I am starting to really dig into the new animation system in Blender 2.55. It seems like utter madness, which probably just means that there is a very clear idea to it that I am uninformed of. I badly need to figure this idea out.
My current battle is in adding a walkcycle to an armature that is already in an animated scene, and already has its own non-walkcycle animation (if I use a blank scene with nothing but the armature and a walkcycle, I have no problems, but that’s not going to be a very interesting animation…). The two serious problems are:
1: When I want to turn a set of animation keys in the Dope Sheet into an NLA Action, I seem to be screwed. Is it really either everything or nothing? Is there no way I can point out the three keys from a foot movement that I want in the action, while leaving the remaining keys out? I literally had to copy the keys not meant for the Action (the Action is for repeated foot movement during a period of walking, which is after a period of getting up from the ground), then turn the remainder into the Action, then copy back in from clipboard. Very impractical and timeconsuming. Plus, it doesn’t really work
2: When I finally gave up on it, I just erased every key except the ones for the Action. Then I made the Action from that (in the NLA Editor). Then, I loaded up a previous copy of the file and appended the Action (after twenty minutes to figure what in the NLA Editor actually constitutes an Action. There is a LOT of stuff crammed in there, in no visible order). Then, it just sat there, doing nothing. I know I probably just need to throw a switch or something, but it is not a logical one, and there are barely any tutorials out there dealing with NLA Actions (plenty for pure Dope Sheet animation) in 2.5
I am desperate. Please, help me. You’re my only hope! :spin:
Also, detailed NLA (not Dope Sheet) tutorials that deal with complex use are highly welcome. I guess it is an area still not covered by Blendercookies or Guru
One way around it is to write a script. I have just posted a script to python and plugins that goes thru the fcurves of an action to modify them. It wouldn’t be difficult to modify it to copy your action and strip out the offending keys.
just playing around with the new blender-2.55 nla-action-editor thing.
It looks like i can overlay different actions (like in the old 2.49 version) - there are options for
replace
multiply
add
subtract
the keyed actions with the other actions in the nla-editor.
If you have already keyed some armature-bones in one action
and this action is already part of your whole nla-playing animation.
Now you want to change, for example only one hand doing a wave-thing.
Then you dont need to redo the whole action to change the hand-anim inside
this action.
You can create a new action with only the hand waving.
Then you put this hand-wave-action on top of the other actions in the nla-editor
and move it there to the time, where the handwave should happen and activate
in its properties the option to “replace” (other actions-strip-keys for this one).
I think what Pipeline is on about is not having to go through the process of creating the new action for only the required bones, from each action he already has. If it’s only a couple, using the NLA is no worries, if it’s a lot i can see his point. I’m thinking perhaps a “filter script” that lets you select actions and bones and create filtered actions.
That said, I’m wondering if you can create masks that remove the influence all but the desired keys?
With the NLA editor, one of the greatest sources of initial confusion for me was the little solo button.
Thanks for the input, guys. I have looked at it hard, and I think my problem is actually how the NLA interacts with constraints. There seems to be a horribly complex relationship between the tow, or maybe even just with constraints and animation in general. The Offset in constraints (copy location and copy rotation) makes things fly all over the place in no logical fashion when combined with the NLA. Sadly, the biped script still relies on constraints, but I may have to change that.
And yes, the NLA is a powerful tool. I wish it had better documentation, but with a bit of luck, that’s coming around the corner!
I am going to see if I get the time between xmas and newyears to rewrite the biped script entirely, using layered animation somehow instead of constraints. Any advanced tuts (pref video) on the topic are insanely appreciated
The main annoyance remains that it is so demanding to make a select few keyframes out of a bunch into an Action. Maybe I am overlooking something…
You can already copy/paste keyframes from the dopesheet into different strips - that doesn’t take long. When you paste, just make sure to be on the frames where the animation started.
This is just a comment of …
very useful discussion …
Ominous statements about the
intricate and mysterious NLA editor documentation.
Another comment.
I would be interested to write documentation
Do we have access to the software writer?
Do we as users have enough use cases to undesirably <concoct> a reasonably accurate picture of the NLA E?
If you know or can familiarize yourself with the NLA, I bet people would be very grateful for that. I would!!! Not sure what a software writer is, do you mean the programmer? I think Ton and one or two others cooperated on this feature, I can check. But what is “undedirably concoct”?
IMHO
I call people who write software in C++ and Python… software writers. Just a personal choice.
I call people who write software documentation for the user … documentation writers
Programmer seems vague to me. Blender does not need more vagueness.
I am struggling with the NLA E.
If we have cooperation from people who accurately know the intention of the software that saves time and effort. For other learners this is desirable.
If we do not have knowledgeable resources regarding the NLA E… we are in the realm of guessing and slow exploration… I consider that undesirable and wasteful … and concocted … the documentation writers are not quite sure they are presenting the NLA E in its full glory accurately.
Concoct … doing something ad lib, perhaps low quality, never fully expressing the great architecture and features that are Blender.
I get what you mean now I think we are still at the level where the basics are covered, but not much advanced. There is not that much information to build on, we need to experiment, ask programmers, and document from that. This is what I do, but it is giving me some problems. Maybe we should start learning circles, where we can learn together and discuss our results, and then later turn it into actual tutorials??
Aligorith’s tutorial is very informative, but the NLA is pretty complex, and without some pictures and handholding, it is still a bit tough to get through it. The basics are nicely covered, though, and the needs are more in the advanced areas.
And eventually, I’m planning on writing the NLA section of the wiki, or at least helping.
Sweet! Did you originally work on the code, or are you just doing a Lone Ranger and helping us understand the NLA before you ride into the sunset??
I thought you might be one of them in disguise. The Lone Ranger wears a mask
Perhaps this ‘learning circle’ you speak of is more viable. How would it work?
No great science, really. We could just set up a single thread somewhere discrete and keep a mutual ‘blog’-like series of posts on everything we do in NLA-related experiments. Like I have been doing some experiments with a walkcycle that directed the steps via constraints to objects (I created, by script, a series of bones in a seperate armature and made the character armature ‘walk’ by stepping to each one in a sequence). Didn’t work well, constraints using Offset (Copy Location and Copy Rotation) are a mess, and I couldn’t combine it all with a detailed walkcycle animation, so I was stuck with absolute foot positions, which sucks. Soon, I am going to try to make a walkcycle in two layers (i.e. two Action strips on top of each other), one with the placement of the feet, one with the movements during a step. The idea is to make broad-use walk motions that are independent of where the character is actually stepping, and then automate where the character is stepping regardless of how the feet are moving.
That’s the very condensed version. The learning circle (bad name?) would do insane things like this and describe results publicly, so we could all gain from each other’s experiences. Then, or now and then, someone (singular or plural) could do a tutorial on what we found out.
I think we are still at the level where the basics are covered, but not much advanced. There is not that much information to build on, we need to experiment, ask programmers, and document from that. This is what I do, but it is giving me some problems. Maybe we should start learning circles, where we can learn together and discuss our results, and then later turn it into actual tutorials??