TASVideos

Posted: 8/28/2014 8:36 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

WIP6: Link to video Now is the time to think about the sketch glitch route before continuing the TAS, now that we get the espers.

Posted: 8/25/2014 11:47 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Mr. Kelly R. Flewin wrote:

Mind you, it'd be awesome if there would be subtitles to explain some of the stuff you're doing.

Yeah, I will do that. Sabin scenario is finished. 399s ahead of Erokky's TAS. Encode is following. I ran into Templars and Soldiers on the Veldt :( Edit: WIP5 Link to video

Posted: 8/22/2014 3:42 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Here are WIP3 and 4. See the YT descriptions. 279 seconds ahead of Erokky's TAS, due to doing the Celes skip. Link to video Link to video

Posted: 8/21/2014 11:02 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

I finished Ultros 1 without any level up. I lost about 4s compared to the optimal fight, which is not bad considering Terra is two levels lower than on Erokky's TAS meaning her DA deals 250 less damage. Edgar is also 2 levels lower and everyone else is one level lower. This should not be a problem for the next bosses, but I fear I can loose time on some fights during Sabin's scenario. I'm currently 71s ahead of Erokky's TAS.

Posted: 8/20/2014 6:43 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

So, I started over from Mog fight. Here are the improvements: - Took the Mithril Pike and the Mithril Shield. Mithril Pike will be helpful for Ultros 1 at least - Better MTek fight, I got one Metal Kick quite early - Waited during the castle submersion to manipulate the Lete River. This is where the counter increased the fastest (one unit per 2 frames). I computed how much time I had to wait based on the previous run. I waited a little less as a safety measure - I only leveled up Terra to level 6. I fought all 3 crawly battles because I really needed the money, and Locke escaped two of them so Terra got to level 6. No battle was fought in Mt. Kolts. - Vargas was a little better I think due to the DA been done a little earlier. Also, because of the low level, I needed Locke to do a critical to kill the bears, which made the fight more complicated. I had to start over once because Sabin ATB was very low :( - When arrived to the Hideout, the NPC blocked the way up :(( However, after that, I didn't have to wait a single frame to manipulate the Lete River, so I guess without the bad NPC I would have to wait anyway. I didn't have time to go on the save point, but this is not much of a time loss. Also, I think that in general, getting the genji glove early is not worth it. After Vargas, I'm 25s ahead of Erokky's TAS. Before the Lete River, I'm 47s ahead of Erokk's TAS. Now I have to pray for Ultros fight. With the mithril pike but with one level less, I don't think I can do a perfect fight (i.e. killing with the DA as soon as it unlocks). I may have to level up Terra during the fights on the river.

Posted: 8/19/2014 12:14 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

I do not see any pause in South Figaro. But yeah, I think I went up too far when exiting the town. Thanks for the info about the encounter rate. This is quite strange indeed.

Posted: 8/19/2014 11:04 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Thanks for the feedback!

LCC wrote:

Taking more than 2 steps in the forest outside narshe.

It doesn't matter. All the grass that is situated south-west of the forest has the same encounter increment counter (high) as the forest or the desert. As soon as I enter the forest, whatever route I take is the same.

LCC wrote:

Allowing mtek to do tek laser before crossbow (delays his 2nd turn).

Yeah, it was quite bad. It is not possible to do Autocrossbow before one MTek has his turn, but I could try to get a Metal Kick instead. I also can set battle speed to 1 to make them act faster, actually. Might save time.

LCC wrote:

The NPC block in South Figaro (might be part of manipulating river?).

I wasn't block by any NPC in SF, unless you're talking about the shop. Is the NPC in the shop sometimes elsewhere?

LCC wrote:

Also i'm fairly sure you could get less battles in narshe area in general with a better encounter route, but it's hard to say how that would affect the rest of the run and you're doing this from clear RAM anyway.

Yeah, I cannot do anything about the encounter route in a TAS, as a reset counts toward the total time. I tried to save/reset in the middle of Mt. Kolts, but it only saved one battle, and I need to save at least two battles to save time, because the intro sequence in ff6 is very long :(

LCC wrote:

Leveling in south fig cave/koltz surprises me when you're able to do crits at will, but I run the game not TAS it. D:

Well, I can do crits but desperation attacks are very dependent on the level of the caster. I'm afraid that some boss fights would take much more time if my DAs are weaker. Anyway, now I'm at the Ultros 1 fight and I really need the mithril pike :( So I'm gonna start again from the Mog fight and I will try to level up one level lower (ending Vargas at level 6/7/8) and see how it goes.

