Crossing the uncanny valley WIP

#321

thanks Andre.
im gonna add hims some arms and shoulders with hopefully an interesting pose, and render out a portrait like still.
honestly cant wait to tackle the clothing, doing hair is a pain, atm most of the time im just waiting for some test render to finish. :slight_smile: the more hair there is, slower the progress

#322

This is some great work, but I noticed in the animation that the upper teeth wiggled a bit. Have you been able to nail that down? I wonder if the slight variation is coming from transferring the mocap be hand, possibly.

#323

thanks man.

some irregularities were present due to the subtraction of the facial performance from the head movement. and since i messed up during shooting and forgot to add some solid head tracking points, there were some issues i had to fix by hand.

#324

That good to know, cause this is just awesome work. It’s amazing how far realism can be pushed these days. We’ve come so far even in just the last 10 years.

You’re texture capturing trick is quite clever. Sorta a way the “rest of us” can get scanned data. heh :thumbsup:

#325


///////////////////////////\

#326

Thanks for the front page plug CGS. It really is an honor

#327

Congrats man

you deserved it :slight_smile:

#328

Wow man, there are some super realistic faces on CGSociety, and then… there is yours…
Just to make sure you stay grounded though… The uncanny valley waits you at the animation part. But hopefully you can cheat it well enough with a nice facial rig and capture.

#329

thanks guys.

Dennik, dunno if youve checked, but the rig/animation part is already done. its just a matter of returning to it some day, polish the rough edges and do rendered sequence.

But for now, finishing this still image is my priority

edit:
a creepy bonus

//youtu.be/7XDZmDe6T_c

#330

Absolutely stunning work, man! :buttrock:

I’m bookmarking this thread. :thumbsup:

#331

Maybe I am missing something but… How did you create this animated texture?

#332

Great Job :wip:

#333

This is the craziest and stunning cg thing I’ve ever seen…

#334

thanks guys.

////////////////////////\

and about the dynamic texture:

░
firstly i needed to finish and polish the mocap, so my mesh followed the reference face as closely as possible.

[img]http://ft.trillian.im/fbe8ac09d8955b867eb63005b20cb31900ee32bc/6ieDgNCJ60gUn4LINHs08Lz8uSE85.jpg[/img]

then i simply projected the video sequence onto my mesh, and baked out a texture for every frame (818 frames in total) i did this for all three cameras (left45° | front | right45°)

░
i combined these three image sequences together in AE, overlayed them with the base diffuse texture for extra details. the final image sequence is what you see in the youtube video.

░
than i assigned it to the model just like a regular texture, with the use image sequence option checked. and thats all there is to it. its a simple process actually

#335

This texture projection technique is something I’d personally try to avoid.

I’m not criticizing at all, just spilling out my own opinion :slight_smile: The reason being is that when you need to mo-cap an actor, 99.9% of the time you’ll be re-targetting the mo-cap data onto something different (a creature, different actor, etc…) so the topo won’t always match and also it’d be cool to actually develop a system for blending the maps in the shader dynamically instead of having it projected from a video, because this way, you can skip the whole complex shader work altogether and just use projections instead.

#336

jo jasne, chapem…a suhlasim.

it was just a test. kinda DIY version of this process:
[VIMEO]28443054[/VIMEO]

when i saw it the for the first time, i was like _looks easy enough…maybe i could do that, and so i tried .)

#337

Just awesome, been looking into this lately, blows my mind how much you’ve been able to do with it!

#338

I’m new mem here.
Which software did you use for converting bunch image to 3D Object ?
Thank you very much.

#339

thanks Dave

#340

Good stuff! It took three years to make?:slight_smile: Now that you have this head done, do you think you have learned things that would allow you to speed up the process to make heads without having scanned data yet still maintain this realism? I think it would be interesting research.

The result is perfect anyway:)