Animation Design Philosophy

[Springheel]

I know virtually nothing about animation, (beyond what our animators here have told me), but I do know a little bit about project management. In all projects you have to weigh three factors--productivity, desired output, and timeframe.

 

Our desired output is to have a full list of basic animations for each AI model. Looking at our animation list, that's about 20 animations if we don't count combat animations. That also doesn't include any of the additional animations I've listed as 'desirable but not necessary', like yawning, shaking a fist in frustration, or conversational gestures.

 

Twenty animations. For each model.

 

I'm assuming that we aren't going to suddenly get more animators who show up (and stick around), so our recent level of productivity is probably going to remain consistant at best.

 

(BTW, let me say again this is not a criticism of our modellers. Ascottk has been one of our most productive members since signing on, and I know Solis is dealing with a steep learning curve. It's just reality that animating takes a lot of time.)

 

So how realistic is it that we can create twenty animations for each AI model, if some of them have to be made from scratch? How many new animations can we make in a year? Even if we assume that we become staggeringly productive, I think 2 brand new animations a month is pretty optimistic. That's 24 a year.

 

At that rate, we will be lucky to get all twenty animations done for the AI models that can share them (the existing guards) by two years from now. If we have to start making twenty separate animations for AI that *don't* share animations with other AI (like the nobleman) then we simply will not get done. Period. It won't even be close.

 

So what are our options?

1. Get a lot of new animators.

 

We've tried to do this continually. I don't see any major changes happening here in the near future. Frankly, even adding one or two more animators would only ease the problem, not solve it.

 

2. Push back the release of the mod.

 

We could simply admit that we won't be done the animations by 2008, or 2009, so the mod will have to spend another two or three extra years in development. I don't think that's an option favoured by anyone.

 

3. Start producing far more animations, somehow.

 

Without new animators, I don't know how this would happen in any significant way. I suppose we could really cut back on our standards and use whatever clunky WIP animations we can whip up, but that wouldn't be very satisfying.

 

4. We could release the mod in a reasonable time with all the AI models we have available, but with only a partial set of animations for each.

 

I don't think that's a desirable option. Animations are a very important part of the game, and are often needed to support features--no one would be happy with the bottle-throwing feature if the bottles came shooting from an idly-standing AI, for example.

 

5. We finish a complete set of animations for some of the AI, but set aside other AI models that we don't have time for.

 

This is the only remaining option I can see. Any model that requires a unique set of animations is put on the back-burner until something changes. Meanwhile, we focus on getting the 20 basic ones done and shared among the AI that can share animations.

 

Since the nobles and the forger can't share animations then let's scrap them!

 

You're being sarcastic, but that's exactly the kind of thing we have to seriously consider. Who has time to make twenty unique animations, just to support a model that, frankly, is of borderline quality (nobleman) or usefulness (forger)?

 

I'm not bringing this up because I like the idea of cutting back--I've pushed for a lot of variety in our AI, and a long list of colourful animations, because I think they're very important to a successful virtual environment. But we have to be realistic about what we can achieve in a reasonable time, and what we can't.

 

I personally would rather have fewer AI with complete animations sets, rather than lots of AI that are incomplete (well, actually, I'd rather see half a dozen new animators show up, but...). I also don't want the mod to be unfinished by the end of the decade.

 

If I've made some mistake in my reasoning up there, please do tell me where. If not, then it just comes down to which of the options above people are willing to live with.

[sparhawk]

Well, we can still release the other models as well, just without animations. It's not as if a free project is obliged to do everything if nobody is willing to help out. If people complain, they should learn the skills and contribute. Since our animators continously run away, we have to be lucky to at least have found some that are now staying. There is not much we can do about this.

[Springheel]

Well, we can still release the other models as well, just without animations. It's not as if a free project is obliged to do everything if nobody is willing to help out

 

Well, yes, I wasn't suggesting we hide them or anything.

[New Horizon]

I hate to suggest this but, given the reality of the situation, it is probably our best course of action. Go back and set things up so our AI can definitely share animations. It means some extra work now, but in the long run...I believe it will be for the best.

 

As ascottk pointed out, this would mean our AI would need to be pretty much the same size all around....although I don't think this would affect an overweight character....just the height. Basically, we choose a standard size and resize the rest of the AI to that standard. All males are one size, all females are one size. Perhaps we should even look into what T3 did and have one or two 'base' AI models that have nothing on them but cloth. The gear could be added through the attachment system. This is flexible for both animators and in the end, FM authors...as they can easily mix and match heads, helmets, pauldrons...etc.

