Portrait, Director Werner Herzog (in progress)

Werner is Done!Well folks, I’m declaring this project finished. A huge thank you to the over 7,000 views as folks have followed along and offered advice and encouragement. In particular, shout outs to the following forum members who offered input that truly helped make this piece better than I could possibly have done on my own:

Samratek
Chafouin
SterlingRoth
Julperado
RickyBlender
Bonrw1
Photox
PSMan012
Kerub
Minoribus
FXR
Clockmender

My final render, output at 5,000 samples with post-production in Krita and Photoshop. I’ve cleaned up the bump map and eye shaders since my last post. I encourage you to click on the image for the full resolution for the best view…


===============================================================

Hey all, figured I’d start a follow-along thread as I start my most ambitious attempt at photo-real in Blender. I’ve still got a lot to learn, so that’s part of the exercise here. I’m attempting to match this portrait of one of my favorite directors, Werner Herzog. He’s got a great face, plus he’s a super interesting guy.

Please, please, I want and need your feedback. Looking forward to hearing all your great hints and tips!

Here’s my base mesh created by vertex pushing, no sculpting, just using front and side references and a lot of nudging. Laying in quite a bit of geometry here given I want to get this photo real…


Starting sculpting. Working everything symmetrically for now. My plan is to use a small rig later to alter the facial structure to make it asymmetrical. Love the ability to mask during sculpting…


Using side reference imagery for his ears. Swapping back and forth between an image plane, and an image reference (right panel…)


More Dynotopo sculpting. Have also established base meshes for his shirt (visible) and jacket shoulders (hidden…)


Dynotopo complete, retopology will be next to prep for a lower resolution mesh… This is as detailed as I’m going to take the sculpting given I know I can get a lot of finer detail by painting in bump. Looking forward to trying the GLSL viewport bump painting something I’ve been experimenting with on some smaller test models…


Hi - this looks extraordinary. Great start and very inspiring. One thing of note and it may be because of the mirror modifier but in the views you’ve provided the distance between the eyes seems to be a wider than it looks in the reference photos. Thanks for sharing this!

Thanks Samratek. I’ll check the proportions on the eyes. It’s amazing how you can get so tied up in details that you loose the larger proportions. Just after I took the last sculpted screen grab I did go back in and make some adjustments. Swapping between ortho / perspective, and viewport and camera views can make a huge difference as well.

I’ll check…

Checked the distance on the eyes, Samratek is right, the gap between the eyes is a little wider, but this is caused by the fact that the sculpt so far is symmetrical. The (camera) right side of the face matches up to the reference image, but the left side doesn’t. Once I use the rig to make him asymmetrical I’ll take a look at some adjustments to match, plus his head is likely tilted some in the photo.

Here’s a quick snap of the face lining up on camera right with the reference image…


On to retopologizing the mesh. Here’s the original base mesh topology…


And here’s a snap after dynotop…


Here’s the re-topologized mesh…


I mentally wrestled with the idea of making this a project purely completed only in Blender, 100%. But then I realized that sometimes being purely in one application becomes about struggling unnecessarily when there are other tools that do things better.

I started retopo in Blender and was pleasantly surprised by how great the mesh snapping tools are, but even though it worked quickly and smoothly, no matter what settings I checked, I kept getting visually lost between the X-ray of my new mesh and the underlying hi-res sculpt or turning x-ray off and loosing sight of my new mesh (I welcome suggestions.) In the end, I switched over to Topogun. Hey. I’m lucky enough to have it so why not use it? I think it does a better job of visualizing the high and low res meshes at the same time in the viewport and compared to most 3D packages, it’s very affordable. 3D should be about doing things quickly and efficiently, no? Here’s a shot of the lower-resolution target mesh in progress…


I tried a series of normal map bakes in Blender and after some frustration also in xNormal. I finally got some acceptable results in both, but still couldn’t get my new retopologized low resolution mesh to look like my high resolution sculpt.

In the end, I decided to try displacement. I’ve had a terrible time in the past trying to get displacement maps to do what I want in Maya / Mental Ray so I had initially avoided it. It was great to see how quickly and easily they’re both baked and applied in Blender.

Here’s a quick test of the low res mesh with a subdivision modifier and a displacement texture. I also used this chance to try a quick sub-surface-scattering test with the default Blender Render “Skin 2” preset. Super fast.

Eventually I’ll render this in cycles with a more complex SSS material, but as a quick test, it’s a great way to preview…


Here’s a quick shot of the final displacement map I’m using without using sub-surface-scattering. You can also see an initial eyeball texture I’m working on.

It’s amazing to see how much of this displacement sinks away using the default SSS preset.

