As some of you may know, when it comes to rotoscoping, there are two core software solutions that studios integrate into the vfx pipeline. The first and probably most common being Nuke, used by studios such as Animal Logic and the other being SilhouetteFX, used by studios such as WETA VFX. So I thought that I would do a comparison of the two tools to establish which is the better software to learn (or simply to learn first)…

VS

Firstly to compare these programs a distinguished definition of what a roto/prep artist’s responsibilities are is necessary, so i have broken it down into several key points:

- 2D tracking
– Matte extraction
– Rig removal/paint fixes
– 3d camera projections
– Clean plate creation
– Spline based roto creation

So to compare these two packages I have broken down this comparison into 5 categories, layout, rotoscoping, tracking, keying, paint and stereoscopic workflow.

LAYOUT & WORKFLOW

When comparing the layout of Nuke and Silhouette there are two very obvious differences. One is node based (Nuke) and one is layer/node based (silhouette).Immediately this really makes Nuke a clear winner with just this fact alone as the more fundamentally node based a compositing-roto  package is, the less linear it is and there for the easier it is to trouble shoot and faster the program will be. However, in saying that, it would be unfair to immediately declare Nuke the winner based on this alone as we are not talking about these two programs in the context of compositing, but in the context of being a roto/prep artist which really changes the essence of what is important in the layout.

The first thing that you will notice when opening Nuke is its distinctive and simplistic layout, the beauty of this layout is that it can be customised anyway you want and is extremely maneuverable depending on your workflow. It also allows for layouts to be saved depending on what sort of tasks you are performing. However Nuke’s beauty can be a little cluttered and counterproductive without multiple screens, this is only a slight draw back though as you can expand a screen your using (whether its the viewer or node graph) by a simple hit of the space bar.

Now to compare this with Silhouette, on opening you everything is laid out neatly, when beginning a new project it will ask what particular nodes (or functions) you will be using for the course of the project (whether its keying, roto or paint ect.). This is a nice function and its also notable that a single screen setup for silhouette still works really well as everything is ergonomically placed. The down fall of the silhouette layout is its limited customization options which is understandable given its not used for anything other then roto and prep work.

Now for node ease of selection it Silhouette is instantly a clear winner as all the essential nodes for roto/prep are directly available without any need to search for them like in Nuke. This does increase work flow efficiency as you can quickly start using a node without searching for it in a drop down menu like Nuke’s. Of course in Nuke hot keys are available as does silhouette, however the difference here is that all of the essential nodes in silhouette have hot keys available, which is simply not the case in Nuke. However in Nuke it is possible to re-assign the default hot keys, but it is a painful process.

One huge down fall of Silhouette is the fact that you can’t import video files, unfortunately you are limited to only using sequences, now this was an issue with Nuke as well back in the day but they realised the need for video as well as sequences, something the silhouette hasn’t managed to modernize itself with.

Another aspect of silhouette’s layout that rises above that of Nuke’s is the ability to do colour defined roto/key quality checks (e.g. having a red overlay where the alpha is to check that the roto/key is covering whats intended for the duration of the shot) with ease. In the case of silhouette all that is required is the simple hit of a button from number pad to switch between these checks and views. I do realise that this exact thing can be done in Nuke and some would argue is better as you can define what number press effects each of these views, but the difference is this viewing method has to be set up in Nuke prior to being able to do this with the use of several nodes,

but in Silhouette it is ready at startup. However in saying all this, the huge downfall of Silhouette is that you can’t actually create custom channels, you are restricted to RGBA. Ofcourse most of the time that’s all you need, however it can be handy to do multi-channel roto passes when compositors have specific needs for your roto.

ROUND 1 VERDICT: NUKE

So in summary, although Silhouette has a lot of efficient and ergonomic designs for setting up and executing a roto comp. The fact that Nuke is so non-linear and customisable makes it a winner regardless of what hot keys and viewing options Silhouette has to offer.