 

Just one option, but I think it would be flexible.

[ascottk]

I hate to suggest this but, given the reality of the situation, it is probably our best course of action. Go back and set things up so our AI can definitely share animations. It means some extra work now, but in the long run...I believe it will be for the best.

 

As ascottk pointed out, this would mean our AI would need to be pretty much the same size all around....although I don't think this would affect an overweight character....just the height. Basically, we choose a standard size and resize the rest of the AI to that standard. All males are one size, all females are one size. Perhaps we should even look into what T3 did and have one or two 'base' AI models that have nothing on them but cloth. The gear could be added through the attachment system. This is flexible for both animators and in the end, FM authors...as they can easily mix and match heads, helmets, pauldrons...etc.

The closest generic models w/o any armor are the Black Guards I did for "The Cabal" I was planning on revising for TDM. One is bigger & heavy set while the other is skinnier & average looking. & before anyone has a chance to say their meshes probably won't good for TDM, compare only the meshes of TDM & TDS & you'll realize that the meshes are not that different. It's the textures that can make or break the model.

[Springheel]

Go back and set things up so our AI can definitely share animations. It means some extra work now, but in the long run...I believe it will be for the best.

 

What does this actually mean? Remodelling the AI? Do we have anyone who is comfortable doing that? Is it even possible without doing a huge amount of work?

 

Basically, we choose a standard size and resize the rest of the AI to that standard. All males are one size, all females are one size.

 

But there's more to it than just size isn't there? Otherwise, I would think it would have been fairly straightforward to make the nobleman the same height as the guards.

 

Looking at the existing AI models, five of them are robed characters (4 male, 1 female). Is there any chance they could all use the same animations?

 

The nude pagan should, IMO be dropped, since we're not aiming to support pagans anyway for TDM1.0. The builder forger, as much as I like him, is a character of limited usefulness. The elite merc guard and wolfwere could also both be dropped without being missed if they can't share animations with the other guards.

 

If we got those in game, all we'd really need would be some kind of generic civilian male townsfolk, which could also be used as a house guard with a few attachments. That would, conceivably, leave us with only two types of AI to animate--robed characters and non-robed males.

[ascottk]

Basically, we choose a standard size and resize the rest of the AI to that standard. All males are one size, all females are one size.

We can resize the meshes but we'd have to use the same unaltered skeleton. Also if we resize the animations it'll be the same as resizing the mesh.

What does this actually mean? Remodelling the AI?

The elite merc guard & the builder forger are not the in the T-Pose so we'd have to adjust the mesh so fit an unaltered skeleton. Only the meshes that don't fit an unaltered skeleton need to be fixed. I'm actually modifying the elite merc atm.

Do we have anyone who is comfortable doing that? Is it even possible without doing a huge amount of work?

It's a lot of work & aside from oDDity I've been the only one that's bringing AI into the game.

 

I'll see what I can do the next few days

But there's more to it than just size isn't there? Otherwise, I would think it would have been fairly straightforward to make the nobleman the same height as the guards.

*Replied to NH's post . . .

 

Looking at the existing AI models, five of them are robed characters (4 male, 1 female). Is there any chance they could all use the same animations?

I would think so. Meshes may need to be altered in order to fit an unaltered skeleton.

[Springheel]

The elite merc guard & the builder forger are not the in the T-Pose so we'd have to adjust the mesh so fit an unaltered skeleton.

 

Well, like I said earlier, those two models aren't exactly crucial. The forger is cool, but he's really only applicable to certain kinds of builder maps. And we could make elite guards just as easily by making different skins for the elite city-watch and giving them a different head (maybe taking the elite guard's head/helmet).

 

In my (fairly uninformed) opinion, neither one is worth a lot of work, especially right now with so many more important models (like the peasant woman) still pending. How do you feel about it?

 

It's a lot of work & aside from oDDity I've been the only one that's bringing AI into the game.

 

Hey, give me some credit for the fire elemental and zombie!

[SneaksieDave]

Short of option #1 happening, option #5 is the most reasonable one.

 

If they get a toolkit with two types of guard and one type of civilian, there's still room for plenty of gameplay. Edit: And skins can take care of too much duplication. And if they don't like it, they can get to work and add more.

 

I think it's better to have fewer, complete AI, than a bunch of sketchy ones.

[Ishtvan]

Well, right now, it's not as if animation is the only thing holding the mod back from release. We still have a lot of work in all departments.

