Very impressive technology, but this is a pretty low rez. video. I'd be very curious to see how the outputted models hold up under closer inspection, under different lighting conditions for instance. I suspect the data is actually very messy, and only works really well from a distance. The Hepburn head looked odd with so much lighting info baked, and the texture and model didn't seem to match closely enough. Still, impressive.
I wish that instead of having to tweak the "average" head to the individual, we instead just start breeding people to look more "average". Make it easier for us modelers in the long run.
Awesome, the part that it works from a single image is imho its biggest strenght. Average gamer wanting their face in a game cant really handle front and side views and putting dots on them or something like the facegen thingy requiers, specially if it would have to be done on a console.
Isn't this facegen? Based on the list of features and some of the "sliders" he is talking about I'm pretty sure its facegen. http://www.facegen.com/
I played with the demo a few months back. Its pretty neat, but there is almost always some poor UV stretching around the nose, due to the way they did the base layout.
Oh and how is this any different from what was in Oblivion? Granted Oblivion looks like a cheap knock off of facegen but same principle. I don't see an option to create new and interesting bodies, where are the monster and demon sliders? Where is the option for power armor?
Um vig, if you don't see the difference between this and Oblivion, you're missing the point something fierce. This can (apparantly, allthough i have my doubts) reliably generate a morphable mesh from a single picture. The morphing itself didn't seem very good yet, which is likely also why they didn't show that off very much, but as a step in a workflow this could be fantastic.
Combine this with topology-brush/surface-snapping tools as can be found in topogun, silo, and at least one plugin for max, and creating a digital likeness become a whole lot easier.
The generation of faces in Oblivion wouldn't be much of a workflow enhancement to me, i think. If it would be i'd use makehuman which seems to do much the same thing. I'm imagining just creating a quick scan with this from a concept you made, creating a realtime model with proper edgeloops and using the highres as a base for displacement, tweaking it in your favourite displacement-paint application
Not only that, but it didn't work remotely like this. It was a head-generator, not a way to use a photograph or concept to generate a textured model.
Aside from that you're right though. It's just that the part that overlaps with Oblivion interests me a lot less than the picture-thing.
also, yes i did, Mop. So does this mean we're not being thrown out onto the streets just yet?
also, yes i did, Mop. So does this mean we're not being thrown out onto the streets just yet?
[/ QUOTE ]
Yeah. Personally I think the significance of how long this has been around is that its possibly indicative that it's not *quite* as good as it appears. If it was incredibly good at generating head models from single photos, it's likely that after 8 years it would probably have been made available and in prolific use by now. But I could be wrong. There may be other reasons why it never really got off the ground. There are various softwares around that do it of course, but I don't think they work so well with organics.
It looks like it's a student project of the author. There is always the possibility that he hasn't turned it into a comercial product, hence the lack of it's availability.
Ah, the Matrox Parhelia had instructions for all this fancy pants morphing stuff built in. It was being pimped as the next big thing for a little while a few years back, but I think it kinda cratered when there wasn't any support for it whatsoever (or none I ever heard of at least).
At first it seems cool, but after a minute of watching it becomes painfully obvious that it's just a pre-rendered bunch of bull. Besides the fact that daz pointed out how old it is, just the fact that it show no interface whatsoever besides a slider is a dead giveaway. I would be more ready to believe it if it was a real time demonstration, in which a random person gave the random reference pics, but this is about as true as that Killzone video being touted as real-time two E3s back.
Replies
That was freakin amazing.
http://www.mpi-sb.mpg.de/~blanz/
http://www.facegen.com/
I played with the demo a few months back. Its pretty neat, but there is almost always some poor UV stretching around the nose, due to the way they did the base layout.
Oh and how is this any different from what was in Oblivion? Granted Oblivion looks like a cheap knock off of facegen but same principle. I don't see an option to create new and interesting bodies, where are the monster and demon sliders? Where is the option for power armor?
Combine this with topology-brush/surface-snapping tools as can be found in topogun, silo, and at least one plugin for max, and creating a digital likeness become a whole lot easier.
The generation of faces in Oblivion wouldn't be much of a workflow enhancement to me, i think. If it would be i'd use makehuman which seems to do much the same thing. I'm imagining just creating a quick scan with this from a concept you made, creating a realtime model with proper edgeloops and using the highres as a base for displacement, tweaking it in your favourite displacement-paint application
Um vig, if you don't see the difference between this and Oblivion, you're missing the point something fierce.
[/ QUOTE ]
I was drawing a connection to the same tool set "in principle". Of course Oblivion looked worse, as I mentioned.
Aside from that you're right though. It's just that the part that overlaps with Oblivion interests me a lot less than the picture-thing.
also, yes i did, Mop. So does this mean we're not being thrown out onto the streets just yet?
did anyone read daz's post that this is 7-8 years old? ¬_¬
[/ QUOTE ]
read?! no one reads on the internets!
also, yes i did, Mop. So does this mean we're not being thrown out onto the streets just yet?
[/ QUOTE ]
Yeah. Personally I think the significance of how long this has been around is that its possibly indicative that it's not *quite* as good as it appears. If it was incredibly good at generating head models from single photos, it's likely that after 8 years it would probably have been made available and in prolific use by now. But I could be wrong. There may be other reasons why it never really got off the ground. There are various softwares around that do it of course, but I don't think they work so well with organics.
One of the authors homepage, doesnt look like he is about generating faces to be used in media.
Maybe thats why the project never got out.
edit: oppsy doopsy didnt really notice the links Daz posted.
Interestingly enough the video would appear to be about 8 years old (first presented at Siggraph '99) : http://www.kyb.tuebingen.mpg.de/bu/people/volker/
http://www.mpi-sb.mpg.de/~blanz/
[/ QUOTE ]