jducoeur

You're viewing

jducoeur's journal
Create a Dreamwidth Account Learn More

Reload page in style: site light

So I've confirmed what I suspected: the new camera on my new desktop computer at work lets me use "live avatars". That is, it detects my facial movements, and instead of showing my face when I'm communicating, shows an animated avatar. Current frame rates are very slow (maybe four frames/second), but Moore's Law will take care of that.

I'm curious about how long it will take before online services like Second Life begin to use technology like this, allowing you to simply "talk" to your camera and have it show up as your in-game avatar talking, mimicking your lips and facial expressions. There are some standards that will have to be developed before this is possible, but it's clearly coming: this toy serves as a nice little proof-of-concept.

So if you try to do a video IM with me, don't be surprised if I show up as a bug-eyed alien, or a shark, or even a smiley face, all of which I have currently installed...

Flat | Top-Level Comments Only

From:

bkdelong.livejournal.com

Interesting - is this built into AIM or something else?

From:

jducoeur

I believe this particular implementation of the idea is happening at the device-driver level -- it's modifying the camera output before it gets to programs. That's supported by the fact that ASAP (the IM system that *we* wrote, which I know has no explicit support for anything of the sort) shows the animated avatars with no difficulty...

From:

dsrtao

So we're what, about three generations to "my videophone me is always properly dressed, wearing makeup as appropriate, sits in the office, and never looks hungover", even when none of that is true.

From:

jducoeur

Hmm. Needs to have about ten times the density of detection points, needs to be able to synthesize an avatar from a scanned model of yourself, and needs to be able to be able to provide a suitable environment for that avatar. Yep, three generations seems about right -- it's Playstation-5 or 6-grade capabilities...

From:

corwyn-ap.livejournal.com

Cartoon avatars are much easier to synthesize than human-looking ones. There is the uncanny valley to consider. Might be easier to photoshop on the fly.

From:

jducoeur

Oh, sure, but I still think PS6 is about the right timeframe. I give it about one more generation before top-end movie-grade animation starts to get humans roughly right, using advanced algorithms and big rendering farms. After that, two more generations to get that down to the desktop level.

Mind, I'm using game consoles as my "generation" marker, so that's several years per generation -- about ten years to get this feature down. Ten years is an *eternity* in this business, so I think it's a plausible guess. Yes, I'm explicitly assuming that the uncanny valley can be overcome -- that you just have to get *all* the details right (including things like skin texture and eye movement) to pull it off...

From:

metahacker.livejournal.com

There was a plugin for iChat that did something like this, sort of -- it posted a new icon for you every 1/2 second or so, taken live from your camera. Unfortunately it changed your actual icon, so it was very disconcerting in a style where every line of chat has your icon attached -- they'd all change at once, in a ripple.

How does the tech you're using work? Does it assume you only have one icon visible on your receiver's screen? Does it change your default IM icon?

From:

jducoeur

Has nothing to do with icons, and everything to do with video -- note that I specified "video IM". Far as I can tell, it's intercepting the camera output at the device-driver level, interpreting my face, rendering the avatar based on it, and injecting that in place of the actual camera stream. So whatever I'm doing with the camera shows my avatar instead...

From:

metahacker.livejournal.com

Blink. Okay, that's completely not what I was thinking. I thought you meant it just noticed motion, and hit "play" on an animated avatar...

From:

jducoeur

No -- this is really quite a bit cooler. It doesn't work all that well yet, but the fact that I can, say, tilt my head to the side and have the animated character follow suit is really neat. (Although I immediately found myself wishing that the shark's fins would follow my hands -- I talk with my hands enough that it just doesn't come out expressive enough without them.)

And there's something deliciously *wrong* about showing up as an animated smiley-face icon (or a stick figure) in live video conversation. It's a whole new realm of expressiveness: when they get the technology really up to snuff, I think this is going to become *really* popular...

From:

gyzki.livejournal.com

Technologically, looks like Mr. & Mrs. Everywhere will be able to get here by 2010, on schedule. Even more so than Brunner guessed - if memory serves, they "looked like you" racially in the book, but this implies they can be made to look like you-you.