 

That being said, I would think something like choosing some "core" animations and characters would be good. Once the core animations are done on the core characters, we can spiral out into animating secondary characters with unique animations, and making some animations that are somewhat lower priority (I don't think anyone will mind if yawning is abscent from release 1.0 and doesn't appear until patch 1.1, for example). I don't know anything about animating, but that would seem to make sense.

[OrbWeaver]

I will freely admit that I am largely ignorant of the issues, but it is not the case that only three sets of animations are needed - swordsman, archer and civilian? Perhaps in real life a nobleman would walk differently to a peasant, but I don't think anybody is going to mind if they use the same animation in-game.

 

Or is the problem that the animations can be done like this, but the models are not ready to use them?

[oDDity]

There's no need to resize the models. Using motion builder, the same animations can be applied to different sized characters, sometimes with minimal tweaking required, or a few new volume bones added.

I thnk you can get the list down to quite a limited set of generic anims for all characters to share, plus some extra small but essential sets like archer. swordsman and civilian. The undead can be used from Doom.

The female set will only need a few generic ones for female civilians, I was always against have any other female types in the game and never felt it was worth the time or trouble, either modleing or animating.

And yes, you're going to have to accept that the animations aren't going to be of movie quality, I don't think many of you appreciate how difficult an art it is to produce really realistic human animation sets. It requires a lot of study and practice, and the fact is that we're never had a real character animator who's done anything, and probably never will at this stage, so you;re just going to have to accept the efforts of other team members who try their hand at animation and do their best.

[Springheel]

That being said, I would think something like choosing some "core" animations and characters would be good. Once the core animations are done on the core characters, we can spiral out into animating secondary characters with unique animations, and making some animations that are somewhat lower priority (I don't think anyone will mind if yawning is abscent from release 1.0 and doesn't appear until patch 1.1, for example).

 

I've already done this in the animation list. It's divided up into necessary animations and desirable ones. There are already nearly thirty on the necessary list, not counting combat animations. Even trimming down the necessary list severely, we would still need:

 

idle (Standing in place, shifting weight slightly side to side, breathing) *

Stand, turn 90/180 degrees (a standing guard turning to watch a different direction)

Walk forward, casual, male/female *

Walk forward, alert (weapon out, looking around, could be trouble)

Run forward, (alerted, weapon out) *

Run forward, defensive (weapon out, arm up to protect face from arrows)

Run away, male/female (panic, no weapon--help meeee!)

Jump back, startled (totally surprised by something right in front of them--not used if alert)

Search forward (cautiously walking and looking around, weapon out, ready for trouble)

Draw / sheath weapon (side--swords, daggers) *

Draw / sheath weapon (back--hammers, bows, axes etc) *

Throw random object (at player out of reach) *

Archer knocking and firing arrow

Pain Animation, generic (torso)

Failed KO stagger (ouch! --just took a hit to the head but isn't KO'd)

Sidestep (dodge to the side, primarily for arrows) (?)

default attack ready (sword up and out in front)

block front (sword held out vertically in front)

generic attack

Stagger forward, blinded (can't see anything)

Reach out (waist level, to open door or retrieve object from table)

 

Now, the civilian models won't need quite that many, but still more than half of them--and this doesn't include ANY colour animations at all.

 

It's not so bad when AI can share animations, as many of the above already exist for at least one model. But any model that can't share animations would have to have them all made from scratch.

 

Or is the problem that the animations can be done like this, but the models are not ready to use them?

 

I don't know, Orb...I keep getting different answers about this question, ever since starting to ask it six months ago.

 

On the one hand:

Using motion builder, the same animations can be applied to different sized characters, sometimes with minimal tweaking required, or a few new volume bones added.

And the other:

I tweaked the skeleton for most of the new AI so we can't use other AI animations due to the different joint rotation.

 

I'm lost....

[Darkness_Falls]

can lightwave be used to animate, or is Maya or 3DSMax or something req'd?

[Darkness_Falls]

Okay, clarification: I know Maya is used for some of the anims here. I don't care about that. If LW can be used for TDM anims, even if it takes 10 times as long as Maya to do an animation, let me know. I may try to train myself to make anims in LW if it'd work.

[OrbWeaver]

I don't know about LW, but it is certainly possible to create animations in Blender, therefore Maya cannot be strictly required. I guess the critical question is -- does LW have an export filter for MD5 available? Maybe Doom3World has an answer.