ROTOSCOPING

Now to judge whats really the most important part of a software package for a roto artist: how well you can roto with it.

Let us start with the roto tools available in each of these two packages. Firstly in Nuke the roto menu has 4 shaping options: Bezier, B-spline, Ellipse and Rectangle. These four are really all you ever really ned for doing your shapes, however there really is no harm

done in having more. Which is exactly where Silhouette comes in where there is in fact an extra type (also found in after effects) called the X-spline. For those not familiar with this it is basically a more maneuverable version of the bspline, with every point you make being a transformable one.

Now when it comes to the actual roto node options for each program, there are some rather large differences, and by this I mean Nuke is so much more flexible and non-linear.  Now these differences start with the blending options for each program. Silhouette features the basic Add, Subtract, Multiply and Difference blending modes, and whilst this is generally all you need, it never hurts to have more options. Nuke on the other hand has the option to do all these, as well as the option to merge: atop, average, colour-burn, conjoiunt-over, copy, disjoint-over, divide, exclusion, from, geometric, hard-light, hypot, in, mask, matte, max, min, minus, out, overlay, plus, screen, soft-light, stencil, under and xor. Although most of these wouldn’t be used often for roto, it is good to have, and I find myself often merging with Stencils, Max’s and Disjoint-overs.

Next up with the roto options, where Silhouette has all the basic shaping modifiers like blur, erode, opacity, invert and channel options. Nuke in fact has all these plus foldering options, multiple layered transform options, shaping options and even the ability to determine the lifespan of a bezier without effecting its opacity.

Lastly the actual performing of the roto and animation of the roto: arguably the most important aspect of this whole review. However when it comes to this, the two programs work pretty much exactly the same, the major difference is with the timeline. In Nuke each key frame that is made for a roto shape appears as a blue mark on the frame in the timeline that the change occurs. In Silhouette it works exactly the same, however with Silhouette the time line shows the key frames for each separate roto shape as different layers, compared to it all being dumped together. This is really helpful allowing you to instantly see which key frames are associated with each different roto shape. Now of course this can be done in Nuke, but for this to be done you either need to have the dope sheet constantly open in some corner or you need to keep flicking between it and your node graph, neither of which are very efficient or ergonomic.

ROUND 2 VERDICT: NUKE

Although for the animating and drawing of shapes in each package is the same, nukes huge amount of modifiers and options make it a clear winner in this category.

KEYING

Regular joe’s often get confused when keying is mentioned for roto jobs, but as any roto artist would know: if you can’t key it THEN you roto it. There’s often times where you can key something out that will give you the smooth animation and edge fidelity that rotoing by hand or with trackers can never provide.

So lets start this comparison by looking at what different options we have for keying in each program. In Nuke, the main keying nodes (and I do say main as there are more) are:

• Keyer: which is the most basic keying node, featuring your different types like: green, blue and luminance.
• Key light: a very advanced keyer that is great for getting accurate edge fidelity, yet lacks in getting the alpha perfect across the whole key.
• Ultimatte: is a very advanced keyer that allows for maximum control without the need for multiple nodes, in fact I believe ultimate has the most options out of any node inside Nuke.
• Primmatte: The easiest and fastest keyer for good results. The beauty of primatte is its ability to add more than one colour to the key. You can basically keep picking more and more colours you want to remove and more colours you want to keep from you footage until you have the exact key your after. It is also very effective for marker removal work.

In Silhouette there are only two nodes for keying, these are:

• Keyer: very similar to that of Nukes “keyer” node and in fact has indistinguishable results.
• Power Matte: is a very cool, effective and simple gem that Silhouette has over Nuke. Silhouette’s matte chops don’t end there. Power Matte puts a little magic in your workflow. It’s similar to that of the Rotobrush in AfterFX’s, but using roto shapes for each area you want to keep and remove instead of painting the areas. However although its like AE’s Rotobrush it really can not be compared, Silhouette then automatically generates a high quality matte even on intricate details  such as hair and fur. What makes this keyer so cool is that you can key frame and animate the roto shapes (being used to key colours and luminance) as the shot goes on for a very effective result.

