Cloning the Climbing of Shadow of the Colossus in Unity | Part 2
Вставка
- Опубліковано 23 січ 2024
- I will be sliding in my nightmares after this.
For clarification this isn't going to be a Devlog of a full game... yet. I'm just aiming to get the system working, but if I do, man, I'd really love to make a colossus type game.
I'll be doing more one off videos too and coming back to this series when I have big updates. Leave your video suggestions in the comments I'd love to hear them :)
Thank you so much for all your continued support. Love you all!
Music
/ @flowerheadmusic
Flowerhead - Wire Wire Rocks
Flowerhead - Flowerboy
Flowerhead - Jazz Music Playing
Flowerhead - After-Hours
Midnight Commando - Full Duplex
Purrple Cat - Gentle Breeze
Alex Productions - Lifestyle
Barradeen - Asleep - Ігри
It seems like in addition to precalculating everything in one frame to hit high speeds, you've accidentally created pathfinding on these arbitrary meshes as well. You could have little buddies up there with you!
good point
Actually, generating a nav mesh and using it as one might have been a solution. Nav meshes don't really need to be flat, it's just how they usually are.
Rob could make shadow of the colossus x pikmin
@@Girugi i'm not familiar with what goes into baking a navmesh, but I would imagine that would become performance intensive if the mesh deforms.
@@AngelofD000M That’s so horrifying from the enemy’s perspective 😬
Even if i only understand 10% of what is happening i feel compelled to watch because the jokes just hits
Colossus devs: “wait we were supposed to handle all that instead of just smooth lerping our way out of edge cases and covering it up with furry decals and baked animations?”
Having done some 2D and 3D video game math with a Batchelor’s degree level education, I sincerely salute you!
Much respect for handling all the edge cases (no pun intended) especially when the model starts moving and suddenly float rounding changes the outcome making it flip flop and stuff.
I would not be surprised if your solution is already more generic and robust than that of the SotC final product.
Sorry to hear about your back Robert. Glad to hear you're doing better. Thanks so much for making these videos, they're both entertaining and inspiring. Looking forward to seeing where your project goes next!
I understood very little of this, but I appreciate your effort to explain it to me.
These are still the best devlogs on UA-cam, I'm really looking forward to seeing more of the animation work!
I want to make character design for this ! This project looks super cool ! Also I really enjoy this more nerdy style of videos, reminds me of Sebastian Lague's devlogs where i just sit there staring at the screen not understanding a word but still in awe
I would love to have Sebastian's brain for a day, to be able to understand those formulas would be next level.
Oh my god a new episode already!?! All of your devlogs have always been the perfect balance of entertaining and in-depth/educational IMO, but I am loving this newest series where you really dive deeper into your logic. Good stuff, man.
The earth actually is flat and on the other side they keep all of the waterslides that all of the billionaires can use"
You are actually a comedic genius
One thing that I really appreciate here is that this route of solving the problem is extremely performant. One thing that's becoming increasingly common (and problematic) is that developers of all sorts, be they game devs, software devs, backend devs, etc, are really quick to use generalized brute-force a-priori solutions. (in this problem I imagine that would mean something like having the hands be independently searching for valid grab-points based on the current position or something) While that definitely can be worthwhile in many cases, when it's the thing you *_always_* default to it becomes a pretty big problem for performance. (and a lot of the time you can get a lot of free-specialization performance, i.e. parts that you can specialize without actually sacrificing any current or future applicability because it's a non-achievable state, purely internal logic, etc.)
It's just really refreshing to see someone really going at it from a very low-level and performant perspective when it's so easy to just default to extremely high-level and generalized approaches.
You make the best devlogs Robert :) But fair warning, I made a flat earth joke in one of my vids awhile back, and months later, half the comments are still from people thinking I was serious lol
yo its the packaging guy with only 2 likes
jesus _christ._ glad you're ok! that sounds like a nightmare, not being able to do anything for _8 months_
these are still by far my favorite dev logs to watch on youtube, the way you present everything is so fun to look at
The animation when the character is climbing is hilarious 10/10. Nice vid.
Incredibly entertaining video, I love this series
Me: Nice another fun devlog from Robert
Robert: Now we detect the EDGE, remove the EDGE, recalculate distance between the EDGE, and now we move along that EDGE, while despawning all the other EDGES.
💀
Right when i think, I am making a really hard game, this guy shows up with SOC remake 😌💀
So sorry to hear about your injury, my friend, I hope you really get well soon.
All the best for this system and I can't wait to see the how this turn out when finished.!
Well I have to thank you for these 2 videos you made about SotC's climbing system. There is a fan project called BTFL which tries to recreate SotC with the colossi from the cut content and I made a custom climbing system myself using raycasts but your approach is just better, you are genius! Thanks again, this will be really helpful!
i already know this is gonna be good n technical how i liket it
For this series, since you are using so much adjacency information, I recommend the "wing" or "half edge" data structure for triangles. Those make searching super fast and give triangle and edge adjacency information. Might be useful!