Posted: 8/18/2014 10:27 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Catastrophe wrote:

Mt. Koltz was solid. I assume you need all the xp for something? Or just getting Terra to level 12 for the magic route?

Well, that was a really really tough decision. I ended up Vargas with level 7/8/9 so that Sabin joins at level 10. The next time I will gain XP in on the minecart, and I fear that I can miss this extra level for some boss battles. I really don't know actually.

Catastrophe wrote:

If you're willing to fight encounters now then maybe grab the ThunderRod early? You've got Edgar and I noticed that you were already xbow'ing all those 3x Crawly formations. Cons: no sprint shoes for those ~50 steps. Pros: you seem to be wanting the xp for something and it makes Locke's scenario faster.

When you first arrive in the cave, there is no thunder rod in the chest. I will try to grab the ThunderRod with Locke just before TunnelArmor and use it against him, but it is possible that a DA is faster (and I can steal the Air Lancet).

Catastrophe wrote:

But each manipulation costs time. Each of the 16 ATB possibilities are reached by waiting n*16 frames before initiating the battle. (This battle might be special because it is initiated with the A button.) So if manipulating the Marshall's ATB doesn't just fall into place with 0-1 delays then it's not worth it to try. And with the Lobos you need a character ready to act who can sit in the item menu for 1-10 frames. Again pretty much just to save 20 frames. So although keylie's Marshall battle isn't perfect, it might be the best possible given the RNG.

Yeah, you have to wait each time for 4 frames before getting another ATB configuration. The problem I had with this fight is that I have to manipulate criticals. Mog is always the first to act and we need a critical from him. Here is what can happen: - I crit with Mog and Lobo attacks - I manipulate Mog's attack so that the next Lobo skip his turn. Unfortunately, Mog's attack is never a critical - I use item menu or whatever to manipulate Lobo, but then I'm very late in the fight, and the Marshal has enough time to attack twice To continue Catastrophe's comments on battle manipulation, you have normally access to 15 ATB startup out of 60, which I can frame rule. To modify the frame rule, there are multiple ways: - go in and out of the menu (very costly) - delay the win of a previous battle (e.g. the XP win message) - add lag frames when escaping a previous battle (escape for a very long time, go to submenus) - delay the naming of a character

Posted: 8/16/2014 5:10 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Second WIP: Link to video

Posted: 8/13/2014 5:14 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

First WIP: Link to video Some details are in the comments of the video. Edit: I finished Vargas. I did the first possible desperation attack, but somehow, I lost 1 seconds compared to Erokky's run. I used the same commands (1 Autocrossbow, 2 Fire and attacks) so it's not a matter of spending more time in submenus. I'm wondering if it's because of the emulator differences. I also lost one second because Vargas cast Doom Fist one second later in my run :( But maybe I can manipulate that. Edit+: I managed to skip both Gale Cut and Doom Fist, nice! :)

Posted: 8/13/2014 8:57 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Link to video I managed to skip a lobo's turn, by attacking at a specific frame. That's weird.

Posted: 8/9/2014 9:28 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Catastrophe wrote:

Any progress on the TAS? I don't think the SG can be improved further unless a better exploit is found.

Hey! I was on vacation. The TAS is currently before the Mog fight in Narshe. The Intangir SG is good news! On Bizhawk 1.7.0, the freezing takes 1-2 minutes. We still have to compute if it's faster than Engulf, because of the extra freezing and the longer magic route. The 312 setup really sucks to get :(

Posted: 7/23/2014 9:27 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Brossentia wrote:

The first is about the queue of the 28th spell availability. It looks like the battle that we'd be using to get goggles is mould 0 with 41FF. FF is the availability, correct? Increasing it by one (by imping and unimping) should put it to F0; is that right? I'm completely unsure, though. Could you possibly clarify this for me? It would be greatly appreciated.

Hi man, and welcome. For the spell availability, you have to write the number in binary. FF = 11111111. When you do something that possibly changes the availability of the spell (consuming MPs, getting imped/deimped), the game checks again if you can cast the spell. If you can, it will push a 0 from the left of the number and throw out the rightest one, becoming 01111111 (or 7F in hex). If you can't, it will push a 1, becoming 11111111. Also, the game defaults this value to11111111 but makes a single check at the beginning of the fight, so if you can cast the spell, it will be 01111111. The three common setups that are described in masterZ faq are: - normal setup: 01111111 (7F) - imp: 11111111 (FF) - normal then imp: 10111111 (BF) You can see that with the 41 type spell, these three lines in my doc are present (417F, 41BF and 41FF) but there are plenty of other ones.

