Crossing the uncanny valley WIP

#341

Quoted for agreement.
I am shellshocked.

Can’t wait to see what you’ll do with the raytraced skin shader coming very soon with 3.0.

#342

Where’s the tongue? :stuck_out_tongue: :stuck_out_tongue: :stuck_out_tongue:

#343

thanks for the comment Phil. well more like 2 years (27 month to be exact, with shorter and longer breaks of course)

and basically everything i did on this project, ive learned on this project. i didnt know how to do stuff like this before. i didnt even had a single commercial project when i started working on this. that all just came along. i didnt want to rush anything, i just had a vision, and i want to develop my skill set until i got there.

so yeah, i could definitely done this much faster, since the next time, i dont need to learn how to render skin, make eyes and teeth etc, i just need to make em :slight_smile: + theres lots of stuff i could reuse from this project, which can speed up things further.

about the scan data. i feel that people still dont fully understand, that there werent that much usefull scan data involved. i read a lot of comments saying, ‘wow, look at this, its scanned…etc’ which is a bit misleading, it wasnt as simple as that :slight_smile:

basically, this is where the ‘scanning’ ends. by retologising the main face areas :

after that, it was a traditional modeling task, with one major difference


the amount of reference cameras.

for me, this was the major benefit of using 123D Catch, that i had plenty of cameras calibrated and accurately placed in space. thus i could check my model from various angles and make sure that the mesh stayed on model.

i did this cause i dont think that modeling from a front and side orto is appropriate. i wanted to do things differently (the right way)

when texturing, i did a similar thing in pftrack, calibrating the camera positions first, and after that projecting textures on the model. that also had nothing to do with scan data. same thing with the motion capture- nothing to do with scanning.

so for me it was not so much about scan data, but about the workflow.
sure it can be done without it, its just a matter of your preferences. there are many very skilled zbrush artists (much more skilled than i am) that could sculpt a face like that in few days (hell…in few hours)

…so for clarifying some things.

Dave, the tongue is comming, dont worry ,)

#344

Wow this is “THE ONE”! Ive read through your forum 3 times without looking away from the screen!!! Wow. People said everything about your work and I don’t want to repeat again, but I do have a little question( more than a little): can you please run through your shader network , the setup. I understand it and how it works it’s just I never created any custom shaders and I tried to recreate yours but I don’t know what goes where connection wise. I know this is more off topic but I would be forever in dept to you, and if you chose not to answer and that this is off topic, no hard feelings I already learned more that I could ever ask from a forum thread. Thanks and keep being amazing!!!
P.S.
WOW

#345

thanks Papp

what exactly don’t you understand connection-wise ?
basically you can tare down a skin shader to three main components :

░

  1. SSS
    now im using two sss2 materials for skin scattering

░
2. Reflectance
im using couple of standard vray materials for the reflections and speculars for mimicking the multi lobe reflectance appearance of a real skin. having it broke down like this allows you to control the bumpiness of each individual reflection layer separately.

this comparison between a random real skin (left) and my render (right) nicely demonstrates why that is necessary

░
3. Diffuse
is for adding a slight roughness and fuzziness to the surface.
more on that in post #229

try also checking my thread on CGFeedback for more info regarding the shader

#346

Thank you for that, Basicly I understand the layering , what I don’t understand is this:

| Main material - the one that gets applied to the object |
From what I gather is a Vray Material??
||
V
(What shader is this?)
There are 2 Skin Blend??(one for the diffuse part going into the Vray , one for the Reflectance part) and the displace.
||
V
Now it gets tricky for me ( there is a skin reflection material ( I think its a surface shader) that goes into the reflection and it has all the textures going into it like reflectivity, roughness and so on
How do you make up all the slots in which everything goes into???

And one last thing, how did you obtain the reflectivity edge map?
Thank you again for all the help:D you are my hero…:smiley: (sounds gay sorry …Thanks dude:P)

#347

the material thats plugged to the mesh is a vray blend material of course,
and at its disposal you have all the slots youre ever going to need.
the base material is SSS2 shader, and everything else is vray standard material. (u cant use surface shader for something like reflection material, since its not reflective at all)

i dont know what do you mean by reflectivity edge map

#348

I was looking for the Blend in a wrong place:D I found it and as for the map I’m talking about it’s here
Thank you again

#349

whou, thats a very old network. theres almost nothing left of it.
you’ve read through the thread 3 times, but seems like not very carefully ,)

the current one looks more like this :

even this one is like 8 months old…but the core structure remained the same.

now the network is a mess, but who cares .)

#350

Oh yeah I didn’t mean it like this actually. I work in game production and scanning heads is pretty much unavoidable if you want realism, otherwise proportions will usually be wrong and a lot of details are missed even by talented artists. Also, on a production where you have multiple artists, the consistency will be affected. I don’t believe much in the long term prospects of hand-modeling heads, I’d prefer rationalization of how a human head is made, why there are differences, and find a way to parameter this from a base head (tag data, then cross-ref it against parameters to produce heads of all types as far as the data will allow it).

So if the scanning process and the retopo process could be fast, you could easily grow your data bank and start making a good software to make heads fast.

Really looking forward to your future work, either as a continuation of this or anything else:)

#351

interesting idea Phil, that would be really awesome

#352

Hi

I just found this thread through a recommendation on twitter, just finished reading it all!

Absolutely amazing work and you are super cool for sharing your steps with the community. I have a couple of questions please since I’m working on something similar at the moment for a feature but definitely not on this level.

  • In pftrack, why do you use the mocap solver for stills? I always use the normal camera solver, since they are stills then its as if im solving an object to get the cameras.

  • Can you please expand on how you achieved the “global transformation subtraction”, you mentioned motion builder? what are the steps if you dont mind. I was stuck with this issue and i decided to solve the face motion in the camera instead of the mocap by tracking only the points that moved rigidly and then duplicating the camera and freezing it to essentially stabilize the face. It worked fine but not 100% accurate

Thanks

#353

Thanks man.

If you’re asking about calculating the camera positions in PFTrack for the texture projection phase.
Than yes, i did use the camera solver node. Dunno, but if i wrote somewhere that i used the mocap solver, than that was my mistake. Cam solver is the way to go obviously .)

And as for the head movement subtraction, just folow these steps, which is what i did
Just make sure you have some solid head reference points. i forgot about them and had some huge problems, i had to use the ears and create a custom locator for the nose. that introduced a bit of waviness to the final solve, which i had to fix manually(painfully)

#354

Awesome work man!
Is the tongue also mocap driven, or is it keyframed?

#355

thanks.
its keyframed

#356

i like what you have done so far and i was curious about how u get the colors for your beard and hair and make them look so nice … of course you used maps for that (maybe :slight_smile: )but can you please show me that map of yours ?

thanks for your detailed information that you have showed us
really impressive :thumbsup:

#357

im plugging the texture into a one slot ramp, so i can then mess with the HSV color noise to introduce some variation.+ different textures for root and the tip color + im also using vray dirt for more shadowing

#358

hey guys, id like to hear your opinion on something :

l. or ll. ?

#359

I find II more believable. Awesome work :slight_smile:

#360

I would say II as well. More appealing and realistic (?)