Mechanim vs Legacy Animations

I'm just curious how many of you use Mechanim or Legacy Animations for your Projects, or if you are using Legacy currently do you plan to move on to Mechanim eventually? and why?

I honestly can't say I have put to much effort into learning how to work with Mechanim, I'm pretty comforable with Legacy Animation System it has suited my needs so far, so I just haven't had a real need to move on to Mechanim just yet.

What would say are the biggest benefits of Mechanim vs Legacy Animations besides the fact that you can easily share animations?

Those of you familiar with both Mechanim and Legacy, which would you say is easier / harder to work with?

Sorry if this has been posted before, I guess you can just link me to a similar thread if this has all been discusses already.

 

While I still like Legacy Animation, because it gives me a lot of control, you have some nifty features in Mecanim as well.

Perhaps the best feature, bipedal rig re-targeting: Make a set of animations once. Then share them among every other two legged character you'll ever make.

Root motion: you give Mecanim an animation that already has movement.. and you don't have to worry about the feet sliding on the floor. Especially when it's a movement with irregular speed... such as zombie walks, pushing heavy boxes, or any motion that doesn't translate at a constant speed.

Blend trees: You can blend several animations using a float value. (Yeah, same as legacy, hehe).... but in Mecanim is faster to setup. And I especially like 2D blend trees... take two float values (such as forward speed and strafe speed)... and then start adding the animations to the respective values. And if two float values are not enough (say you want to add angular speed, or even acceleration) .. use a blend tree inside a blend tree (nested blend trees).

Layers: you can override all animations with a layer on top at any moment. For example I use a layer when the character is mid-air... so whatever the character is doing, he blends into fly mode and goes into a free falling state, and lands as soon as he hits the ground.
*And layer mask are very useful, to mix animations. Such a walk animation + torso animation. Just make a layer with a mask that only allows torso and arm animations.

Transitions: They're great to switch and forget, to any state you need, without any conflict. Switch from normal movement, to combat mode... back to normal, to climb ladder, to dying, etc.

All of the above are only the Unity free benefits.

 

@dogzerx2 awesome thanks for all the info! yea those sound like some pretty awesome features. I guess when, or if I ever work a huge project mechanim might be the way to go.

 

Yep! Legacy is still very powerful, and you can do everything you need with it. And also Mecanim workflow is slightly confusing at first... but I think once you know how to use it, it can give very nice results in very little time!

Later!

 

Do you happen to know the difference in performance between the two, especially on mobile? At the moment I'm using legacy, but I've been considering switching to mecanim, just a little worried about performance on an older iPhone.

 

Legacy.

Mecanim is to complicated for my simple stuff and confusing.

Why should I change something when it works good enough for me, right?

 

I use legacy for simple stuff, but if I were to do anything that required non0trivial state management Mechanim would be my preference. At the end of the day if you're making your own complex animation management system with legacy animations you're essentially re-inventing the wheel now that Mechanim is available.

 

ive tried both. im going to use legacy. i have very large collection of mecanim animations, but sadly i cant use them cause of issues with mecanim. and cause i dont want to create crappy low quality game just so that i can use technology that is not ready for primetime.

in my opnion mecanim is not ready for prime time and is still an unfinished work in progress.

dont get fooled/lured into it as a result of marketing hype and nice ui

when petal hits the metal the flags come out

just try any of the mecanim samples and they all have the same issue

sadly without unity supporting events for mecanim animations it is little more than a toy

basically you end up with game that is somewhat unresponsive to input and animations that feel sluggish/laggy

all this is because the code does not dont quite know when an animation completes

legacy work just great non of the above issue exist.

and if you are worried about the same re-targeting animations to different models then you can inexpensively buy fragmotion to re-target them...


for more detail look at my recent posts in character system.

http://forum.unity3d.com/threads/143609-Character-System/page52?p=1400560&viewfull=1#post1400560

basically author has resorted to add WaitForSeconds () in his code and while i guess he could have put better falues to mask the issue its not the correct way of doing it. what should happen is when an animation completes it should cause an event. and while events can be faked it still fake and not proper solution. so mecanim needs proper events.

so keep using legacy until the finish mecanim or get it to a level of functionality equivalent or exceeding legacy...

now having beat up on mecanim as soon as they add events and give more control i'll switch to them in a heart beat cause the ui / unity integration / conceptually so much easier. its just being crippled by such little thing. i just ant understand why its taking this long to finish up... i mean events cant be that hard. internally know when an animation is complete, just have some delegates / events defined that we can tie into, dont have it like black box with very limited functionality toy api.

perhaps someone should take a look at using reflection/instrumentation to get around limitations. of the other side is in .net and even in native unmanaged c++ it can be intercepted / directly accessed and events may even be possible to be added. im not sure the version of mono .net supports such stuff, but microsoft has added api for doing such things in their .net.

