TAS (tool-assisted) demos: part 2

@alps, perfection takes a lot of time. There are a lot of other things to do. I can't spend all my free time on building "perfect" Doom TASes. No-one will thank me for winning every last second, no-one will donate me for that. I just want to make entertaining movies. What I do is more than enough for most people who watch my TASes. I know how to use XDRE. When I see it's needed to be used, I use it. Everything else is my personal business!

3 hours ago, alps said:

you know though, you really need to switch to full building, if you want to make optimal demos, with slowmotion its simply imperfect.

3 hours ago, Dimon12321 said:

 

It doesn't matter. I'm not trying to win every last second.

Dimons demos are among my favourites, and it seems to me like he's doing what he enjoys and is doing a great job at it. Investing significantly more time by building in order to save what ends up often being a small amount of time isn't everyones priority. Not to mention that building a max demo is significantly more nuanced, difficult and frustrating than simply getting good player movement, in a NoMo demo for example. 

Edit: Whoops, I didn't see that Dimon already replied. 

1 hour ago, almostmatt1 said:

 

Dimons demos are among my favourites, and it seems to me like he's doing what he enjoys and is doing a great job at it. Investing significantly more time by building in order to save what ends up often being a small amount of time isn't everyones priority. Not to mention that building a max demo is significantly more nuanced, difficult and frustrating than simply getting good player movement, in a NoMo demo for example. 

the only thing i meant was potentially using xdre in certain areas to get doors, switches on earliest possible tic etc. for example you can get 0.40 off a prboom demo by adjusting the doors. i didn't mean 10 hour perfection, just simplistic type.

scythe map03 intended route nomo in 12.89

intsc03xo1289.zip

Doom E2M3 (Refinery) in 20.46 (NoMo)

e2m3xo2046.zip

On 11/15/2020 at 4:27 PM, alps said:

Doom E2M3 (Refinery) in 20.46 (NoMo)

e2m3xo2046.zip

Built runs usually look like that, so don't worry about the "eyes bleeding" part, I think. Some people don't care what the run looks like as long as it is fast :-D

22 hours ago, Looper said:

Built runs usually look like that, so don't worry about the "eyes bleeding" part, I think. Some people don't care what the run looks like as long as it is fast :-D

lol yeah, i think the future of TASes sits at those sharp turns, when I try to smoothen out things it just becomes slower, when going for tic perfect demos, smoothing it out isn't very viable. and smooth demo playback already takes care of that problem. speaking of fast demos however...

e3m1 nomo 13.97

been experimenting with this map since i finished the shitty mess that is d1e3-236xo, today i got lucky, everything went according to plan, and got the 13.

e3m1xo1397.zip

rb01xo849.zip

rebirth map01 nomo 8.49

Doom E1M8 NoMo in 21.97

e1m8xo2197.zip

its good, could redirect off of those corners near platform (with wobble effect) and use door boosts, easily 21.7x or 21.6x, too much effort however.

and its still 21 so nobody would really care about it.

doom episode 3 run (Work in progress)

e3m1-e3m4

in 1:08

Don't upload to DSDA

e3owip2.zip

a second off in e3m1, e3m3 and e3m4

10 sec e3m2 in progress

3 seconds saved from d1e3-236xo (garbage) so far

EDIT: e3m4 is now built, updated the wip

I've noticed something weird and can't find a cause for it. There are a pair of XDRE commands, transpose left and transpose right (turns you by 90 degrees (64 shorttics) and automatically adjusts your MF/MB & SR/SL values to produce equivalent movement), that are fairly handy for making movement look a bit cleaner, however they occasionally don't properly work - specifically, when you transpose left/right between byte angle 101, 165, 229 & 37 (longtic angles 25856, 42240, 58624 & 9472), your position and momentum is ever-so-slightly changed. This effect appears to be unique to these specific angles. A few times recently, a tas I'm working on desynced after transposing on these angles and it took me a while to work out what was going wrong. Does anyone know why this happens? I've been using XDRE 2.20 if that makes any difference.

5 minutes ago, almostmatt1 said:

however they occasionally don't properly work - specifically, when you transpose left/right between byte angle 101, 165, 229 & 37

That's weird. These angles are close to ones needed to perform a wallrunning trick.

I never used it because I try to avoid sharp turns.

its actually weird, turns out shorttics work in vanilla complevels or others that you're normally not meant to have shorttics in demos. and then going back to longtics produces strange effects. its kinda weird how it happens, i dont think i can exactly explain it but after reloading the demo it desyncs. transposing feature may have produced that same effect for you.