I love these types of videos so much! There aren't enough videos about the technical sides of them mechanics ❤
Major congrats on the 100k subscribers, more than deserved!!!
Fascinating to hear about the issues you ran into making this. It looks great and can't wait to see more!
I started watching your game dev is hard vid! Was searching for Shadow of the Colossus videos and found you again! Love what you are doing
A timely upload? Splendid.
I was just casually watching this in my subscriptions when you brought up RASCAL and I was like, hey wait, that's me.
Kinda surreal but cool nonetheless.
It's not surprising you chose not to go with the collider route as it's definitely a tricky thing to get right in the way you'd attach the player and such especially with update order issues being so annoying.
Still I think personally I would end up going with the colliders even if it's harder to nail the basics because it allows everything to be more physically based, and if you want to do anything more than just climb the monster, you'll need decent collision anyway. Plus I think it'd alleviate a lot of edge cases (pun intended) and modeling limitations from the system. But realistically they both have their pros and cons and I'm sure you can do it decently either way. My hunch is just that (assuming you want a nice physical feel with the SotC style ragdolling and jumping around on ledges on the monster and stuff) it'll feel nicer and more cohesive if everything is handled on the physics level. Even if it would probably drive me insane trying to make all that behave lmao.
Your approach with the interesting techniques used to efficiently navigate the triangles themselves is really cool though and at the end of the day you could easily still use like probably 2 different traversal points on the mesh to guide a physical character controller's hands and such and use the colliders on top of that to allow it to bonk around and walk on the monster etc.
No idea how far you wanna take this idea though.
Anyway, cool stuff, keep it up.
It's always boggled me how they managed to get all that realtime skinned mesh collision and physics based animations onto the dang ps2 in the original SotC.
I wish there was more detailed behind the scenes stuff on that game.
Also, I'm curious if the performance of the approach took a hit or how much after robustifying everything like this.
Aw amazing! Well I definitely didn't expect to see you here either! I love RASCAL, it's such a cool asset, I have no idea how you managed to achieve splitting up the mesh in such a way and well as making the colliders persistent. I also purchased Grassflow as a potential way to do the fur of the colossus but never really tried to make it work aha. Can't wait to see what other tools you come up with if you're planning any!
Thanks for the insights! I agree I can see someone doing a good job with colliders, but that someone would have to have more of a knack for figuring out the maths I think, rather than just brute forcing things with planecasts aha.
And yeah, absolutely nuts from team ICO, the mind boggles.
@@RobertThomsonDev lmao, I was almost going to ask if you wanted a free copy of GF for that, but thought it would probably be too difficult to get it to work well on a skinned mesh. I did a test of that ages ago actually with GF1 but it was just brute force like duplicating the skinned mesh a bunch of times to get more source geometry and it was awful but i just wanted to see what it'd look like. But yeah idk, grassflow just isn't designed for it, obviously. I do think it'd be really funny to do it though.
You could probably get away with using GF2 and just calling the update positions function every frame to replace them on the mesh. But youd probably want to do some small tweaks to allow it to make more sense as fur like adding some kind of local directional control so it can like lie down in the right direction like the grain of the hair yknow. As well as probably wanting to add some kind of per-bone simple physics to give it some drag as it moves. But both of those things sound pretty tricky. I'd say probably better to just go with something bespoke. But if you wanna try anyway, just hit me up and I can give pointers and help out.
Rascal wasn't too complicated, just split the mesh into sections based on which bone each vertex is most strongly affected by, and then do a little matrix transformation to reposition the vertices so that when placed at the origin of the bone everything lines up. I figured for a lot of cases you could get away with not rebuilding colliders all the time and it'd be a lot more efficient. Which was extra important at the time since I don't think the ability to bake collision meshes on another thread at runtime existed yet.
But honestly for something like this, I wonder if just using burst to bake the whole mesh as a collider each frame would work well enough 🤔
Though you still need a way to specify which parts are climbable, rascal does give decent control with splitting the collision up based on materials and such so you could probably figure out something with that.
A raycasthit should also give secondary uv data though, theoretically you could bake info into that.. No matter how you slice it, climbing a skinned mesh is an interesting and challenging problem.
Did you say how you're marking which parts of the mesh are climbable? I might've missed it.
Also I can relate to not being good at the maths part. My eyes glaze over any time I'm researching and see some big equation with a million symbols that make no sense. I feel like if they just wrote it out in code it'd be so much easier to understand. Maybe that's something multimodel ai can do with translating them.
But yeah I feel like I can't really understand just about anything unless I work through it and figure it out myself lmao.
