/alaktorn91
Current TAS advances 31 steps. I need to advance a further 44 for my Giga Aspis kill to sync. +6 +3 +1 +1 +3 +1 +1 +10 +11 – The amount of steps I advanced over my old TAS: Kill first enemy, kill its copter, shoot wall ×2, kill copter, shoot wall ×2, shoot wall ×10, kill spinning blade. old RNG 498813948 on frame 16760 new RNG 73299994 on frame 16761 (+ 38 steps compared to old) Lost a frame (may or may not be necessary for that specific RNG manipulation), and only got + 37 steps over the current one, so I’m 7 steps away from syncing the boss fight. I don’t think I can advance the RNG any more than this without losing a few frames, which sucks… This manipulation is so fucking complex I really hate TASing this game. I’ll probably have to re-TAS it trying to advance the RNG as little as possible, and do Giga Aspis manipulation all over again to check if there’s a better RNG earlier than what I currently get. Video of my attempt at RNG manipulation: https://www.youtube.com/watch?v=-l7DC6HLly8 Also input file: http://tasvideos.org/userfiles/info/23603629336034618 Edit: Also I tested not charging a shot at Giga Aspis, but the lag frame still happens if you move first frame. Edit: 599552 X 15484 667648 X 15845 599423 X 15484 667647 X 15845 So to do a +1 step I’m losing a frame due to a single subpixel. WOW. I take part of what I said before back, my old TAS already destroyed the copter and got the +3, it just self-destructed off-screen so I didn’t notice it. I was confused at enemies’ health so I found their HP address and by extension the whole enemy data RAM section. Starts at 0x02154560 and ends at 0x02154F1F. Each enemy is 208 bytes of data, 12 enemies total. I wasted my life on that shit. Never again. Edit: I lied.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
ALAKTORN wrote:
I was confused at enemies’ health so I found their HP address and by extension the whole enemy data RAM section. Starts at 0x02154560 and ends at 0x02154F1F. Each enemy is 208 bytes of data, 12 enemies total.
Correction: it starts much, much earlier. Address 0x02153A00 (26 total slots, 5408 bytes). Its value: 0x00000000. This is a pointer. Also, this isn’t just enemy data, it’s more like an “object” area. NPCs use it (when spoken to?), and things like Giga Aspis’s rocks from his attack use it too. Strangely the biometals don’t use this area during cutscenes, nor do items use it as far as I can tell. +0xD0 offset for the next one. 0x02153AD0 value: 0x02153A00, points to the slot before itself. Image of before and after an enemy spawns: http://i.imgur.com/ohqcNVp.png Before: +0x8 and you get a 0x06. What does this mean? +0x20 and you get a byte that counts up for each slot. Starts at 0x49 (ends at 0x62). What does this mean? Everything else is empty. After: The pointer now points to the next usable area as opposed to the area just before it. At some point it points to 0x0215E804 which is outside the “enemy” RAM area but that just points back at a usable slot within it and finishes the loop. Dunno why it has to go to 0x0215E804. There could be more weird pointers like that. Other 2 bytes untouched. X/Y pos/speed, health and other such information is inserted. If all 26 slots are used up, nothing that uses that area will spawn anymore: https://www.youtube.com/watch?v=hxpuS7HZJsc 0x0215E800 points to slot #26 at game start. Maybe if I looked at what reads that address I could figure out where the inserting of data comes from? Idk what I’m doing.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
My despawning cheat allows me to check for things that affect RNG steps. I figured out the spinny blade enemy. The blade does 9 steps when destroyed, and each chain part does 2 steps. Since the top chain is always visible, you can get at the least 11 steps, and at most 15 steps when all chain parts are visible (meaning the blade is down).