Yeah, I think vita has pointed this effect out before, and yeah, it's only between specific angles. I think it's probably some rounding issue in the source code, but I'm not sure about the details

Doom E2M9 Fortress of Mystery No Monsters in 5.97

e2m9xo597.zip

Scythe MAP01 Get Going No Monsters in 7.77

sc01xo777.zip

behold

i present to you

my greatest demo

e1m1 hangar nomo 7.97

uv-speed probably coming soon (tmr)

e1m1xo797.zip

and also

e4m9 uv-speed 6.94

e4m9x694.zip

doom 2 map01

uv-speed in 4.77: lv01x477.zip

no monsters in 4.49: lv01xo449.zip

plutonia 2 map15 where hate runs red

uv-pacifist + reality in 5.74

p215x574.zip

kama sutra map09 the holocaust nomo in 10.97

11 in uv coming soon

ks09xo1097.zip

scythe map08 uv-speed in 7.20

sc08x720.zip

Doom Episode 1 UV-Speed in 3:25 - ep1-x325.zip

For the first 80 tics of this demo, the inputs are taken (though slightly, but not significantly, altered) from e1m1xo791.lmp by 38_ViTa_38. I sent him a message a few days ago asking if he was okay with me posting the demo but haven't had a reply, and it looks like he's not been on Doomworld for about a month and a half, so I hope it's okay that I still post it. 

9 hours ago, almostmatt1 said:

Doom Episode 1 UV-Speed in 3:25 - ep1-x325.zip

 

 

For the first 80 tics of this demo, the inputs are taken (though slightly, but not significantly, altered) from e1m1xo791.lmp by 38_ViTa_38. I sent him a message a few days ago asking if he was okay with me posting the demo but haven't had a reply, and it looks like he's not been on Doomworld for about a month and a half, so I hope it's okay that I still post it. 

very good, missed tricks in e1m3, e1m5 are the only small errors an eye can see

there is also an error in e1m7 i believe where the lift didn't raise as soon as possible (might be wrong) the linedef is on the left there and not in front (again i might be wrong but it didnt trigger for me up front)

oh and the tics in e1m1 are transposed, (at start before wt)

16 hours ago, alps said:

very good, missed tricks in e1m3, e1m5 are the only small errors an eye can see

there is also an error in e1m7 i believe where the lift didn't raise as soon as possible (might be wrong) the linedef is on the left there and not in front (again i might be wrong but it didnt trigger for me up front)

Thanks! Yeah, the door stuff in e1m3 & e1m5 is a shame to have missed. 

I just had another look at e1m7. The western linedef to raise the lift, linedef 906, was apparently triggered on tic 6207 but nothing happened. If I had have not slowed down too early here, I'd have crossed it on tic 6206. If I changed my angle to activate the northwestern most linedef to activate the lift instead, I could have done so on tic 6205. But yeah, weirdly even though xdre says I did activate linedef 906, the lift doesn't raise until I activate the southern linedef to raise the lift, which occurs on tic 6217, a full 0.28 seconds after crossing linedef 906. More egregiously, I just now noticed I bumped the wall on tic 6191 and lost half my speed, I don't know how I didn't notice this. If I correct this to not bump, and then cross the closest linedef as soon as possible, the lift raises on tic 6200, a full 0.51 seconds sooner than it did in the posted demo. Considering the exit time was 53.60, with better movement at the elevator I'd only need to save 4 tics elsewhere to get 52 seconds instead, which is certainly possible. 

The conclusion: :(. Might have to rebuild. 

@almostmatt1 Great work! It's not as watchable as Andrey Budko's TAS, but it's still a great improvement

plutonia map01 uv-speed 6.11

5 is very difficult

pl01x611.zip

15 hours ago, almostmatt1 said:

Doom Episode 1 UV-Speed in 3:25 - ep1-x325.zip

In e1m8, did you try using the nomo glide + void glide out of the courtyard? The wait for the pinky void glide is actually quite long, maybe it's slower.

1 hour ago, kraflab said:

In e1m8, did you try using the nomo glide + void glide out of the courtyard? The wait for the pinky void glide is actually quite long, maybe it's slower.

you have to wait for lift to raise outside, dimon did 21 sec nomo so it is slower to use that glide.

@alps, kraflab probably refers to this route. I'm also wondering if it's possible to perform the second glide using a boost from a single rocket blast.

I didn't try this, no, but it does seem as though it would be faster. Has that void glide from the courtyard been confirmed to be possible by anyone? I'll give it a try. 

Edit: Thanks Dimon!