 

There is an even system in the asset store that you can use.
Robin didn't have a choice but to do what he did for the Character System.
Mecanim is a vastly more powerful system then legacy. You just have to know how to use it. I have seen a lot more complex systems beyond that of the character system.
According the the keynote from Unite 2013.. there are a lot of updates to Mecanim with 4.3
http://www.youtube.com/watch?v=gKYrdTZc5YE&feature=youtu.be&t=1h33m13s

 

The ONLY reason i use Mecanim : re-targeting. I hate the Mecanim workflow.

A tutorial would be nice

 

+1 haha yes, thats exactly what i say... but i think i see the value of using Mechanim based on all these features dogzerx2 mentioned, and what angrypenguin says. I'll probably continue using Legacy for everything unless i plan to work on some huge game project that requires tons of animations for many unique characters. it will probably save alot of time and money.

 

Thanks for sharing your experience @im , I kind of get the impression that it will be a big learning curve, and will take time to really master mechanim and get it working right, so it seems... ahhh I'll just stick to what works lol! I vote Legacy all the way! haha...

 

The number of animations and/or number of characters isn't really the issue. Well... I guess the number of characters could be since a lot of animations could be very easily shared, but to me that's not a big draw at the moment.

It's the state management that gets my attention. A couple of years ago I wrote a really complex system on top of the legacy system. Characters needed multiple layers of animation, each of which had to play multiple animations at once blended together in order to get the required functionality. With legacy it took a lot of code to get it working, and was pretty model specific. With Mechanim the impression I get is that it'd take no code to get similar output, and would not be model specific. That's two huge wins before even considering that it'd take less time (and thus be cheaper) to implement, or that the animation states could potentially be updated by our animator rather than the coders on the project.

 

i dont think its a learning curve problem. its very easy. its a missing functionality problem that they desperately need to address. so you see that the controller doing strange things to get around that fact that required functionary it missing. so its needlessly like square peg in round hope. and you see kludges of different kind to get around the missing functionality, but that never quite do so it looks / feel unnatural / strange. just try the demo you will see what i mean. and this is supposedly from the people who know mecanim. if they cant make it look/feel natural there is little hope that others will. and they dont.

i sending up with legacy myself. i would be willing to jump ship cause legacy require using 3rd party solutions and we all know that as soon as mecanim missing features are added its the place to be cause its builtin and very easy to retarget and do other stuff with. its just things like limited api, not having events.

@ angrypenguin

yeap, but without like getting an event when animation ends and getting api its sort of wasted since the controller is sort of like in its own little black box that cant be easily interface with. so you dont have ability to micromanage things. and you end up with case where user presses a key and it dont quite response as expected or as it does in legacy.

anways i think there has to be way to bubble this up and have things like mecanim events/expanded api go up priority list.

one way is to continue to use and support legacy. that will cause it to bubble up in priority...

so is there unity site where we can add / vote on what is to be fixed / added in upcoming release. that way the community can influence development...

from some of the posts it sounds like it. ive not seen the site. if someone posts a link im happy to go there and vote for mecanim events/expanded api

oh yeah and update update to latest mono or at least to garbage collector

 

Actually, if someone were to write a good tutorial demonstrating everything mentioned on dogzerx2's original post in this threat it'd probably do well in the Asset Store.

 

I do feel mecanim, like a few other unity addons have been rush out the door. Legacy gives alot of power as is and for coders is the better choice right now imho due to a full api. Mecanim is still a great tool but has limited control where your stuck using there state machine to set it all up. I have played with mecanim quite abit but will still say im not an advanced user so far.

I do think its greatest feature is retargetting. Root motion is an ok thing as is its controller/statemachine and ik is something possible in legacy with code.

Mecanim appears to me, so far, as a quick n easy tool to get animations in and jumping to different ones with little coding ie for artist or novice coder. When you get into serious character stuff ie complex mechanics things go wrong and this system becomes very painful and limited. In my system characters have animations, physics, movement, transitions, control etc all working seamlessly and offering total control. Mecanim takes that away where i have limited api control and having to setup node graphs then painfully syncin with my stuff.

When mecanim offers the ability to use all its features via code then thats when it will be a more complete solution and make me switch from legacy.

 

Pretty sure if this feature existed with Legacy, few people would use Mecanim.

 

Since I'm using Blender to do all animation, I can do "retargetting" pretty easily. I just keep separate .blend file with armature that is fully animated, then after modelling new character, using this shape:

Code (csharp):

1.    O
2. ---|---
3.   / \
4.  /   \

(simplified, but it shows how they're modelled), I just import armature from said "animation" file and skin to new character (rescaling if needed). Works surprisingly well.