ALAKTORN wrote:
old RNG 498813948 on frame 16760 new RNG 73299994 on frame 16761 (+ 38 steps compared to old)
Tried new route. 16760 73299994 + 38 steps. Improved 1 step and 1 frame. Not nearly enough… Edit: I FREAKING DID IT!!! NEW ROUTE TO ADVANCE THE RNG 44 STEPS AND GIGA ASPIS SYNCS!!!! YESSSS!!!! Now I’m trying to save the lag frame, and in fact it advances the RNG 2 steps more so now my manipulation in-level is too much LOL. I don’t want to change it though, it looks so cool. Edit: Saved the lag frame. Everything syncs. I have achieved a truly perfect Giga Aspis fight. MARVELOUS!! Video to come. Might as well upload the input file in the mean time: http://tasvideos.org/userfiles/info/23719392473820814
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Encode: https://www.youtube.com/watch?v=_IZAPyJFsVE Didn’t really check, but I should be 64 frames faster than the published TAS by this point. 1 frame for each of the walljumps, 62 frames for Giga Aspis. Edit: Actual comparison: Lag 70 – 72 lost 2f, I think that’s due to going from v9.6 to 9.11 72 – 74 73 – 75 85 – 87 1905 – 1907 1917 – 1919 1920 – 1920 saved 2f, possibly again due to emu ver differences 1933 – 1933 1934 – 1934 1935 – 1935 1936 – 1936 1937 – saved (Giga Aspis start) 1938 – 1937 1939 – 1938 110f saved from selecting Vent (with 2f lag lost). 3f saved in A-1 (2 from lag, 1 from walljump). 3f saved going to Giga Aspis by better walljump and full speed dashing to start the cutscene, all 3f are saved during the cutscene walk. 62f saved on Giga Aspis, 61 by better RNG manipulation and 1 by better lag manipulation (at least on this one I was correct…). Good thing I went and compared. No idea how, but I had 2 more lag frames right at the start of input after the intro cutscene (savestate’s fault). Saved 2 frames by cutting them off. Edit: Don’t know why Moth/Rolan didn’t do this, but I just improved the intro by 95 frames. After picking a character, you can press either A or Start to quickly skip the animation. Now the input file should be good. Didn’t bother to check these things earlier because the game is super easily hexable, so I wasn’t really worried. /// Oh man I’m getting ridiculous lag now. 87 – 87 1907 – 1910 lost 3f for some reason 1919 – 1922 1920 – 1925 lost 2f for some reason etc. So I lost 5f in lag (for some reason) but saved 95 by skipping that animation. So a total save of 90f, not bad. /// My save is adding up to 92f. I don’t understand. Maybe the animation skip actually saved me 97 frames… wasn’t very meticulous in calculating that. /// No, it really is saving me 95. Wtf. THIS ***** EMULATOR KEEPS PLAYING AROUND WITH THE ****** SAVE DATA. ****** ****** **** WHY CAN’T IT START WITH CLEAN SRAM EVERY TIME HOLY ****** **** IN HELL ********! “1920 – 1925 lost 2f for some reason” I don’t understand how I ever got the 1920 but it’s not happening again. Probably **** save data bullshit. With 1922 a 92f save sounds right. I think that’s a total of 272 frames saved so far.
Experienced player (643)
Joined: 11/23/2013
Posts: 2230
Location: Guatemala
ALAKTORN wrote:
THIS ***** EMULATOR KEEPS PLAYING AROUND WITH THE ****** SAVE DATA. ****** ****** **** WHY CAN’T IT START WITH CLEAN SRAM EVERY TIME HOLY ****** **** IN HELL ********!
Yeah, it happens to me and ALL of my DS TASes every frikin time. Maybe this is a bug, not so sure, but dang it's annoying. BTW Congrats on your perfect fight anyway.
Here, my YouTube channel: http://www.youtube.com/user/dekutony
Editor, Experienced player (847)
Joined: 5/2/2015
Posts: 696
Location: France
So it is now possible! Major props to you, ALAKTORN!!!! :)
Editor
Joined: 8/6/2014
Posts: 37
Starting SRAM is highly bugged in both 9.10 and 9.11, I recommend you do not use either when TASing.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Thanks guys.
Suuper wrote:
Starting SRAM is highly bugged in both 9.10 and 9.11, I recommend you do not use either when TASing.
I went to 9.11 because it adds a scrolling bar in the Lua window and you can mass-highlight cheatcodes in the Cheats window so it’s easy to delete them in bulk. I’ll try to be careful with SRAM and hope nothing goes wrong, but if I’ll run into SRAM problems again I might just go back to 9.9.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Damn I’m lazy. Zinfogel has made a theory TAS for Normal mode, some pretty cool stuff in it that I think can be implemented in my improvement to the published run. https://www.youtube.com/watch?v=iR63-AspOWc Edit: New link I think? https://www.youtube.com/watch?v=ReH0cu5MMh4
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
While randomly playing Serpent with Hx to figure out RTA strats, I think I came up with a strategy that’s faster than what the published TAS does. Anyone know of any better? https://www.youtube.com/watch?v=L2jRW__1CdE I haven’t compared it though so I’m not sure. But that video is RTA, not TAS.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Protectos re-fight: https://www.youtube.com/watch?v=NyvAOSULvDk Strategy is pretty much the same as the published TAS, but for some reason Moth/Rolan missed the double hit with the orb and ended up doing an extra attack to the boss. My vid is a slight variation for RTA and it shows that the extra attack is unnecessary. Zinfogel also tested going left after defeating Leganchor, and it seems faster: https://www.youtube.com/watch?v=RbvQ_dEnN-g
