mos6507 - all messages by user

2011/6/15 16:24:30
Facial animation My opinion is that there should be an open API/SDX for the more ambitious and technical people to do what Muvizu currently does internally. There's a HUGE gap between working with something like Muvizu and something like Maya, not the least of which is price. A small team like Muvizu could benefit in having motivated 3rd parties extend the system.
2011/6/15 15:19:36
Facial animation The technology driving all this is rapidly changing, so rapidly in fact that it's hard to keep up with it.
The Kinect motion capture stuff is very promising and I expect this will be a standard way to input animation into machinima. I don't know how long that will take, but it will happen. It's inevitable.

There are some big milestones that I'm really waiting on for Muvizu. I'd like Muvizu to be an extensible animation engine rather than one that is hardwired to making Veggietales squash-and-stretch style animations. So that means not just custom objects, but custom characters, custom animations, expressions, and idle behaviors that help define the style, performance, or theme. What if I don't want them to heave so much when they breathe? What if I don't want them to shimmy their hips so severely on their idles? What if I don't want them to gesticulate ala the Penguins of Madagascar? Maybe a little more subtlety and grace?

Gaining more control shouldn't mean going to iClone and keyframing every joint.
I really like the promise of Muvizu, but it's in a creative straight-jacket where using their system means wholly embracing its cartoony, humorous aesthetic. That really precludes it from being used for a lot of stories where the technology is otherwise capable of doing it.
2011/5/6 16:04:15
What actions would Muvizuers like to have? I'm waiting for the ability to have different character themes or personalities. Right now the style that Muvizu evokes is "Veggietales". This is through both the character styles and the way they move around and squash and stretch. I know that custom character imports are on the cards but I would also like to see the ability to ratchet down of some of the exaggerated movement, the extreme hip and arm swaying and the abrupt Penguins-of-Madagascar style movements. These are great for comical stories that rely on punchlines and pratfall style humor, but not for anything more serious, where you want to develop empathy for the main characters.
2011/2/15 21:34:17
What actions would Muvizuers like to have? Motion capture doesn't have to pop. It depends on how the engine eases into and out of the motion capture, and how it blends with whatever else is going on with the model (head nods, walk cycles, idles, etc...)
2011/2/15 18:59:01
xtranormal no longer free... As the resident Xtranormal interloper, let me say that when I write about it, it's not to bash Muvizu, as I'd like nothing more than for Muvizu to wipe the floor with Xtranormal.
However, XN is where I've invested all my time, and I've grown really attached to their character designs. The lack of stylistic variety in the Muvizu characters is a big impediment to me jumping over. I just don't identify with the Veggie-Tales aesthetic. I fully expect that to be remedied sooner or later, so I do watch the progress here with great interest.
XN continues to be a comedy of errors, the most recent screwup being the unannounced removal of the Nuance voices in State, which has buggy side-effects.
But on the plus side, someone on my mailing list revealed a crack that allows us to retexture all of the characters and sets, something that will at least give us some more flexibility which heretofore was limited to their "Suitz" package in which they erroneously deployed with external textures.
Since that time I was unable to convince them to officially switch over all their assets to external textures. They promise some form of user-friendly "customization" within six months or so. I don't believe in waiting, and they have a history of overpromising and underdelivering, so we will proceed full speed ahead with the crack.
I much prefer the cordial and respectful way that Muvizu devs interface with the users to XN. XN power-users have had to huddle together and treat the company as more of an antagonist than anything else.
Here is the first result of cracking the XN object format:

I wish we had some really good texture artists over there.
2011/1/11 14:12:27
New character? "We need to stay close to the original proportions of the char or the animations which have been made for the original character will look terrible. "

There is a solution to this. The way Xtranormal works is they describe the various body parts (think, bones) of the model and they use this information to help offset the animations to conform to different body types. Even for animations that appear to use motion capture, they are able to reuse the same animation files but have them work with characters with big or little heads, etc... I can tell they are reusing them because they point to the same physical files.
One of my next hacking projects with Xtranormal is to do more research into these settings, but it's a pretty flexible system they have there. Some of the animations are actually algorithmic rather than canned mo-cap, as such you can script them out just with a handful of settings. It works well for arm/hand gestures.
It's something to look into.
2011/1/10 18:41:08
Disabling blink behavior Another question about the Trek bridge set. Is there a way to stop the blink cycles entirely with the consoles? I find it distracting but when I roll the slider all the way to the left all it does is slow it down a lot but it is still cycling.
2011/1/10 18:39:12
Disabling shadows Is there a way to disable shadows on a given light, object or scene?
I tried using the Star Trek bridge set and because of the upper console overhang, it results in a really ugly shadow cast over the viewscreens. When I lower the light, the shadow from the railings start to get ugly. This is one case where I just want more of a simple fill light effect with no shadows at all.
Is that doable?
2011/1/7 17:26:39
Movie in background 83fgh5t1 wrote:
Nah, Muvizu can't produce a render of a single character or object. It grabs the whole viewport as a 24-bit image without Alpha channels.

But, if you're forced to do chroma-keying then you'd get the best results using Muvizu's TGA sequence output.

