This has been a very busy week! I am not completly happy with the results and I think I will work on my hand some more before I am complelty satisfied with it. Unfortunally I messed up some of the shadowlayer frames when I rendered, so they are twitching a bit. However this is the result so far:
Here is the final ncloth test I did.
Now I have implated this into my final version. You can see the reslut down below. I have been rortoing and testig some grading with the handshot as well, but since I am not compelty content yet with how it looks this is a verison without them. This week I have to finish up the handshot work and render the other shots. Luckely I only have to render one frame of these other shots since the camera is still in all of them and the hand is static.
So, I have played around a little with the fissures. The original plan was to use a boolean operation and then animate its history. It worked great with a lowpoly mesh but not so great with a highpoly one. The procedure is the same, however animating the history becomes to much for my computer to handle and impossible to view in the render view. So the boolean plan is scrapped.
Instead I am currently turning my attention toward an Nclott solution with animated cluster deformers. This is a render with only cluster deformers. I need to redo the skinweights though.
I tried to do a version using only ncloth, but I can’t get it to work as well on the actual hand. Furthermore I like the cluster animation better as it divides the geometry much more precisely as it goes.
I have now tracked my hand and animated the fingers with joints. It was not easy.
In the end I chose to do a geometry track of the arm in pftrack, send the tracked joints to maya and add a jointchain for the hand and fingers. I then animated the fingers as accurate as I could with the clip playing for reference as an image plane in the background. This was my back-up plan but I think it will work just fine.
When I started I wanted so badly for the ordinary object tracking to work. I tried many things but both Matchmover and pftrack just turned up with red or wierd solves. Earlier in this project, before I filmed my material with the RED camera, I did manage to get a decent object track in pftrack of my arm (just the arm, no fingers!) filmed with two static cameras. I did the same procedure for this shot, but perhaps I just moved around too much this time. Or it may have been that thet one of the 7D cameras was a little out of focus, making the tracks glide to much. Either way, it´s nice to have a back-up plan!
Okay. So, this is a small render of some of the animation. The hand plate is visable behind the cg hand. Unfortunatley, you can see that sometimes the outlines of the fingers on the plate is not completley covered by the cg hand.
What I did here was to simply dilate the fingers in nuke. I came across this outline problem when I started to animate. My first thought was to simply give the sculpture some additional thickness. But, in order to completely remove the outlines (and unwanted motionblur) was to increase the thickness to the point were the Hand of Adam actually looked quite fat. Instead I will dilate the fingers and avoid to show any skin at the edges (only for the fingers). Perhaps there will only be smaller cracks at the fingers joints. I will see how it looks when I start on the fissures.
This is how far I have gotten with the texturing process. I think I will try to make it look like ruff limestone with some light cracks. I saw another sculpture at Nordanå which looked sililiar to this whom I thought had pretty nice texture. However, right now it looks like the cracks are the dark lines you can find in marble material, so I will rework it a little more again. I also want some more detail and definition on the stand texture. Now when I also have some idea of how the stone will look like, I will go back and adjust lightning a bit as well.
Here is what I been working on this weekened. Lightning tests and a slap composition. The lightning is not perfect yet and the slap comp’s rotos are of course not 100% but this is how the sculpture looks right now. My real hand is behind the cg-hand, having the same pose as the sculpture. I will proceed with animation and texturing this week so hopefully I will have a test render on this shot on sunday.
In the beginning of this project I planned to model and sculpt my hand from reference images alone. However since my neighbours, from the project “Sabotage” who both are sitting next to me, were exploring the capabilities of Agisoft Photoscan I could not keep my curiosity at bay. They took a series of photos 360 degrees around their deer and then created geometry of those in Photoscan. For my project, I did not have to have a perfect copy of my hand, especially since my cg hand is going to be slightly larger than the real one. But since I now know that I probobly have to animate the fingers in maya, I thought it would be nice to at least have the correct proportions. Therefore I did a few tests…and a couple more. In the end, I did not expect the scan to go as well as it did.
My first thought was to only use the scan as a proportion reference. The first tests turned out useless. Photoscan did not recognise the hand in the images. This was due to three main problems:
1: The skin. Photoscan calculates its point cloud like tracker points. My skin was simply to smooth without any features to track.
2: The background. Although the background was blurry and the hand took up most of the imagespace, it disturbed the calculations.
3: Hard not to move. If one is to take a 360 degree photo serie of a hand, the rest of the body should preferably not be in the picture. This means holding the hand away from the body and above the head. You also have to keep that pose for as long as it takes to photograph approximately 30 images. Even a slight move can create an extra thumb, which did happen.
With help from Johan Lilja, who took the pictures and helped with some of the problems, I ended up like this:
Strapped against a table with tape, having the armrest of a chair as support for stablilty, a grid litterally drawn on my hand by an ink penn, inside a bunk draped with the Universitys white curtains. Oh we had fun! The table was there so I could cover my head under it. Yes, it looked pretty strange.
In total we took 32 images around my hand with a 50mm focal lenght:
Still, this is the result from Photoscan after this session:
I was very happy with the results. The model was not perfect, but good enough to use as a base for the highpoly model. I took the model inside mudbox, smoothed it out, increased the polycount and made the fingernails a bit more precise. I then sent it into Topogun, baked the normals in Xnormal, and extended the arm in maya. This is how it looks right now:
You can also see the beginning of the stand/stone that the hand originates from.
Now, did it go faster with Photoscan rather than sculpting my on hand? In the end, mabey no. The time it took to test and solve the problems I encountered could have been spent on sculpting. If I knew how to do it right away then the scanning definitly would have won. Still, I am happy with the results, I got a great bake and I now have an exact cg-version of my hand in my computer. Cool!
Milestone 2 and presentation is now over and posted. Here is my camera material, edited in premier. I am a bit unsure about the final clip but it have to do for now. Something else I have noticed is that some of my clips have a line in the middle of the image with dead pixels. This is something I have to fix in post. Although I’m not so far behind on my schedule I am starting to have this overall feeling I have to work up my pace. I really want to have as much time I can get on the texturing part. There is alot to be done and get done. Here we go Milestone 3!
I have cut my movie in premier and started to track my selected shots. My plan was to use pftrack for my whole tracking process. However, for some reason I can’t seem to find a way to make a good or sensible solve even for just the scene and with the reference images. Therefore I tried to track in Autodesk Matchmover instead. It was still a hard track. The camera movement has very little paralax. But together with the reference frames I finally got a sensable solve. YES! Analyzing this I now know that reference frames definitely are a great help solving camera movements like mine. It would not have worked without them. Also, the reason I think i couldn’t get a good solve in Pftrack is becuase I just don’t understand how to make the constrains to work properly. I have to look this up. But for now I am happy it worked out well with Matchmover.
Something else I have tried out is the geometry tracking within Pftrack. And it works…sort of. From the start I had already taken into account that geomtrey tracking could not solve all of the animation on the hand, like the fingers for example as they constantlycover each other. Exporting the geo track from Pftrack results in an animated joint chain in maya. Now when I have experimented with it I think I will go with just keeping the joint which drives the forearm and then later in maya, proceed with building on that chain and add joints to the hand and fingers I myself will animate. The forearm joint will add more realism to the overall animation, providing small movements that would be hard for me to animate.
The geo track is a fast way to get some additional animation. However, from the start I wanted the object tracking to work, but once again the solve is not quite right. I am aware that I can only track features that are static to eachother. I have tried to track the forearm’s markers since there isn’t much twist movements from the arm. If I have time I will try this again, but for now it’s good to now the geo track is working.