In Render Settings >> Quality >> Adaptive Sampling, increase the max samples level. As you go up in sample levels, the render time will increase but the lines will be cleaner and smoother.
If the Anti-aliasing contrast is lowered, it will get closer towards the higher limit of samples per pixel. If it is increased the bias will go towards the lower limit of samples per pixel.
This determines how pixels get blended together and how many pixels are used for this process. You can use this if you need even more anti-aliasing. The edges will get really soft and blurry so be careful.
Just turn on Interactive Photorealistic Renderer (IPR). There’s a button for it or it can be turned on in the menu. Choose a smaller, relevant render region so that things don’t slow down too much.
This can be done in Render Settings. In the drop down menu, go to Presets >> Save Settings as Presets. There is also an option to load presets later.
You can also save materials as presets! However these presets will not reflect nodes connected to the material (e.g. bump maps, file textures), only the attributes directly applicable to the preset itself: color, transparency, ambient color, incandescence.
Sampling Mode: Adaptive Sampling
Max Sample Level: 2
Multi-Pixel Filtering: Triangle.
Filter Size: 2 and 2
Final Gathering turned on under Indirect Lighting tab.
Use the blue sphere icon to turn on the resolution gate. This will be right on top of the viewport. Also View >> Camera Settings >> Resolution Gate. This option will show you exactly what will be shown in the render.
Select the material. To export all the connected nodes, not just the child nodes, click on the input/output connections so that it also shows the shading group. Select the shading group, go to File and then Export Selected Network. This way you can import and export shading networks.
In Settings and Preferences, go to Display>>Performance. Increase the Max. res for swatches.
If part of the object disappears it could be that the camera is too far in relation to the object and information is lost. This has to do with clipping planes. You’ll need to adjust the near clip plane and the far clip pane.
You can set the file prefix, folder names (using slashes) and then insert a variable in the file name, e.g. scene name, camera name, date and more. This is accessible by right-clicking in the file name prefix field.
It’s best to use images (e.g. tiff or tga) instead of video files because supposing the render crashes in the middle it’ll still have saved all the rendered frames as images and you can pick up where it left off. With videos, you would have to restart the render from the beginning.
Change the Frame/Animation ext. For example, name.#.ext. This format works with most other applications as well.
Frame padding will change for example 1 to 01 or 001.
You can set the start and end frames under Frame Range. You can also get it to render every X number of frames by entering X into the field By Frame.
This will render according to your render settings but you can continue to work in Maya, although it obviously won’t be as fast. This is under Render >> Batch Render. You can also cancel the batch render from the same menu.
This can be done under Renderable Cameras under the Common tab of the Render Settings. You can add more than one renderable camera but clicking on add renderable camera in the dropdown menu. Maya should then create a separate folder for each camera. Remember you can set a variable for that camera in the file name as well.
Colour grading/correction is something that I’ve found every shot needs. The deer footage that I previously used for rotoscoping was pretty washed out so in the spirit of recycling I decided to use the same video for colour correction. I wanted to make the skies as blue as blue can be and prove that grass really can be greener.
And so, with a little too much excitement, I dropped in a HueKeyer node and connected it to my footage. There are two curves, the amount curve which determines the colour to use, and the saturation threshold. Using command+shift I sampled the colour of the sky and the yellow line in the hue keyer shifted indicating the colour I was sampling. I switched to the alpha channel, then dragged the amount curve until the sky was black and everything else was white. I had to use the scrollwheel of my mouse to make the graph bigger because I couldn’t get a solid key otherwise. To do this the colour I want to change actually has to be white and everything else black so I checked off the invert box. I switched to the matte overlay (hotkey m) to see how things were going in glorious colour since the gradation of the sky meant that my whites weren’t exactly white in the lighter areas. At first I thought that it might just work but it didn’t. When I added my colour grade it was quite splotchy in the areas where the sky was lighter. I went back to the hue keyer and mouseovered the lighter parts of the sky and noticed that it was indeed outside my curve range. I then adjusted the curve to account for all parts of the sky.
I laid down an additional roto for my sky and set it as an alpha. I then added a ChannelMerge connecting my sky roto with my hue keyer. I outputted the ChannelMerge as a new layer, sky.mt, copied the channel and connected it back to my main roto. At long last I was in a position to make colour changes — and who doesn’t like playing with colour?