Brossentia wrote:

Also, for item targeting, what if more than one aiming is available? Does the game default to any kind of aiming, or are all those possible? Is it the same with the item flags, as well?

With the Goggles glitch, it will use the targeting of the item to cast the spell, so technically, you can cast spells with other targetings. However, if the game didn't implement it, it is not possible. Exemples of this is Engulf and Bababreath. Even if the hand cursor says it is multi target, the spell will only be casted on a single target. Catastrophe, for the MC equipability, did you also check if the weapon or shield flag was on in the doc? I think it's necessary.

Posted: 7/22/2014 8:21 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Sorry, I didn't update this for a while. It is possible to get Engulfed from the Triangle Island, saving us to do the Floating Continent and the Airship glitch. Using the Goggles glitch, you can cast Engulf by launching a Black Belt. As the Black Belt is not a usable item, we have to flag the item as usable before, by using the sketch glitch. All the information that the sketch glitch is doing on the inventory, arm slots and commands is summarised in this speadsheet. Sorry if it's not very clear. Using the same technique (deflagging), it is possible to equip Moogle Charms during a fight. Some spell setups for the sketch glitch can generate Moogle Charms in the inventory, including Mute type setups, and other setups can flag it as a weapon (or shield) and equippable. We might go for double Google glitch + Slot for Kefka tower or triple Goggles as Catastrophe suggested. I don't know yet what is the fastest. Also, Sneeze has an advantage over Bababreath as it can be multi target and doesn't remove your character from the party. The problem is that the associated item is not usable so a sketch is required like with Engulf. Catastrophe, we (Kadmony, Simon and me) are totally planning to do the TAS. We are still setting up the Engulf strat because it requires multiple sketches. Also, I started the first inputs last week. The Narshe route is way from obvious though, I have to test 2/3 different strats. I think your strat is faster than Erokky's one, it's just that you didn't have a preemptive on the first fight and on the fifth one. There is another Whelk strat using Battle Speed 3 (only this value works): if Whelk body gets his turn before the second turn of the head, he doesn't Gruuu and the ATB is emptied again. So it's possible to kill him with 3 Tek missiles and 1 Beam, saving two attacks. However, the head attacks three times instead of 1 :( I just found something else today: you remember that for most fights with a preemptive or side and a large party, not everyone is escaping on the first try? Well, you can avoid that by setting the Message Speed to 2 (or more, probably) instead of 1. It costs you 16 frames by message (Starting messages, magics, commands, etc.) but it saves 2 seconds per escape. So I think it's worth it, starting South Figaro at least. Well, using previous TASes, it is possible to compute exactly how much time it saves (if any) and where to switch it on or off. Man, doing an RPG TAS with an already existing one is so much easier.

Posted: 7/20/2014 5:20 PM

Post subject: Guidelines regarding uninitialised RAM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

In ff6, the encounter increment is a 16-bit value that increases by a constant at each step. When this increment reach a threshold, a battle happens. However, this increment is not initialised by the game. It is using the value stored in the RAM when the console powers on. With snes9x (at least 1.43 and 1.51), the entire RAM is set as 0x55. With Bizhawk 1.7.2, the RAM is initialised with what is seems to be random values (but same values every time). For my example, the increment set by Bizhawk is bad (0xFDFB) and a battle is triggered after the first step. I was wondering if it was possible to set other initial values of the RAM.

Posted: 7/1/2014 9:48 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Patashu wrote:

Hi keylie, this thread was split off to curb a big debate over what did and did not constitute as ACE in the super mario world TASes, so you might want to ask in the FF6 topic or in a new topic for such a technical question. ;)

Sorry. I made another topic.

Posted: 7/1/2014 9:47 AM

Post subject: How to jump to RAM?

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Hey! I'm currently investigating the sketch glitch in ff6, and my goal now is to get to ACE. I've found many ways to jump to random places in ROM, but as opposed to Super Metroid for exemple, the game's code is executing in banks C0-EF which only contains ROM data. Also, jump instructions whose destination is variable are almost always staying in the same bank. I'm currently trying to execute long jump opcodes: JML (5C, DC) and JSL (22), but there are very rare, even in data. I managed once to jump to $FF:FFFF, which wasn't very useful. Do you have any lead on other ways to be able to jump to RAM in that case?