With all this being said it probably sounds like Nuke is a clear winner for its abundance in keying options, but a very important aspect of keying that has to be mentioned is the combination of keys. It is near impossible to get an effective key of something with only one type of key, even when its a blue screen being keyed it’s often the case that lighting will be different on throughout the blue screen

forming different shades of that blue and thus for the perfect alpha channel, different keys need to be combined.

With Nuke this is done through using different merges like the multiply to combine the alpha channels of each, in fact to get a really good key a lot of different nodes need to be used from time to time. This of course means that its not always fast and simple to get the matte that you are after. However with Silhouette it is very easy to combine keys (to an extent) as it has the option in the keying node for primary and secondary keys as you can see on the picture to the right.

In saying that, the problem with Silhouette is you are limited to exactly that: two passes of keying. Whilst its layout makes it fast and easy, this can be very limiting as you can simply continue to add more and more passes to you key until you have exactly what you are after. Further on this, Silhouette is also limited by the fact that you can’t choose the exact colour you want from your footage, your locked into the basic realms of its standard blue and green screen defaults.

ROUND 3 VERDICT: NUKE

Once again Nuke takes the cake for the same reasons: flexibility!

TRACKING

When performing a matte extraction the first thing to always keying, however when roto becomes necessary the next step is to try and make life easier in any way possible. That’s where these bad puppy’s come in!

2D Tracking:

Nuke has two core 2D trackers:
- Tracker: The tracker is Nukes default 2D multi-point tracker. It does all the basics like Stabilization, Matchmoving, jitter removal and channel controlled tracking.

- Planar Tracker: A plane based motion tracker that uses a roto shape you make to judge a plane and follow its movement. It can provide very accurate and jitter free results in those times when you just cant get a movement right. A very hit or miss tracker.

Silhouette also has the same two and has pretty much all the same features as described for Nuke above. However differences can be seen in the Planar Tracker, where Nukes tracker forces you to sit there and watch the tracked area it captures for each frame in the form of a point, Silhouettes actually allows you to see inside the  roto shape and indicates where areas are being tracked well (green) or could cause problems (red) like that of 3D camera tracker. It then allows you to disable certain points that are considered defective (red).

Putting these two trackers together on the same 26 frames of footage resulted in very similar results. However in saying that, Nuke seemed to handle the point tracker better and smoother (slightly) and had less random pops, also applied stabilization more effectively. But Silhouette masters the Planar tracker with better results more of the time. Although in Nukes planar tracker a separate tracker node can be generated, it still doesn’t compare to Silhouettes often flawless tracks.

3D Tracking:

To a roto artist, having access to camera’s, axis/locators is invaluable, it can literally allow you to do one frame of roto for a very complex large duration stereo shot and have the whole thing animate and correlate perfectly.

This is the biggest reason why Nuke is better than Silhouette:

- Nuke allows you to import camera’s tracked in different programs

- Nuke allows you to perform 3D tracks in the program itself

- Nuke allows you to import and create axis points (Points to 3d)

- Nuke allows you to project roto and paint onto cards

vs.

- Silhouette can not work in 3D space at all or use any cameras.

For such a large and helpful tool to be missing from a roto software package is beyond me but that is why…

FINAL ROUND VERDICT: NUKE

This post is still a work in progress, to be continued…

———————————————————————————————————-
Driven Shoot
Earlier in the year in June I worked on a shoot as a Focus Puller (1st Assistant Camera).
The shoot was directed by none other than Carlo Ledesma, who is known for his recent feature film The Tunnel (2011). On top of that two trucks full of lighting equipment  were also used by a top of the range Gaffer.