There’s more than one ways to do this in Nuke, e.g. the ColorCorrect and the Grade nodes. Although the ColorCorrect node has many more options, I prefer the simplicity of the Grade node and tend to get much better results. So I added a much-awaited Grade node, using the sky.mt mask so that only that part would be affected by colour changes. Adjusting the gamma, I took away the reds and some of the green to make the sky more blue.
I repeated the process for the hilly area, making it more green. However, I noticed that the deer’s hooves had become rather green too. I wondered if there was any way I could cut out the deer roto that I had previously done from the hill roto. Sure enough, I was able to do this by outputting the deer roto as a mask, deer.mt, doing a ChannelMerge between the hill.mt and the deer.mt and in the same node. I then added a grade node using the newly created HillWithoutDeer.mt as a mask, once again adjusted the colours and used the D key to disable and enable the grade node so that I could assess the difference.
My node graph is below:
I’ve worked with beziers in Illustrator for years so the concept of Rotoscoping wasn’t completely alien when I first started out. I began with the exciting process of scouting Vimeo for rotoscoping reels to see how things were done so that I could do this not just for the purposes of my university assessment but for my demo reel as well. Not to piss off PETA or anything but as it happens I am a big fan of the ‘two birds, one stone’ philosophy. After watching a few dozen demo reels and performing the obligatory ritual of oohs, aahhs and popcorn, I set out to work on my own.
Based on my boyfriend’s fallacious promises that I would acquire inordinate amounts of footage for rotoscoping, I conceded to his unrelenting requests for long, hilly walks and resultant muddy shoes. We headed out to the middle of nowhere (also known as Welton Dale), blissfully unaware that there exist such things as wellies. There was no wildlife to be seen. I take that back. I came closer to nature than I ever wanted with the full force half eaten birds and unfortunate rodents, which luckily for everyone involved I decided not to film.
What we did manage to spot was a murder of crows (most foul) sitting on a tree. My boyfriend succeeded in making them fly off with a few rather eyebrow-raising screams while I filmed (fervidly). Once we reached the town there was a beautiful little church with ducks who were very much alive and more than happy to humour my camera. But at the end of this rather laborious afternoon and much to my dismay, my Flip HD camera turned out to be not-so-HD and all the footage of ducks and crows and a car that almost ran me over now lies somewhere in my hard drive waiting to be forgotten.
Having decided better quality was more important than blurry original footage for the purposes of rotoscoping I scrambled to find public domain videos online and downloaded a few gigs worth only just staying within my ridiculous 40GB bandwidth limit.
I worked on a couple of pieces, starting out easy with a flower with basic up/down camera movements, then a lizard that I chose because it was very well camouflaged and therefore more challenging, and then a fast-paced deer galloping across a hill (in hindsight I could’ve just taped myself clamouring around Welton Dale — that would’ve been sufficiently hilarious).
Using Quicktime Pro, I cropped each of the videos to the exact frames that I wanted and then read each of them into Nuke. I brought in a roto node and attached it to the footage after which I started laying down beziers. Using tracking points and offsets I tracked forward and backward until the the entire footage was tracked. With that out of the way, I went back and manually adjusted the beziers in each frame until I was left with very little hair.
Refering back to the roto reels I had watched, there seemed to be a general theme. First, an unedited clip was shown, then a matte overlay and finally black and white alphas.
I know that to turn matte overlays on and off the hotkey is M. However I needed to figure out a way to “bake” them into the video. To do this I connected my roto node/tracker into a ColorCorrect node and changed the colours. I also lowered the opacity so that the roto object would be visible underneath.
It took me a little while longer to figure out the black and white alphas. While my settings appeared to be correct it either was turning everything white instead of just what was within the beziers or offsetting everything to the side. After testing one thing after the other I finally got it to work by changing the FORMAT of each of my roto nodes. If there ever was a time to say hallejuah this was it. I dutifuly broke out into a croaky rendition of Rufus Wainwright.
Update, March 16: I’ve now learned more about advanced camera tracking and will experiment with that in my next bout of rotoscoping. For now, I’m working on greenscreen keying and compositing render passes.
The final output:
As part of my Visual Effects assessment I’ll be creating a train and integrating it with live action footage. I’ve done some research about similar work below.
From the movie Transformers. Some interesting camera movements, smoke and the kind of environment I’m trying to go for.
This is a piece by a masters student at Bournemouth. The lighting is excellent. I would make the train a little more gritty though. Never seen a train so clean!
Here’s another CG train. I like the camera shake at the end.