That said, I don't think it'd actually be that mathy to use colliders.
A raycasthit gives you the triangle index along with the barycentric coordinates, and with some *simple* math, you could easily maintain the exact deformed point on the mesh you should be holding onto, and just use additional raycasting to figure out movement and such. But that kinda implies that the collider is all one thing that matches the base skinned mesh. Unless you build like a lookup table to quickly convert from the bone based mesh index to the main one, which could work.
Sliding along edges would still be a tricky problem, but I think with enough raycasts you could figure out something. Though the 'pathfinding' type beat you have going with your edge system right now is pretty cool.
Anywayyy, sorry for rambling so much. It's my favorite pastime.
I'm still very new to programming/gamedev so I've been watching videos like these to ease me into the kind of mindset I should probably have as someone who has never made a game before. I appreciate the humour, but also the explanations of how these things work, even if I don't fully understand every single thing being said, I find it all very engaging. Thank you for doing this series, and I look forward to the next part!
Really fun devlogs. Can’t wait for more
I'm really excited to see where this project/system goes, as making something with a mechanic pretty much exactly like this is the entire reason I want to learn how to make games.
I'm sorry to hear about your back, I hope you can fully recover soon! Looking forward to the next video!
These devlogs are so great. All the jokes, and animations, and animated examples like the normals and smoothness D20 (Which is very satisfying to watch) make everything so much easier and more entertaining to learn and understand. Thank you for being so fantastic. Take care of yourself
Yippie a new Robert vid, I adore your stuff!
Your format is great! Really enjoy getting to see your problem breakdowns.
I absolutely love the in depth problem solving, super interesting to watch! Especially with these amazing visualizations. These last two videos were great!
Great stuff! Ready and waiting for when your Patreon goes back up, you deserve it!
Awesome job the would of loved to see all of the other explanations illustrated but really enjoyed what we got this is a pretty fun series thus far excited for the next one
Love your videos! Very happy you're back at making them!
Dude, I enjoyed this video so much! Really glad you are back!
Woo looking great! Awesome to see you back (pun unintended 🤐)
i'm in awe at the amount of dedication and work that went into this project. fantastic job!! hope to be able to do something similar one day:))
super duper excited to have you back, these are always super entertaining. much love bro.
Man this series hits the perfect spot. I really like solving 3d problems with math and your hard soft edge problem is one I've tackled multiple times already. Just at 12:20 where the character starts to slowly twist his logical forward position I was reminded how walking on angled sufaces doesn't ensure "forward keeps being forward". I'm really excited how you are going to solve future problems, as I've already learned some nice concepts here!
My guy broke his back and decided he's going to remake Shadow of the Colossus. What a champ.
He broke his back wtf??
Love ur devlogs man
This looks very cool! I’m excited to see new progress
Oh my! A very quick upload turnaround! Gotta say, my condolences for the spine, I had a debilitating injury last year as well. Regardless, it's really cool to see the problems that one would run into when developing such a system and how you chose to overcome them. Keep it up, dude!
I love these! I'm a games programmer and it's nice to see someone finding all the edge cases!
Wow! I didn't understand most of this! Great job though, you continue to be a super inspiring developer
crosses arms, rubs chin, nods along
throws in the occasional "interesting"
Amazing video, can't wait for more! Hope you're doing better, take your time, we will still be here and waiting for the absolute Gold you produce ❤️
Hello from the other side of the earth! Can confirm that I operate a waterslide for billionaires and there's nothing else here.
Love the new video! I barely understand any of the corner math but i am excited to hear about it. Hope your back is doing ok and hope everything goes well!!
The Greenland size revelation is where it really got scientific
I really reaaaally love your videos!!!!
"are you still listening? man, you must love corners as much as me, you bloody weirdo..." you didnt have to roast me like that 😥😥
Ahhh yes yes the edge of the triangle. Yes the triangle edge. Indeed.
Fantastic, can’t wait for the next one
At 4:00 where you describe the plane and illustrate it pushing the angle to make the ball normal to the coner/point you want - it seems that may be a pefect function for an auto adjusting racket to hit a ball/bounce a ball exactly to the location you want.
This should be a comment in the previus video, but i will write this here so you got more chances to read it.
Before you uploaded the Part 1, i completly forgot about your existence, the algorithm just recommended the video in my homepage and i thought you were another gamedev from the thousands that i see daily. But then i saw i was already suscripted to your channel, so i checked it to know more about this mysterius game dev... AND THEN I SAW "I make my suscribers game ideas" video and i rembembered the exact place and moment i was when i saw ir OVER 2 YEARS AGO. You dont know me, but i usually dont remember that kind of stuff, so yeah... your videos are that good, they have good dinamic and a nice explanation of the math behind the games, you make a difference between all those gamedev youtubers that just want to be dani so bad, but sadly you upload videos once a decade, understandable for such a masterpiece, but now i will miss them till the next one in 2025.