That's what I thought, although graphics cards certainly understand alpha channel, at least with 2D images. The codecs that Xtranormal uses aren't exactly lossless and it maxes out at 720P so I do have issues with pixelation on the edges. It doesn't help matters that there is no camera animation in Xtranormal so I have to often zoom in on these images. This would hopefully be more manageable with Muvizu. Probably the first thing I'd do with Muvizu is try blending Muvizu characters in Xtranormal and vice versa, so this is on my mind.
2011/1/7 17:23:13
Facial animation I would like to again mention how Xtranormal does things as a point of contrast. Not to say how it does it is best, but it is is pretty effective.
It just has an icon for expressions and you pick the type of expression (happy, sad, surprised, disgusted, angry) and control the percentage and duration.
The problem with this, in my experience is you can not have someone gradually get sad or gradually get more and more angry. It resets the expression before firing off the next expression.
I have managed to have a character get more and more sad by repeating the same sad icon a few times and ramping up the percentage, and it worked well in that isolated context, but there is a noticeable stair-stepping effect when it moves from one icon to the next.
Anyway, the way this works is a series of morphs. The character can still talk over this, but with the more realistic characters that have fully modeled mouths, it can interfere with lipsync. If a surprised expression pulls the mouth open and face back, they won't completely close their lips during lipsync. It's hard to do this right.
This is one of those things that is hard to simplify. You really need to provide some advanced switches. Like to turn off mouth influence or be able to raise individual eyebrows, control ease in-ease-out, etc...
I have played around a little with IClone and they have a new facial tool that lets you puppeteer the face in realtime with the mouse and other controls. This is awesome, but it's kind of overkill.
As I said in an earlier thread, emotion is conveyed by the whole enchilada. You have the dialogue, the face, but also the gestures. For stylized characters rather than the Beowulf-types of IClone, a simplified expression system is more than enough.
2011/1/7 17:10:51
Movie in background I'm using Sony Vegas, not after-effects. I understand the keying is better in after-effects. I was just wondering what the technical limitation is. With full raytracers they have the ability to output with alpha channels. Is it the realtime nature of this the reason why it can't be done?
2011/1/7 17:04:14
What actions would Muvizuers like to have? Hi, I thought I'd be able to contribute to some of the theoretical discussions like this before I really dig into Muvizu, based on my prior experience with Xtranormal. If I say something that Muvizu already does, my apologies. Take what I'm saying as the wisdom of a complete outsider.

What Xtranormal does is separate animation into discrete channels. So there is a head zone (for yes/no) main gestures (which are mostly upper-body) postures (which are more full body) and expressions (the face only). Then there are walking animations and the lipsync.
It is then able to blend this together. So if you tell your character to do "hands on hips" and walk at the same time, they will do just that. Somehow they determine how to establish priority and blending between the two animations. So while a walk cycle would normally have swinging hands, instead that part of the body has the arms at the side.
In this music video, at 1:10, the character walks while hands are on his hips. It's kind of a weird shoulder position he gets into, but it works.

Likewise, when a character is in a sitting posture, then all the gestures only impact the upper body and therefore don't disrupt the sitting pose.
What this does is allow you to create a richer performance than you'd ever be able to do with just one main animation "channel" for the entire body.
If you only have one channel then you need a zillion variations. You know, hands on hips with head nod yes. Hands on hips with head nod no, etc... That's a game you just can't win.

The kinds of gestures that I would expect first are ones that you'd use in the course of dialogue, as body-language. Think of how people move when they speak with their hands. These are the things that Xtranormal added first. The reason these are important is that if you use text-to-speech, the hand gestures go a long way to filling in the emotional content that is missing from the vocal delivery. So when Xtranormal added the Starz pack, they added stronger gestures like hand-punching-fist that work well with people like Mel Gibson. They can speak like automated call centers, but by doing the hand-punching-fist move, they will come across as menacing.
They also tailor the way these animations play out based on gender or body type. So the female version of a particular animation may have more hip-sway or graceful movements. At the very end of the youtube clip I embedded, that coy little head tilt move and the eye movement is all part of the idling system.
What I would like to see from Muvizu is an emphasis placed on these subtle aspects that make characters appear more lifelike, more Pixar and less Veggie-tales. Then it will be possible to empathize with them. I don't know if the company wants to move in this direction, but that's my wish.
2011/1/7 16:16:10
Movie in background Is there any way for Muvizu to just save a 32-bit PNGs with alpha channels so that there's a nice antialiased edge around the characters? Greenscreening has obvious limitations. I've done tons of keying with Xtranormal and you always notice the edge to some extent no matter how good the keyer. Then you've got to worry about green in your foreground objects.
Here is an example of chroma-key in Xtranormal. You may think it looks okay but if you watch it at 720P and look closely, you'll see the hard edge.
2011/1/6 14:23:01
xtranormal no longer free... Hi, I'm a long-time vet with Xtranormal, disgruntled with the company as most are, and looking for another pony to back.

Here is a recent example of the stuff I've done with Xtranormal lately.

I think that some of the areas around character movement/idles and expressions are more advanced in Xtranormal than anything else I've seen, certainly better than Moviestorm, but the underlying engine itself is very weak. No character interactions. No collision detection. Static sets. No texture modding, etc....
I've learned a lot about how Xtranormal works under the hood, as this video represents a texture mod and character-sharing hack that I came up with, and am interested in comparing it to Muvizu and/or assisting them directly in bringing over some of those ideas.
pages: 1 2