Once I get into Cycles node development, I’ll have to play with bringing back as much of this detail as possible in SSS while keeping it natural. I wonder how much of his lines and wrinkles I’ll be able to bring back with bump mapping? Time will tell…


Great step by step illustrations. I’m getting anxious to start a human model face myself.

Here’s a quick snap of the UV set up used to bake the displacement map. I baked a 32 bit float EXR to get the maximum level of detail out of the modifier.

The UV squares aren’t as even as I would like, but Blender’s handy stretch detection confirms that at least there’s little to no stretch in the map.

I tried making some additional seams under the chin and around the neck to reduce the disparity of the checker size of the chin and the rest of the head and the shoulders, but nothing really made much of a difference. In the end I did a simple T-cut across the forehead and then over and behind the back of the head. This is actually my first time cutting a map for a head. I welcome suggestions on how to get the best even map.

I’m mapping at 4K so there should still be plenty of detail…


Rather than move on to bump mapping before texture I realized a better approach would be to lay in my skin texture first and then later derive a bump from that so that I can match it up more accurately…

Given I can’t get access to Mr. Herzog himself I’ve been hunting for an approximation. I found this guy over at 3D.SK. He’s a similar age with a similar skin type plus he’s lit pretty flat compared to most of the folks over at the site. I’m also using a couple of shots of a nose of a younger guy who’s nose complexion better matches my reference photo for Mr. Herzog…


In texture paint mode I’ve set up a 4K base color map using the “add texture slot” function. Starting to lay in some color using the texture stencil painting method to pull in my reference images. Switching between “shadeless” and shaded object in the viewport to check accuracy and using the clone tool to blend together areas of the head …


Remember back in post #10 when I said I felt I couldn’t get a satisfactory Normal map out of Blender? (or later out of frustration, xNormal) - turns out the Normal maps out of Blender were just fine. It was human error. I had forgotten to set a flag in Blender Render to tell it that the texture being used should be treated as a Normal.

Here’s what the Normals looked like before I figured out my mistake:


And here’s what they looked like after I figured out where I was going wrong. The “switch to flip” is in the marked red box…


Much better! Of course, like all Normals, there were a couple of minor issues. In this case where the ears got close to the head. Understandable. I used the paint tools to fix them up as best I could with a clone brush. They’re not perfect, but they’re much better, and will mostly be covered by his hair anyways…


Moving forward I’m going to use a combination of my normal and displacement maps now that they’re both where I need them to be…

With the normals issue fixed, it’s time to turn my attention back to texturing. This is my first time using the slot painting system in Blender. I’ve had some experience in Mudbox which has similarities, but works more like Photoshop. The Blender slot system is pretty flexible though with practice. That said, I limited my diffuse texture work to one “slot” or texture which mean’t needing to be more assured about my texturing choices. In a world of undos and multiple layers, there’s something liberating about needing to be more accurate working everything on one layer.

Here’s a shot in progress, working my way around towards the eyes to use the less visible parts of the head to warm up to texture painting directly in Blender. The image on the right is one of the reference images I’m using for stenciling. You’ll also note that I’m testing some light set ups and also using the recent addition of viewport ambient occlusion to better see the form of the head as I texture.

Again, I’m regularly swapping between shaded and shadeless viewports to make sure I’m getting a good clean evenly lit texture…


Here’s a quick render test with the Blender Render SSS preset for skin. Again, in my final render I’m going to use a more robust skin shader, but this provides a good quick preview…


And here’s my finished texture (for now…) Once I had a base layer down I then used the normal map as guidance to stencil and clone in skin from my reference images in a way that better followed the contours of Herzog’s head. Also note that while the head geometry is still symmetrically mirrored, I used a combination of mirrored and un-mirrored texturing to get a natural skin texture laying in larger blocks using mirroring, and then adjusting and adding without mirroring on top…


I’m going to come back to this skin texture a little later down the road, and use it to create some sub-dermal layers for a multi-layer SSS set up, but I’m happy with how this is looking right now. More to come…

Very cool work in progress! :slight_smile:

That’s all very nice, but does he Dream? Does he cry at night?

Seriously though, nice sculpt, and great subject, too! Werner is so interesting, and hilarious!

Thanks for the kind words Chadouin. And yes, SterlingRoth, I’d like to think he does. Ha!

A quick aside on the eyes. Honestly, I’m not getting a great render out of them yet, but I’m still rendering in the internal renderer for now. I think the textures are good, but we’ll see if they need additional work moving forward if I’m still not satisfied in cycles.

Credit to NJakStudio on Deviantart for the original iris texture, courtesy of eye “donator” Elizabeth. So I guess I’m giving Werner a girl’s eyes. Is there such a thing as a gender difference in eyes?

Using an outer shell that is purely transparent/glossy with the texture attached to an inner eye…