[Bf-committers] re: Animation Baking Take II -- response to Ton's questions

Robert Ives robbibob at gmail.com
Mon May 7 22:06:12 CEST 2007


Hi Roland,

you brought up some nice points and ideas there. I totally agree there
should be some bake/merge Strip function in the NLA.
We have had a lot of new functionality added, with the modifier curves in
the NLA and the
offset bones. But the whole approach in the NLA to modify animations with
curves I think
should be added to. Baking is a destructive way of manipulating Animations
and is
always good if there is an option to modify without baking.

I think a good approach in the NLA when modifying Animation would be to have
a "Absolute/Relative" option per strip. So by using the "Repeat" option to
walk loop that has a forward translation and having the strip in "Relative"
mode the loop would carry on walking forward using the animations
translation relative every loop. This together with some kind of redirection
on strip blends giving you a possibility to to turn the Animations on the
blend between 2 strips or on the strip.

I myself are not a great fan of using the NLA in Blender, or any no-linear
editor in any package. I never seem to find
a situation where I will save time using it. And The few times I have I find
It hard to keep the same quailty of animation
I lose to much control over what I set out to do.

Now I am getting away from what your original post is about but I think its
still relevant as to which way our
Animation system is going.

I think we should really look at what our basic keyframe needs are and what
we have got. My major problem
character Animating in blender is the Quaternion Rotation IPO's which
Blender got with the new Armatures
a few years back. I live in Curve editors when I am at  work, it is one of
the most important areas in 3d Animation
but one I think really needs some work in blender. Most other programs I use
have quaternions running in the
background and Euler curves for the Animator to work with, this I think is
the most important change
needed at the moment. At the moment you have the 4 quaternion curves to
fight with, which do a great job of
quashing gimbals, but are hell on earth when you want to tweak that foot
strike or sharpen that kick-off.

I could babble on with what changes/additions I would request in the
keyframe tools but thats for another time
and maybe my own post ;) But I really think before we go any further with
the NLA and the Action Editor
that we nail down a concrete Curve/IPO editor that is consequent for both
object and Armatures.

Cheers

Robert Ives - Nozzy



On 5/6/07, Roland Hess <me at harkyman.com> wrote:
>
> I'll take a shot at answering your questions. Also, I'll see if I can
> get slikdigit to comment/chime in...
> > It still confuses me though... the wiki text mixes up end-user
> > expectations and implementation ideas too much. You might separate that
> > more (ignore implementation ideas altogether?).
> >
> Right -- that's a brainstorming wiki page there, and could easily be
> divided. It's a "first draft" for comments and modification.
> > The wording also confuses me... "key framed motion", what is that? I
> > guess you mean the object Ipo, and not Shape keys? What is a path +
> > speed-ipo, "key framed motion" too? Or what is a lattice with hooks?
> >
> By "key framed motion" I mean motion that is generated by directly
> setting Translation keys on objects or bones in the 3D view, which,
> while not the only method or animation is certainly the most commonly
> used when animating hero or foreground characters. Most of this proposal
> is targeted at enhancing hero character animation.
>
> > Another issue I need to know is "why". What do you require baking for,
> > and which situations really? That is crucial information for being able
> > to implement and design features well.
> > It is well possible you use this baking feature to solve weak parts of
> > Blender animation design? In that case we better solve other design
> > issues.
> >
> The main point of baking animation is for flexibility and work flow. The
> best example is this, a hypothetical project involving some nice
> character animation.
>
> 1. Block out gross character motions by keyframing (LocRot) proxies.
> 2. Swap a character skeleton for the proxy so you can do the character
> anim.
>
> Up until now, everything is as the normal way you approach a project,
> but the user would like to have the character walk around and do a few
> things in between.
>
> 3. Offset bone walkcycling works well only with curves, so to use it,
> you have to bake LocRot keyed motion into a curve.
> 4. Now, you have a walkcycle using Offset Bone in NLA, with a wave, and
> then the character stopping and stooping to pick something up, added as
> NLA strips. Also, you would like to vary certain aspects of the walk
> along the length of the path, like exact foot positioning, etc.
> 5. There's no good way to accomplish this, now. But, if you can bake,
> you bake the entire NLA into a single Action.
> 6. This lets you approach the final round of animation sweetening a step
> at a time -- you're working on a single Action that is the accumulation
> of all the other animation.
>
> You've saved a ton of time, because you used Offset Bone and curves to
> generate the basic walk animation, but were able to tweak it on a
> step-by-step basis afterward, also finely adjusting the transitions
> between the walking and stooping.
>
> One of the comments slikdigit made to me about using NLA is that it's
> not super-useful for "hero-quality" animation because of that. You can't
> finely adjust the output in an intuitive way.
>
> Now, if there were a new blending type for NLA, one that would let you
> add such tweaks as a new live Action, that would be great, but I'm not
> sure of the feasability of such a thing. Like you said, it would require
> a rethink of the entire system.
>
> In short, baking allows you to use Blender's animation tools as
> "construction" tools -- steps on the road to a final, fully tweakable
> high-quality animation. Right now, they are frozen at their point of
> output, and you can take their results no further.
> > Lastly, the proposal doesn't mention animation systems (dozens of
> > related objects). How to "bake" that, and again whatfor precisely?
> >
> I don't know that you'd want to bake entire systems of objects. Although
> you could, it wouldn't be that useful, as far as I can tell. This stuff
> is needed for character animation, mostly.
> > In my idea, we need a really thorough design session once on the
> > Blender (character) animation system. It has severe weak parts that
> > block me from adding improvements or new features. I also lost the big
> > picture or design direction here... :)
> >
> Without a doubt, it does need to be thought over. But I'd be willing to
> bet that people are currently changing the scope and/or shots in their
> projects because "this" or "that" is simply too time consuming to
> accomplish. Allowing more flexibitily in the character animation via
> conversion and baking would help that.
>
> Roland Hess -- harkyman
>
> !DSPAM:18,463decd2876244309314786!
>
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at projects.blender.org
> http://projects.blender.org/mailman/listinfo/bf-committers
>



-- 
________________

Robert Ives.
Character Animator
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://projects.blender.org/pipermail/bf-committers/attachments/20070507/38323474/attachment-0001.htm


More information about the Bf-committers mailing list