[Domarius]

Well now that I'm back from my holiday I intend to help you make #3 a reality.

[squill]

if i need to help out with animation let me know. I still do some mapping in the weekends but the pace is slower and animating is something i do on a daily base right now. I'd be glad to help out

[oDDity]

I've already done this in the animation list. It's divided up into necessary animations and desirable ones. There are already nearly thirty on the necessary list, not counting combat animations. Even trimming down the necessary list severely, we would still need:

Now, the civilian models won't need quite that many, but still more than half of them--and this doesn't include ANY colour animations at all.

 

It's not so bad when AI can share animations, as many of the above already exist for at least one model. But any model that can't share animations would have to have them all made from scratch.

I don't know, Orb...I keep getting different answers about this question, ever since starting to ask it six months ago.

 

On the one hand:

 

And the other:

I'm lost....

 

I don't know if ascotk is using motionbuilder or not. Anyway, once he's sorted out a good base skelelton with joint rotations the way he wants them, he can use that for different sized characters. There's no need to make them all the same size and shape.

When I say 'use the same animations' I don't mean have every AI using the same walk.md5anim, you'd still need to export a walk for each character if the skeleton is slightly different, and parts of the mesh need special attention on different characters, so it's a matter of using the same base walk anim,. but adjusting it to suit each character.

The most time consuming part is properly weighting each character mesh to the skeleton, but that can't be avoided no matter what. Our Ai meshes are a lot more complex than the ones you see in Doom - guys in boiler suits etc which are one piece meshes - so weighting and animation in general is a lot tricker when you have flaps of cloth and pauldrons etc. to deal with.

Thinking about it now, I could probably have made models like the city watch elite one piece, and it would hae been easier to deal with from an animation standpoint, though it might not have looked as good.

I'll have a quick go at that now in fact to see what it look like, it might be worth doing it for all the meshes. THe nude pagan woman for example is already one piece, making her very easy to rig and animate.

[ascottk]

Okay, clarification: I know Maya is used for some of the anims here. I don't care about that. If LW can be used for TDM anims, even if it takes 10 times as long as Maya to do an animation, let me know. I may try to train myself to make anims in LW if it'd work.

There's a few of us with Deep Exploration that can convert between maya, lightwave & max.

I don't know about LW, but it is certainly possible to create animations in Blender, therefore Maya cannot be strictly required. I guess the critical question is -- does LW have an export filter for MD5 available? Maybe Doom3World has an answer.

I really don't trust other md5 exporters since the ones I used in max were unreliable. I think it would be best if the people with Deep Exploration convert the lightwave files to maya.

[oDDity]

Deep Exploration cannot convert rigged models form a lightwave scene to any other format.

[Darkness_Falls]

If I understand correctly, LW is out of the question then.

[oDDity]

THere was an md5 converter for lightwave, but I remember you could only use one material per character, which is a real pain, and you'd no doubt run into other problems as well,.

Check out doom3.org lightwave seciton.

[SneaksieDave]

...but it is certainly possible to create animations in Blender...

Orb: is whatever format is native to D3 built in/available in Blender, then? Or do we have to use some old version, some semi-functional script, some massive format conversion process, etc? If not, this is something I might check out, even if just tweaking existing anims, or making small ones...

[Solis]

Well, hopefully I'll be contributing a bit more now that I have a fully rigged model to work with. So I should be able to finish more animations at a quicker pace now, but I'm not sure yet how much time it would take me to do full animations once I get the system down.

 

Out of those options, another I might consider is to put in placeholder/lower quality animations for some of the more difficult actions (or simply recycle most of them from other characters). That way, it's something that can be updated later without messing anything up, and would allow the people using the art assets in levels to have every animation at their disposal while still allowing it to look better in the end.

 

I'm not suggesting we stop making quality animations for the characters, but given that I've seen retail games have characters that sink through the floor, bend their bodies in totally unnatural ways, and walk with their feet not even staying in the same place that they step, a few lackluster animations in the first release of the mod that are updated later on would probably be acceptable (as long as the poor quality animations are actually replaced eventually). I think it's a slightly more desirable option than cutting AI characters entirely.

 

A guideline would be to expect about 5-10 seconds of higher complexity animation from an animator in a week (things like walk cycles, interactions, etc), with less complex actions such as idles or the lower quality animations extending that a bit. If you can figure out about how long each animation is going to take to be completed and then factor in how many you have, you can come up with a timetable for how long it would take to finish all the animations and start figuring out what would be the best places to cut back.