fredag 12. juni 2009

And the movie is complete

And here is a link to a lower quality youtube version.

Last words


So, over the project I did learn a few things.

Hopfully I am a better animator then I was before I started. I am much more confident with splining. I have also come to the conclusion that I need to learn more about rigging, because I feel I could have done a better job skinning with more knowledge on how rigging works.

I have never really given much credit to animatics, but I must say it did help out a fair bit to have it. Seeing the entire movie at such an early stage made it easier to spot things that should be changed.

I knew from previous animations that having good reference was vital, and this project have proven that even more. The reference videos that we filmed at the started helped a lot.

I will probably render in pass from now on as well. After all the extra work we had to do in post because we didnt render in pass it seems to be the only way. A plusside with all the extra work in post was that I got better at using After Effects, which good.

If I had known what I know now about making this short I would probably try to be on a smaller group. We did have some communication issues in the start. And on a smaller group I feel it would be easier to get a common opinion on how the movie should be. I also feel it would be easier to get stuff done. A large group does obviously have more workpower, but a smaller group is probably easier to work with, and on. Making it more effective.

I would also most likely animate out hero shots first, just to safe it incase something happens.

As for rendering I would, sa said above, render more in passes. I would also try to start rendering at a much earlier stage then what we did now. Maybe as early as the instant the animation is done. Just to get more room for error.

Not sure what I would do if we had more time on this project. If we suddenly got an extra 2 weeks at the end I would most likely have rerendered a lot of shots in passes, to make it easier to work with in post. I'm not sure if things would have been so much differently if we had been given an extra two weeks in the start. We would probably have had a little bit more time, but in the end we would have made the same mistakes. But maybe we would have had time to correct them.


As for inspiration with the movie we did look a bit at how Meet the Robinsons looked like, mainly for buildings. And we looked a bit at The 5th Element for ideas for the apartment. I also found some random pictures that was inspiring.

Liked the dirty look for this, and wanted something similar for the apartment.
We wanted something big for the guard design. And the middle guy in this picture was something we looked at.
We wanted something similar to the above picture for JDs look.

torsdag 11. juni 2009

Post Production

We figured out that we could just do post on our own shots as they came out from render. So I did post on the clips from where he enters the door, and up to the point where he reaches for the handle by the window to close it to block out the sign.

In post I mainly color corrected and adjusted levels to match up the color and light in the different shots. Most of the time I could use the same adjustment layer for all the shots, since the lighting was the same. But at some points in the movie I had small differences in the lighting. For my part I had some major differences in light between shot 1 and 2.

And this is some quick stills in sequence showing the different layers going from scratch -> Gaussian Blur -> Ambient Pass -> levels -> hue/saturation -> curves -> background images.




We did have an issue when JD was walking past windows. We had forgotten to remove the environment map from the Daylight System. So when JD was passing by the windows where we thought we had an alpha channel, we got an odd Alpha Channel where JD had blurry, white edges around him.

At first my teacher suggested adding a Luma Key to the layer, which would take away the white from the picture sequence. The problem with that was that the white was removed from other parts of the pictures as well. So it didn’t look to good. What we did then was to duplicate the layer. On the layer with Luma key on it I had a Mask with Add covering the window. This would then add the Luma Key to only the window area. On the other layer I would have a copy of the first mask with Subtract, so take away the window area. The Luma key then toned down the white lines around JD, making the much less visible. A new problem that emerged was that JDs shirt is white. And lost some color to due to the Luma Key. To solve this I added a new mask with Add (on the layer without Luma Key) to Add that layer on top of the Luma Key one. I had to animate this mask to follow JDs shoulder as he walked past the window. And this was one timeconsuming and extremely slow process. I think the end result was pretty good, especially when I look at the difference between the 2 images.



The first picture is how it looked fresh from 3Ds Max. The 2nd one is with Luma key on the entire picture. The 3rd one is with Luma Key masked out in the window. But that left some black spots on his shirt (most visible on the 2nd last picture). So to solve that I had to animate a mask around his shoulder as he passed the window.