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Trying to port the run to 0.9.9 because I don’t wanna deal with SRAM bullshit. I have no idea how the lag is matching up but the whole thing synced without editing any input, so I’ll take it. Ended up editing a little input before Giga Aspis in the movement to make it look more impressive (to me anyway). In theory all of this would need to be re-done to add entertainment with better movement / drops manipulation but bleh I’ll think about that later. I think jumping into background doors is 384 subpixels more optimal, could save 1 frame situationally. Lol I spent like 1 hour trying to enter the door on the closest pixel only to find out that no matter what, it’s gonna be 1 subpixel away or 1 frame slower to do so. Fuck this game. Trying to manipulate Life Up drops is painful. 9 calls too many. :( I saved a frame jumping into the second background door. :0 Guess that confirms that I was right. Currently 332 frames ahead up to C-1. Vid: https://www.youtube.com/watch?v=9Q9AE6ewf08
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
TAS’d a bit more on stream and also made a speed Lua. I waste so much time on pointless shit, lol. Tested a bunch if walljumping the car right after the door is faster than just jumping over it. Turns out it’s either the same speed or slower depending on when you do the walljump, since walljumping actually has 1 frame of 0 Y speed I was surprised it could even tie. If it didn’t have that it’d easily be faster. Anyway I TAS’d up to entering the door after talking to the first guardian guy lol. Saved 10 frames total, 9 because of Vent’s first dialogue box in C-1 being only 1 box instead of Aile’s 2, and 1 frame saved from jumping into the door. Both the talk to the first guardian and the jumping into the door are within 2 subpixels of being possible, this game is crazy with all these almost subpixel-perfect shit lol. I’d make a video but it’s so pointlessly short I won’t. I’ll make a vid after I’m done TASing the whole C area I guess, which’ll probably take forever. lol That means I’m 342 frames ahead atm btw. Lua: http://tasvideos.org/userfiles/info/31488030838422103 Edit: Why not movie too: http://tasvideos.org/userfiles/info/31488009992023616
Experienced player (674)
Joined: 2/5/2012
Posts: 1777
Location: Brasil
are you done with the boss fight then?
TAS i'm interested: Megaman series, specially the RPGs! Where is the mmbn1 all chips TAS we deserve? Where is the Command Mission TAS? i'm slowly moving away from TASing fighting games for speed, maybe it's time to start finding some entertainment value in TASing.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
grassini wrote:
are you done with the boss fight then?
What boss fight? Giga Aspis was done a year ago. https://www.youtube.com/watch?v=_IZAPyJFsVE Edit: TAS’d up to entering the third dude’s room. Think I saved 1 frame jumping into the door, but didn’t compare.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
Wow, good thing I went and compared… lost 5 frames for no reason and after checking I noticed I had a dialogue mashing as A+B instead of A,B. How the fuck did that happen. =_= Anyway, crouching room. Saved I think 2 frames before the crouching (again by jumping into position rather than walking to it). Aile crouches at 192 speed compared to Vent’s 128 so I start losing time here. Only 314f ahead talking to the guy (29f loss). I save 1 frame going to the next crouch because Moth/Rolan started moving 1f later than possible, lol. 284f ahead after the second crouch (31f loss). And that’s where I’m at now. I have 4 more lag frames despite having saved 1 lag frame at Giga Aspis, will have to look into that discrepancy eventually. Movie: http://tasvideos.org/userfiles/info/31747414173764441 There are 2 enemies in the next room. I’m thinking if I can manipulate a Life Up I’ll kill them, otherwise I’ll ignore them.
ALAKTORN
He/Him
Former player
Joined: 10/19/2009
Posts: 2527
Location: Italy
TAS’d a bit more on stream. Seems impossible to manipulate the LU drops from the enemies so I went pacifist. Saved a frame jumping into the door, saved 1 frame from transforming in mid air (so I don’t have to stop my movement) which also allows me to transform 1 frame later (saving 1f) because I can open the transform menu on the same frame I enter the NPC’s run-away trigger zone which nullifies the trigger. Done with Area C-1, 287f ahead of published TAS. TASing Vent is pretty sad, if I were using Aile I could be saving more frames thanks to her crouching to manipulate my position when talking to the NPC. I kinda hope at the end of the run Vent turns out to have longer cutscenes so I can just resync everything with Aile and be faster… Vent’s knockback speed is 256 while Aile’s is 512 (same as walking). Gonna look into softresetting to skip cutscenes… /// Nope, game is smart enough to only let you skip cutscenes you’ve already watched in full, it seems. Edit: Comparing lag from published TAS to mine (emu v0.9.9 Bus-Level Timing off, dynamic recompiler off) 33 v 34 +1 right on the first screen 70 v 73 +2 (3) intro cutscene 1905 v 1911 +3 (6) after anime cutscene 1937 v 1942 −1 (5) start of Giga Aspis fight 1941 v 1945 −1 (4) right after the dialogue after Giga Aspis, I guess I saved a lag frame here? I think it has stayed the same (+4) throughout the rest of the run.
1 2
8 9 10