And I'm too getting feeling that Mecanim was rushed out. Aside of being overly complex (it's not "import animation, split into sequences by start and end frames, then put Animation.Play(name)" deal like it is with Legacy), there are little to no tutorials explaining how to use it properly.

 

Best of all would be an export of the skinned mesh + animations compatibles with Legacy after a retargeting by Mecanim. A dream...

 

Yep! Honestly im a bit weird that UT create Mecanim as completely different anim solution instead of create extension for legacy anim (like fragmotion) - then we have all power of "full code access" and other nice features which mecanim has now...(retargetting and event system incoming in 4.3...) but i think they have some good reason for this decision....

 

yes its called Not Invented Here Syndrome (N.I.H.S.) a chronic incurable disease.

 

Legacy.

It's my personal opinion that Mecanim wasn't ready for prime time in 4.0 and as of 4.2 still isn't because of missing features. 4.3 addresses the missing crucial animation events and adds the attractive gameobject optimisations, but I've recently read that Mecanim consumes a lot more memory than Legacy, and for 100+ characters on screen that would be an issue. I'd rather see gameobject optimisations added to Legacy.

Animation sharing is fine for human characters, but in fantasy RPG type games, you don't want to be putting human animations on say a skeleton; each 'race' will likely walk in it's own way and need its own animation set. So the retargeting is of limited benefit for games with non-human characters.

We'll look at Mecanim on the next project when its a little more mature.

 

Yes to grab attention for graphics whores who love this stuff when in realworld mecanim has imho too little to offer compared to legacy. As good as it looks in video, in practice, i feel its usage is very out of our hands where we have to follow a very rigid guideline. Most things in mecanim can be and could be possible in mecanim.... look how cool locomotion is in legacy...someone as talented as the creator could probably go further and add retargetting. I know for sure my statemachine blows mecanims out of the water...why? cos it follows unitys own design with component, behaviours and script access/control directly in and from it. Not some software that was created away from unity, bought, integrated and finally hacked around to get usage totally different how unity does things. Why cant we drag a state, or any parts of the controller onto a script field of type mecanimstate etc etc then easily access, modify etc this could apply to all mecanim pieces ie transition connections and so on.

 

Maybe this book will help?

 

I love how it says "guide to the complex new animation tools in Unity" LOL! complex animation tools already is a turn off if i read the cover of this book... not saying the book will be bad... im sure maybe will be good and what's needed a HUGE 500 Page book on Mecanim, ahhh by the looks of this and alot of your responses, i guess i was right to assume Mechanim is a pain in the @$!% to use lol!

 

If I only had money to buy books... Also +1 to above post.

 

Mecanim makes your job 10 times easier in the majority of cases. You don't have to write special controllers for everything, you don't have to jump through hoops to stop your characters feet from sliding. You don't have to mess with goofy physics and scripting when you want your character to jump (you just adjust the gravity on the animation curve). You can do all of this in the legacy animation, if you want to spend 10 times as much time writing code. But look at the word "legacy". It computer terms, "Legacy" usually means "no longer supported." Why spend your time working on something that's not going to be around? I believe unity said at one point, that they would add features to mecanim, and remove the old legacy animation when they had it feature complete. I do not know if that is still the case, but I personally would rather spend my time making my game then redoing tons of animations and controllers.

 

I only got serious with Unity after Mecanim was rolled out, so I thought it would be wiser to learn the new thing right away.

- It took me a while to figure out how to make use of sub-state machines and blend states to keep my state machines readable. For normal animation needs it's quite okay now, but the state machine diagrams still tend to get messy.

- It's hard to predict when and how a state machine will transition to another state. For example, creating a one-shot animation that is entered from the "Any" state will restart the animation over and over because the "Any" transition happens again and again. Trying to unset the transition criterion again might prevent the animation from being triggered. The only workaround is a poll-loop for the state's tag.

- Synchronizing playback of different Mecanim state machines (eg. one soldier helping another to stand up) is hard to do and error-prone for the same reason.

- Without animation events, playing synchronized foot step sounds is also not possible. Luckily, this is being fixed in Unity 4.3.

There are some unique features in Mecanim, like foot stabilization and hand IK, that sound pretty good, but I haven't used them yet.

 

Yes, I agree Mechanim is much more user friendly than legacy. The big deal with animations should be creating original ones and mechanim allows you to retarget.

There is nothing that prevents special walks for races in mechanim compared to legacy and the sharing is still easy and valid between characters of the same race.

In my view Mechanim and UMA are the main tools that will help break games out of the stale war / zombie games genre into new genres, at least for newbies. If you're an expert than either system should be trivial for you and it's likely you have access to a much bigger animation library than newbies do anyway.