I also did some masking on the final shots (when he comes home from the doctor). We made a huge mistake here when we didn’t render in different passes. So when I tried to fade the apartment from being good looking, to being bad looking I had JD on the same pass as the rest of the apartment. This made fading the apartment to ugly, without JD being faded as well, very hard. The first few shots was ok to work with, since JD wasn’t standing in an area that I faded to ugly, but when I started trying to fade the later shots I had JD in the middle of the shot. Masking out around JD turned out to be way to time consuming, so I couldn’t do that (would probably have been faster to do a new render in pass instead of masking him out). In the end I ended up having to include JD in the mask fading to ugly, so JD became ugly as well. This was not our original intention, but if I was to skip the entire fade to ugly sequence we felt that parts of the story were “lost”. And in the end we considered the story to be more important than the looks.

If I were to do this again I would defiantly render in more passes. Having JD on one, background on one and maybe one for shadows. We would have had it much easier if we had rendered in more passes than Ambient and Color. Especially the fading from good looking to bad looking apartment.

Lighting

So, we had a lighting that looked awesome, but due to render time being insane we had to scrap it. We also needed two different light settings. One for when he comes home the first time and everything is supposed to look like crap. And one for when he comes home after being at the doctor and everything is going to look nice. I sat down with the lighting that was supposed to look bad. And this was surprisingly hard. I have never really done a lot of lighting in the first place, so this was challenging. Having the lighting look like it was dark outside, and at the same time have enough light to see what was happening was hard. Also, having the light make the apartment look bad was also hard to pull off.

I tried out a few different options:
Daylight with FG looked very nice, and was fast to set up, but the rendering time was to much. It also looked to “clean” and light in away. Didn’t manage to make it look “bad”.

Daylight with skyportals looked nice as well, but again, rendering time was high.

Lightrigger was a script that I tested a little. It was fast to set up, but we had issues with shadows, where we had to exclude the walls and roof of the apartment to get light into the apartment, but when we did so those objects didn’t receive any shadows either, so it didn’t look so good. Did spend some time tweaking on this, and my teacher helped out a lot, but in the end we had to scrap this idea as well.

Directional Light with Omni Lights as fillers was the idea we ran with to the end. I sat up a direct light outside the window to mimic the moon/sun light coming in through the window. This was also supposed to be the primary light source for the scene. In the end this light didn’t provide much light into the apartment, so I ended up having to use Omni Lights to fill out a large portion of the apartment with light. At first I positioned Omni Lights under the roof to mimic some ceiling lights. This looked better, but the corners of the apartment was still too dark, so I had to position some more Omnis there as well to not get huge shaded areas.

I later added some very strong Spot Lights in the hallway outside his apartment. I wanted the lighting in that hallway to look like it was some sort of industrial lighting just tossed in there because it was cheap. So I set the strength of the lights up a lot. What I noticed a bit to late was that it looked really odd with the strong light flowing in from the door, while the outside was dusky. As if the light from the hallway was coming from a strong sun, but the outside light was coming from a much weaker sun.

Overall I had huge issues getting the shadows to look good, they were way to sharp, and we never really found a solution to that problem, so in the end I ended up using Shadow Map on one Omni, and turning up the sample range to try and soften the edges. This did give an acceptable result.

We also rendered out an Ambient pass to give more contact shadow all over the movie. And this made the shadows and lighting look much better.

I saddly dont have any testrenders of the different lightsettings, mainly because I just did a render, and rerendered when I wasnt happy with the result. So I didnt save anything. So this great wall of text will have to do.

Animation

So, we started out recording some reference clips for our animation so we could have something real to help us whilst animating. We basically just filmed ourselves doing something similar to what JD was supposed to do in the movie. At this point we were working mainly from the animatic. Once we had filmed the reference we sat down and started animating.

We divided the movie into parts that we then handed out to the people on the group that was most interested in animation. And then, whilst we were animating, the other group members would model more props for the apartment, texture props and characters, set light and so forth.