Sidat de Silva was the DoP I was working under that I had previously worked on Scratch with as a 2nd AC. The role of the Focus Puller is (goes without saying) to pull focus on the lens. Now this may sound like a simple task, however as characters move in and out of shot the focal length changes and adjustments to the focus must be made. I would explain this job as one of the simplest yet hardest things to do on a set. It was great to finally get one step closer to a camera and focus pulling is something I have been keen on for a while.

The shoot itself seemed like Drive with the way the actors looked and how the lighting was done. One scene consisted of a women’s ride in a taxi where her conversation with the driver centres on life and relationships.

Now of course the car wasn’t actually driving. Inside the studio, blacks were put up and crazy car simulation lighting was used. This included things such as a rotating set of mirrors that had light bouncing off it in different colours and directions, also a metal wheeled rig that had two head lights on it so it could be dragged back and forth behind the taxi to simulate a car tailing. Also, there was a stage hand that would shake the car from the bonnet to simulate movement. Ultimately it looked amazing and the money spent on lights was well worth it.
I have a link below for you to really see what I am refering to.

So its been a while since posting an update and thats because its been a while since having something worth posting. Reason being is that now I work at Animal Logic there is very little I can say and nothing else film wise I am really doing. Not to say that I am not having a great time, but just that I cant necessarily share my experiences to such a level.

There are things I can say though…

So I have now passed my probation period at Animal Logic as a Rotoscoper. As there is little I can say about what I am actually working on, I thought I would go into a bit of detail about what exactly it is I do as a Rotoscoper.

ro·to·scope/ˈrōtəˌskōp/ Definition

Transfer (an image from live action film) into another film sequence using a rotoscope.

I am guessing that definition doesnt really make much sense, but basically my job is to get a scene ready for the visual effects. The same in which a assistant editor will get the shots ready for an Editor by cutting out all of the unnecessary bits and deleting the accidental bits of footage.

The best way for me to explain is with this example

This shot is a of a man in a fighter jet before all of the CG and lighting elements come into play. Now the first step for getting a shot like this ready is chroma keying the green screen out. This obviously does half the work for me as I dont need to rotoscope (cut) out the whole actor and prop from scratch (like I would if there was not a green screen there). Following the keying I would need to cut out the camera and the crew member that is in place, which is easy in a shot like this.

However its often not this simple, for example a shot I have come across involved a crew member administering car simulating to a prop car against a green screen, to make it appear that the character in the car was actually driving. The problem thats faced with this shot is that the crew member can be seen in the reflection of the side of the car, then meaning that all of her reflection must be painted out in order for it to be ready for the visual effects to be applied. Painting things out of a shot is an extremely laborious proccess as it often needs to be done frame by frame (25 frames to a second of film) and then of course The Great Gatsby being a Stereoscopic (3D) film, it needs to be done twice, once for each eye.

Another thing that needs to be done to the shot of this fighter jet is the tracker markers surrounding him must be painted and roto’d out. Once again a lengthy process, especially when camera moves and focus changes occur. Now I am yet to fully start on marker removal on shots for the Great Gatsby but through others experiences on other projects, a simple 5 second shot can take over two weeks.
People have often said to me: “surely your job must be made a lot easier because of the improvement of technology and software” which is completely true. The power of the tools I have makes what I do incredibly easy compared to how it use to be, with colour keyers becoming more powerful and workflows becoming more innovative it makes a lot of what I do appear easy. However with improved technology also comes stereoscopic (3D) films which makes my job hugely more difficult. It essentially means all of the work needs to be done twice as the second camera used for the second eye on 3D films is not as simple as being slightly to the right. The problem is perspective shift, the actors and props appear on a slightly different angle to on the first camera as well as the depth of field being slightly different. For example if a tracker marker is covered by an actors head completly in the left eye camera, that same marker you didn’t have to roto out before may need to be roto’d in the right eye camera.
This is probably all a lot to take in, but I hope it gives a little bit of understanding about what I do as its a difficult concept to understand.
On another note, a music video I worked on a few months ago as 2nd Assistant Camera is now on youtube, enjoy!