Posted: 7/1/2014 9:17 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Hey! I'm currently investigating the sketch glitch in ff6, and my goal now is to get to ACE. I've found many ways to jump to random places in ROM, but as opposed to Super Metroid for exemple, the game's code is executing in banks C0-EF which only contains ROM data. Also, jump instructions whose destination is variable are almost always staying in the same bank. I'm currently trying to execute long jump opcodes: JML (5C, DC) and JSL (22), but there are very rare, even in data. I managed once to jump to $FF:FFFF, which wasn't very useful. Do you have any lead on other ways to be able to jump to RAM?

Posted: 6/24/2014 3:35 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Catastrophe wrote:

I might be fatigued, but neither of these numbers work for me.
Template 1-1CE: (2,F4), (61,B6)

Thanks for noticing it. Indeed, the glitch doesn't trigger. In my code, I was testing if the 8-bit variable $822D is 0, but I made a mistake, I have to test if the 16-bit variable $822D-$822E is 0.

Posted: 6/24/2014 12:14 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Catastrophe wrote:

Template 0-10F: (3,17), (21,2C)
Fight the Sand Ray x2 formation that is used in the glitch demonstration video. When using 7E20FF = 1703 the first party member will have a certain set of corrupt spells and a Head Band equipped in the right hand. Now switch to the 7E20FF = 2C21 variation. You should see a different set of corrupt spells and have Goggles equipped to the right hand instead. I don't mean to say you're wrong. I'm just saying that there are more factors to the glitch than mold and size template.

Well, I was wrong, I forgot to take into account the information about the pointer to data sprite that can also change. As a general matter, don't forget that this document is really a draft, there can be mistakes in it, don't hesitate to say I'm wrong when I am :) If you exclude the grouping I've made, it makes a total of about 220 spell setups that set the sprite height to 0. You can list all of them using this lua script using latest version of Bizhawk. About variations on monster formations, I don't really know. The glitch copies portion of ROM into RAM based on values in $822D - $842D. Most of it ($8259+) depends mostly on mold number only. However, $83F1-$8589 holds horizontally flips of each sprite, which requires the actual enemy width ($812F,Y). The enemy width can be different from the enemy slot of the mold, right? If so, you can have a dependency on the actual monster formation. Did you notice if differences in monster formations of the same mold were correlated with differences in sprites width?

Posted: 6/24/2014 10:40 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Thanks a lot Assassin! Indeed, you have very well commented the critical portions of the code during the sketch glitch. This would have saved me a bit of time, but doing the work myself has the benefit of a deeper understanding of the code :). BTW, I also used your Bank C2 doc, thanks a lot for all the work you've done on this! Mold index is independent of spell setup, so yes, you can form all combinations. I'm writing a lua script that will predict all the wrong reads/writes occurring from the sketch glitch, and interpret the consequences for known RAM addresses (item list, in-battle spell list, etc.). I don't think I can predict crashes though, so testing will still be necessary for potential interesting outcomes.

Posted: 6/23/2014 8:36 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Catastrophe wrote:

Nice work! I read all of it. But I have a question about the table of sketch glitch values. How do you read it? For example:
List of potential distinct glitch effect based on Spell aiming and availability. Common spells in bold. ... Template 0-8F: (2,40), (69,68), (69,88) Template 0-97: (69,DA) Template 0-9B: (43,CC) ...
Are you saying that when 7E20FF contains 008F (2 bytes) that it has an effect described by the numbers (2,40); (69,68); and (69,88)?

Yeah, sorry, I was really not clear. Numbers in parentheses are aiming byte and availability of the 28th spell. So if 7E20FF is 0x4002, 0x6869 or 0x8869, you will get the same glitch that query the 8-bit enemy size template with id 8F.

Posted: 6/23/2014 10:07 AM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

Hey, I'm currently working on it with Kadmony who designed the sketch glitch speedrun route. I've worked on disassembling the code concerning the sketch glitch, and I found more than 50 different setups of spell aiming byte/availability that can produce the glitch (instead of the usual 3 that are mentionned). You can find my current work (in a mess) here: https://github.com/clementgallet/ff6-tas/wiki

Posted: 5/27/2014 2:14 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

I tested Spelunky HD while dealing with mouse support, and it kinda works (with mouse!). However, you get the same desyncs as in meat boy (during loading screens). Also, the randomness may not supported.

Posted: 5/20/2014 1:09 PM

Quote

keylie

He/Him

Editor, Emulator Coder, Experienced Forum User, Published Author, Expert player (2830)

Joined: 3/17/2013
Posts: 391

I didn't think it could be done even faster. Congratz! Voting yes

Posts for keylie