I was given shot 1-9 (as we had assigned them). This was the entire sequence up until JD is stunned by the sign. We were all more or less guessing how much we would be able to animate. We also assigned the shots fairly loosely, knowing that we could easily transfer shots should one of us be overworked. I had assumed I would be able to animate more then what was realistic, especially with the changes made to shot 3 that extended it by a lot. Luckily Daniel had time left over after having finished his animation, so he stepped in and animated a few shots.

When we started animating we had two options. We could either animate the shots as they appear in the move (i.e. shot 1, 2, 3…). Or we could animate the “hero” shots first, so that the most important shots were guaranteed to be done, in case we ran out of time. We were fairly confident that we would have time to animate everything, so we animated the stuff as they appear.


So, first off I started with blocking out the door entry shot. Here is a clip of the process I used. At first you see the original animatic, then a rough block, and then a splined out version. The final version is in the final movie.



As you may see I start out with a very rough blocking of the entire animation. I also tried to time this block as good as possible. At this stage I was also copying the reference video quite a lot. Once I was happy with how the blocking looked (timing, poses) I tried to get feedback on it from the people I was working with, from my class mates, and from my teacher. After the feedback I generally ended up changing a fair bit of the animation. Once that was done I splined it out and went over the entire animation a few times, cleaning it up. I usually ended up changing and retiming some (or a lot of) poses. And once that was done I went on a new feedback round and made some more changes, before the final result.

We did a few changes to this shot from how it was in the storyboard and animatic. We ended up removing the whole cloth rack that he was pulling down from the ceiling in the animatic (it’s the invisible thing he pulls down with his hand before sliding over to the kitchen area). Reason for this was that we felt that the broken lights, and later, the broken TV, would show that he was living in a dump. And we felt we wouldn’t need more clues for that. We also decided to cut the walk cycle over to the kitchen area, primarily to save time, but also because it felt a bit boring to look at it.


The video above should show my work process somewhat accurately. And I stuck with this process for the remainder of the movie. I won’t be posting the work process (from animation ref -> block -> spline -> final result) as a video for every single shot I’ve done. Reason for that is that I sadly had to delete a lot of previews I took while I was animating (ran out of HD space).


Had to split this in two since a single file got to big to upload. So first is the animatic, second is the animation.


For this shot I started out as in the previous one, blocking out whilst following the reference video. One problem here was that the reference video wasn’t really any good. I had filmed from an odd camera angle, so a lot of the movement JD does inside the refrigerator was hidden. I figured I could just improvise what he did inside the refrigerator, and I think it turned out pretty good. Due to the camera angle in the reference movie being odd I had to change out a lot of the poses to make it look good in the movie, but that wasn’t too much of a challenge. The biggest issue I had for this shot was the interaction between JD and the food and beer can. I didn’t want the mesh from the objects to go into each other, but at the same time it had to look like he was actually touching it. I didn’t come up with an easy fix solution to the object interaction other then keying a lot and making sure the objects and JDs hands looked good. This was only an issue at the start of the shot, inside the refrigerator, where the camera was close to the hands. Later on it wasn’t visible, so didn’t need to animate it that closely.


When I animated the walkcycle in this shot I started out with a very simple 2 pose cycle (contacts and passing poses). I did it primarily to have less poses (and less keys) to work with when I was timing out the steps. Once the walkcycle was timed I started refining it, adding ups and downs. I didn’t start animating the arms until I had timed out the entire cycle. Reason for that was I wanted to know how much time I had to open the beer can and drink from it. I decided to shoot the beercan opening from the back, and using arm movement to make it clear what he was doing. Reason for that was I didn’t want to spend a lot of time fiddling with the fingers. After that I tried out a few different timings for the drinking part of the cycle, seeing when I should start and stop.



When I started animating this shot he was originally going to just walk over to the chair, sit down and turn on the TV with a remote. After thinking a little about it we figured out that he could just as well turn on the TV, on the TV. So we didn’t have to model a remote control. Once I started animating I got the idea of having the TV be slightly broken as well, so we could show that he didn’t have state of the art stuff in his house. Once that was done I wanted to show that he was happy for a short while (when he sees his favorite TV show is on), since he had to be sad again when the commercial started. Going from happy -> sad seemed better than going from sad -> sad. All in all this shot ended up quite a lot longer than we had planned at first, and it also took more time to animate it, since there was a lot more happening.