PD: Your videos just like the videos from many other gamedev youtubers were mi inspiration to change my carrer and now i will start to study videogame development, so thanks i guess
Welcome back!!!
I want more, Bob. I'm hooked on corners.
BROOOOOOO hope your back heals well omg that is so scary. Last year I fractured my elbow and was out of work for months and that was mildly frustrating compared to a back frackture. Be well y'all!
you're a gem based on jokes alone. love these
Yaaaa the goat is back! :D
been here since 300 subs i think, happy 100k
the man himself has returned!
THE GOAT IS BACK LETSSSSSS GOOOOOOOOOOOOOO
I remember your video saying that you were a bad programmer and I don't see anything bad here, in fact you are excellent at programming in my opinion!
you’re amazing, well done big brain boy
oh shit dude, sorry about your back! I'm glad you're recovering, and I'm glad to see your videos again!
Really interesting love the video!
THANKGOD SOMEONE ELSE LOVES THIS GAME TOO. THIS GAME NEEDS A REMASTER!!!!
It got one with Bluepoint Games, solid remaster!
woah, that back accident sounds pretty serious! good to hear that you've recovered well, though!
omg dude sorry to hear about your back!
Thanks I appreciate it 🙏 I'm all good now :))
Wow, the back thing sounds heavy
ANOTHER VIDEO! VERY GOOD!
I love your videos
The function at 2:05 is projecting the character's velocity vector on the vector that goes from your position to the red point. v would is the first and direction is the second (I'm assuming there's another function that calculates direction). You turn direction into a unit vector because when you do the dot product of a vector with a unit vector, the result is the scalar (length) of the projection of that first vector onto the second. Finally, you multiply the unit vector by that scalar to get the final vector with the right direction and length. Add that vector to the player's position and you have the point you want.
im soo getting ready for gogogo online!
this is so cool omg
I wanna play a game using the little blob people in your thumbnails as a character
THIS IS SO COOL!!!!!!!
My climbing system is wayyyy less intricate than yours, but I totally feel you on the rambling guy talking about corners.
i really and truley hope you make this into a full game
The code at around 2 minutes works by calculating the projection of the vector V onto the direction. This is defined as the dot product between a vector A, and a unit vector B, multiplied by the unit vector B. It works because the dot product is effectively a measure of the coincidence of the directions of the two vectors A and B.
Given a unit vector B, and a vector A, you can calculate A • B. One interpretation of this is as a measure of how much of A is in the same direction as B. This is because the dot product is the sum of the product of the components of the two vectors:
A • B = A(1) \* B(1) + A(2) \* B(2) + ...+ A(n) * B(n)
If A and B are orthogonal, this sum is 0. If A and B are opposites, the result is the negative of the square magnitude. If they are the same, the result is the positive square magnitude.
Assuming B is normalized already, then this is the same as calculating a scaling factor for B which when used to scale the unit vector B, will produce a vector B' that is in the same direction as B, but with a magnitude that matches with how much the direction of A and B overlap one another.
Basically, the resulting vector from projecting A onto B is the "shadow" cast by A onto B if you were to shine a spotlight in a direction perpendicular to B from a point very far away.
Not the most elegant explanation, but in my defense, I took vector calculus about 7 years ago.
HEY who told you about our waterslides!!.... no one was supposed to know
cant wait to play your game hit a corner and be instant teleported to the end game.
NEW VIDEO WOOHOOO!!!!
he is back!!!!!!!!!
OMG HE CAME BACK IN LESS THAN A MONTH !!!!
I saw this in my fyp (DESPITE being subbed) and mistook it for the previous part, and was like "huh, might aswell rewatch and check up on the channel"
THEN I SEE IT WAS POSTED AN HOUR AGO
Hooray
damn, I wanted to hear more about those corners...
can’t wait for Shaddo ob da Clobobbus
love ur pfp
Hell yeah
I never knew that the whole "sliding against a wall" thing was something specifically programmed in, I thought that was just what naturally happened.
"The earth is flat because they keep the best waterslides in the other side for billionaires to use them" I'm deaddddd 🤣🤣🤣
Nice👍
the time machine! it worked!
Jesus. I'm watching myself coding the same thing...
Only shit man, we are developing the same thing.
Same ideas (uv, corners., same solutions. Geez man. That freaking scary lol
That means I guess we both are in the right path?
Again the only thing that I guess is different, iis that I use a half edge mesh for the relationship between triangles
Well, I guess you are now way ahead than me. I had to stop the project, so I did started with te active ragdolls. But did not progress that much.
Which makes me feel weird, but I'm dang happy for you. As a person that did basically the same thing.,congrats man, I know how freaking hard and complex was to you get to this point