 

Doable with raycast and this. Years ago.

 

Mecanim is great for quick and easy results but dig deeper and write something much more complex ie custom rigidbody controller with different size and directions and shapes, add full 360 movement ala mario galaxy, then custom blend trees you can totally control from code, add in sound for playing on certain events, add in physics based moved, quick switching of states from code which syncronize physics, animations, variables, sounds etc and it becomes a whole lot harder. The Complex animation system book should say powerful or advanced animation system if it is trying sell tutorials on a product that is good. Complex doesnt imply that whatsoever. It tells you its complex...hard to use, hard to understand, hard to grasp.

Mecanim does have its place and will only get better. It has will continue to allow great releases but it is also limited in many areas. Try and create a complex state tree, then add everything else ie sounds, movement etc and it gets so messy fast. Legacy allows you to write your own tools and write a more structured and intergrated, even unified system.

 

Ouch for me then lol! I guess i'll switch over if i'm being forced to, but ah that would suck especially if I really don't need all those features, and if something i'm working with already works and I like it that would really suck if Unity just scrapped Legacy like that. I wonder though would they really do something like that?? I would think keep adding to the engine not take away from it unless there is some important reason to remove like some kind of conflict but I can't see how they would be conflict there isn't one now so.... i don't see why they would just scrap it like that... i'm pretty sure alot of users will be upset about that when or if that time comes.

 

unity would be shooting itself in the foot, if not outright committing suicide if it removed support legacy

plus there is huge industry of artists / studios that create generic works for number of game engines, ect that use legacy.

lots of members have games, and large investment, library of characters, animations, controllers, and tools for legacy

it will be years, if ever, before legacy support is gone... windows still has dos command box with batch...

they should keep refining it and improving it and perhaps it will replace legacy, but for not i dont think so. perhaps it will for unity, but what about other game engine and things like daz/iclone with focus on other markets.

and then how do you convert all the legacy stuff. i dont see animation / character conversion from legacy to mecanim tools being offered.

the only thing ive seen in primitive tool that will convert mecanim animation to legacy...

 

ahhh in that case ok makes more sense now... but your points are correct. If my usage is very basic, which I kinda feel it is... I have not found myself doing anything extremely complex just yet with the animation systems. I tried to fiddle with Mechanim once and I was like hmmm where the heck is my Animation Component with List of Animation Elements lol! Something as simple as that is not in Mechanim, unless i'm wrong about that, but I didn't see it.

I like the ease of just dropping in all the animations i need in a single animation component and calling those animations when and where I need it. If this kind of simple basic feature is not in Mechanim then that's a problem. *Again unless i'm wrong and i just didn't do enough reading up lol!

 

This..

 

I can honestly say I have no idea how to use Mechanim. I won't say its bad or anything but for all the animation I do I use legacy as ATM I'm scared of Mechanim.

 

Personally I am waiting for UT To add support for creatures with more than 2 legs and/or arms. I would also like support for tails.

 

In defense of Mecanim, I found it so much easier for creating realistic human movement than Legacy. I dont have the expertise to create my own animation system so I was glad to have Mecanim to fall back on. I was able to visually tweak all the blends and transitions, animation speeds, and create loopable animations right in the editor, so no expensive animation software needed, just imported raw mocap. If realism is not your goal then Legacy might be the way to go for now.

 

The thing that irks me about Mecanim is that I have to get a scene running with all my rigs in it before it will build a .exe properly. Otherwise, the animations just don't play. It's really annoying when I just want to toggle off a few development settings (centering and hiding the mouse, for example) before building a demo.

 

hmm I don't know I think Mechanim is Losing the War on this Thread haha

 

Some people just don't like change. Mecanim saves lots of time. Maybe the free version is more limited, but I find the pro version is very useful.

 

I love change, the reason i attempt to learn new features and languages if it will improve both my abilities and new products im working on.

Mecanim just feels like an animation toll designed for realtime in a 3d modelling package where you create your trees etc and test and have basic controls. Once transferred to a game engine, while still a great product, has so many shortfalls programming and game creation wise.

 

OK sounds good, maybe i'll give it a go when 4.3 comes along.

 

Brilliant news.

 

Good

Bad

Come on UT. You can do better than this.

 

That works really well when you control your own pipeline, and is how this kind of thing has been done for ages. I assume it's also the rationale behind the standardized biped in 3DS Max, and other similar things.

However, it doesn't work so well when you don't control the whole thing, like if you're getting assets from 3rd parties. I think that's where the value of retargeting is for Unity, in that it lends the community, store providers and other 3rd parties far more flexibility in the sharing of animation assets, which are often a huge and costly bottleneck.

 

Nice addon. Thanks for the reply.
Can't wait to see the tutorial that will explain us these new functions.