The hardest part with this shot was making him look happy. At first I had him just raise and lower his arms and upper boddy a bit, like a very toned down jumping. But that was way to toned down. So I ended up making him jump up and down on one spot, before walking backwards to his chair. Eventually we came to the conclusion that he should probably jump backwards to his chair. It did look so good with the walking.


I was a bit unsure about the jumping thing making him appear very childish, but my group members liked it, so I went with it. Animating the jump was a bit complicated. At first I just animated it without reference, but that didn't go so well. So I quickly filmed some reference videos, and it went a lot easier from there.




In this shot I had most issues with him standing up. Originally I wanted him to push himself tiredly up from the chair, instead of just getting up normally. It turned out that it was fairly hard to get him look tired, but after a bit of time spent tweaking I think I nailed it.


We made one change here as well. In the storyboards he tossed a pillow at the window in front of the screen almost at the same time as he gets up. This was changed to the beercan in the animatic. Which he throw at the sign much later. The reason for tossing the beercan instead of the pillow was mainly because we wanted to use something JD already had in his hands. Making it easier to animate and work with. And we also had to model and texture one object less.





In this shot the major challenge was the throw itself. I started out with a baseball throw as a reference, but that turned out to be to sporty looking. And I didn’t want JD to look that sporty. So I found some reference on youtube of a girly throw (the term, not implying that there is any difference between the two genders at all). The problem I had with that throw was that it was way too little movement in his feet and upper body. So in the end I ended up with a toned down baseball throw with some influence from the girly throw.




The hardest part in this shot was having him look like he was burning his hands on the hot food, and him shaking his hands as if they were burned while rushing to the sink to get them under cold water.


For the burning his hands part of the animation I tried to copy the reference video as I had done for the previous shots. It didn’t really look that good, but it gave me some base poses that I could build on. In the reference video the actor was more or less holding the food in his hand constantly. I didn’t really think that was realistic if the food was burning hot. So I ended up having him throw the food in between his two hands (that’s what I usually end up doing when I take hot bread or something out of the oven).


Animating the hand movement when he was rushing to the sink was somewhat complicated. It took some trial and error to get the speed of the hand movement right. If it was too fast the motion blur would make it look like one big blur, and to slow would look…slow.

In the storyboard he had some curtains that he pulled in front of the window. In the animatic we changed it to a pull down shader. And in the final movie it was again changed to a nuclear panic lockdown/shutdown mechanism.

Skinning

Just been working on skinning lately. Skinning JD went pretty smooth. The mesh I got from Martin was really well made, and easy to work with, but I still had problems at a few areas. Elbows and knees were somewhat easy to fix with a Joint Area Deformer, and moving the vertexes around as needed at the max bend for each specific joint.

The shoulder proved to be harder to skin. The angle deform didn’t work there due to the shoulder being a bullet joint, and not a hinge joint. So I added two additional bones per shoulder with squash and stretch on them to make sure the vertexes didn’t dig into the mesh when the shoulder was animated.



The rig used is Puppetshop, which is a free 3Ds Max plugin. Its aviable at http://www.lumonix.net/puppetshop.html

The hips also proved difficult. And I can't say I was 100% happy with the result there, but it was good enough for the animation and poses we intended to do. So there was no reason to spend a lot of time skinning them perfectly. And it did work out in the movie.

tirsdag 24. februar 2009

Been few pictures so far, so will post a few now. We are more or less done with modelling props for the apartment, and are almost done with character modeling.

First picture is just a quick overview of the kitchen area of the apartment. The props on the desk was just tossed in there to get the size right. We have all made a few of these props each.

This is just a render of the bed and the entry door to the apartment.
And this is a test render of the exterior, wher our main character will be living in the building nearest the cam.

And here is our main character, the great John Doe.

Here we have a quick 3 part animatic of our movie.
Frist part was made by me.