When these projects were started in September, there were a lot of ideas on how to approach the sequences, a lot changed throughout the process, techniques didn't work, easier ways were found and skills were developed. This blog post is a quick conclusion on each of the sequences, about what challenges they posed, and what would be done differently if they were to be attempted again.
Car Colour Correct (Still)
This was the very first sequence attempted, this is because on paper it was the simplest to do, and the skills learned from this piece, could be transferable to the pan piece. The most difficult part of this piece, was the windows and the shadow. The windows were difficult because of the method used, it was too complicated. It turns out that a simpler method was used in the pan piece to do the windows, which was the rotopaint method. The still piece does use rotopaint to give the windows a glare, but the method of actually getting the right image behind the windows was made much trickier than it had to be, in this piece the windows were roto'd, then transformed, then blurred and edge blurred before the rotopaint was even brought in. So if this were to be done again, the one thing that would definitely be changed is the method behind fixing the windows. As for the challenges that the shadow posed, again it was to do with the method used, in this one the shadow was roto'd separately to the car, where as in the pan shot, it was roto'd along with the car and rotopainted. The shadow in the still was rotopainted in the end to blur out the edges, but because it was roto'd separately, it needed its own colour correct node. So if it were to be done again, the shadow would be done in the same way it was in the car pan piece.
Car Colour Correction (Pan)
This sequence put forward a fair through problems, pretty much all of them do to with tracking. Through most of the process, keeping the car in place whilst the footage moves was constantly bringing up new challenges. Most of these could be solved through adjusting the tracking, but there is something wrong with the tracking, that makes the car jump ever so slightly, and the problem doesn't appear to want to be solved through adjusting, or created new, trackers. There is also some lens flare in the footage, which could explain why there is a problem with the cars tracking. If this piece were to be done again, the first port of call would be sorting out why the car jumps, and how to fix it. Apart from that, the sequence has turned out quite well.
The Altar Sequence
The altar sequence put forward many challenges, from rotos going across many frames to blue screens with annoying creases through them. Lets start with the rotoing. The rotoing was simple, but it was difficult to keep them accurate across lots of frames, sometimes trackers would work smoothly, but with the footage not being of the highest quality a lot of the time they didn't, so they couldn't be properly linked with the rotos. In the first and last shot the trackers did work properly, so they could be used with the rotos. Through personal error more than anything is why the rotos look inaccurate after that point. The blue screen shots in front of the dragon statues and the altar required hold out mattes around the characters, this was difficult because of how much the characters move throughout these shots, which makes tracking them accurately impossible. So the one of the main things that would be improved on if this piece were to be done again would be the rotoing and this could be achieved simply by putting aside more time to make sure that they are as accurate as possible to get the best results. The biggest problem was the blue screens. The crease going along the bottom of the blue screen was one of the most difficult because it meant that it needed its own keylight to sort out, on top of keylights for the dragons and the altar, and certain characters, it makes the shot very messy. It makes it difficult to get rid of the halo effect that happens around anything in front of the blue screen. Maybe it could have been done better with the IBK keyer node, or maybe with less keylights. Either way, the blue screen is the weakest part of this sequence.
Forest Fight Sequence
The forest fight was much simpler than the the altar sequence, the IBK keyer sorted out the blue screen very well, the only thing that proved annoying was the rotos. This is because the footage isn't very clear in many shots, and the object being roto'd is very small in certain frames, which makes it very difficult to keep the rotos accurate. Also since the roto shots require the rotos to stretch across lots of frames, which without the assistance of a reliable tracker, has to be done by hand, and as a result of that, the rotos are not nearly as accurate as they could be. So if this piece were to be done again, more time would be put aside to make sure the rotos are as accurate as possible.
So in conclusion, some of the problems that have been encountered during this process can be sorted once the necessary skills have been developed through practice with the Nuke software, there are a vast about of tools in Nuke that haven't been used during this process, and only by putting in the time and effort can these tools to put to full use and the true potential of Nuke can then be used to get the best results. Other problems, like the rotoing can be avoided in the future through better time management, which is something that will definitely get better by the time the next assessment is issued.
No comments:
Post a Comment