RSS icon Bullet (black)
  • Grading The Graders, Part 1

    Posted on October 31st, 2010 Mike 8 comments

    There seems to be a lot of talk these days about color grading, and a lot of it seems to revolve around what system represents the best solution given a particular set of circumstances. In a lot of these discussions, there seems to be a lot of attention paid to using general purpose platforms, such as nonlinear editors like Final Cut, Avid Media Composer, and Premiere Pro, for this purpose, or using software written for the purpose of file format conversion, like Red’s Redcine-X, for final creative color grading. And while for some this may present a very inexpensive (i.e., essentially free) solution on certain personal projects, in the professional world it is generally not practical or desirable. And it is not pricing that determines these things, rather, it is a combination of factors that together fulfill the practical needs of a professional colorist.

    For any professional colorist, the toolset they use must provide reliability, flexibility, quality, and the ability to assist the colorist in their most common tasks. It must provide a hardware interface that has as many parameters as possible mapped directly to controls on a dedicated control surface. It must be able to help the colorist to achieve what is desired or needed quickly and efficiently. It must allow the colorist to access previous material as both a source of grades and as a visual reference, and do it almost instantaneously, even if the material to be compared is from a previous project. It must provide real time, in-context feedback because anything less than that is not indicative of the final product. And finally, it must provide all of these things in a format that allows for very quick interaction and simple modification, because client supervised grading sessions demand speed, accuracy, and artistry, all in equal doses. For these reasons, I won’t be including things like Final Cut, Avid Media Composer, Premiere Pro, After Effects, Redcine-X, or, quite frankly, Apple Color in this discussion because all of them fail on at least one or more of those basic requirements for a professional grading system. I also won’t be including things like the DVS Clipster because it is not a dedicated grading system, even though it can be used for basic color work. Nor we will discuss Avid DS or Autodesk Smoke, because even though both of them have some pretty advanced color toolsets (particularly Smoke), they are, at their heart, conforming and finishing systems, not dedicated color graders. And while it is certainly not unheard of to do color grading on a Smoke or a DS, even for an entire show, it is also not particularly efficient to do so if every shot requires its own grade. And I say this as a colorist who also does quite a bit of compositing work on a Smoke system, which I enjoy using, but would certainly not be my first choice for a grading job.

    To keep things manageable, I’m going to go through some pros and cons of some of the more popular professional quality grading systems, including in no particular order, DaVinci 2K, DaVinci Resolve, Autodesk’s Lustre, Filmlight’s Baselight, Digital Vision’s Film Master, Quantel’s Pablo, Assimilate’s Scratch, and Iridas’ Speedgrade. Now, I know a lot of you will bemoan the fact that I haven’t included some other systems that might be of interest, such as Mistika, Piranha, or the upcoming Foundry Storm. But since this is my blog, and since I work in Los Angeles, I’m going to stick to the platforms that have a presence in the market I know, which also happens to be the primary world location for this type of work. Before getting into specifics, I’d like to talk about a few things all of these systems have in common. First, and foremost, I’m going to assume for the purposes of this post that all of them are being operated by a professional, experienced colorist. The colorist is the single most important feature of any grading room, regardless of the particular system they’re running. In the right hands, any and all of these systems are capable of turning out stellar work, and in a reasonably efficient manner. Quality is not an issue, as they are all capable of extremely high quality results. In many ways, all of these systems essentially do exactly the same things. They are all capable of basic grading as well as isolated grading of specific areas, that are qualified by either color uniqueness, brightness, or mattes supplied by a compositor or hand drawn by the colorist. They can all work for the most part in real time. They can all store grades for later retrieval. Beyond that lies the specifics, and that’s what I’m going to concentrate on here. So let’s get started.

    DaVinci 2K

    Unique among all of the devices I’m listing here, the 2K is a real time color corrector that is hardware based. It was designed to pass video, in particular, HD video, in real time and implement its corrections in real time. There is no rendering involved at any point in the process. There is also no storage involved as it is designed to work with a video stream that is fed into the device, along with a timeline that is driven by SMPTE time code that is fed with the video. This is usually done from a videotape machine, but can also be done with a digital device, like a DVS Clipster or a SpectSoft RAVE HD, that can play real time HD video and be controlled with a serial VTR control protocol. The 2K is still used for a lot of television programs due to its real time nature and its large pool of trained operators, particularly in the Los Angeles area.

    PROS:

    1. No ingest needed (takes live video stream from any video source).

    2. Large pool of trained talent.

    3. No rendering required (outputs live video and controls a record VTR).

    4. Real time defocus with option board.

    5. Separate luminance channel allows for some unique effects to be done very cleanly.

    CONS:

    1. Cannot work with file based formats directly (although one can use a DVS Clipster or similar device to feed a file based stream as live video to a 2K).

    2. Resizing of the image is not possible.

    3. Random access of source material is only possible with a disk based source device, such as a Clipster, RAVE HD, or other digital disk recorder.

    4. Material must be played, in real time, through the 2K in order to generate color graded output. This requires both a real time video playback device for the source, and a real time video recording device for recording the output.

    5. “Power Windows”, also known as area isolations, were basically invented by DaVinci, but the 2K only allows ellipses and rectangles as the available shapes. Colorists have cleverly used multiple windows to define oddly shaped areas on the 2K for years, but the fact is that only basic shapes are available.

    6. Due to its nature as a real time device, motion tracking is not possible or available. All power window movements must be implemented by hand using keyframes.

    7. There is a hardware based limit on how many levels of correction are possible. This limit is different in different systems because it is based on optional boards that may or may not be present.

    8. Although the 2K is technically capable of 2K RGB image resolutions – indeed, it was used for most of the digital intermediate work done in Los Angeles for the first few years of D.I. – the vast majority of systems are installed in HD video environments, and that is the resolution they generally operate in.

    9. The 2K is a discontinued product. There will be no future development or enhancement of the 2K. Any existing 2K’s will continue to be used as long as it makes sense to do so (as it does in the case of some television post facilities), but newer workflows and the need for resolution independence demand a different approach, one accommodated only by the much more flexible software driven devices.

    DaVinci Resolve (aka Blackmagic Resolve)

    Resolve was DaVinci’s answer to the limitations of the 2K. The original Resolve was a software/hardware hybrid, that was based on file based sources, but whose processing was hardware assisted. Since being acquired by Blackmagic, the system has gone far more to a software base, with much of what was done on dedicated hardware now moved to general purpose GPU boards. There are some similarities between the 2K and the Resolve, but they are mostly related to terminology rather than actual operation or interface.

    PROS:

    1. Implemented on multiple platforms (at the moment, Linux and Mac OS X), allowing some platform specific codec support and multiple price points.

    2. Ability to re-order a conformed timeline allows for some nice workflows, particularly for commercial work. One can start with a full conform, allowing for in-context grading work. That timeline can then be re-sorted to, say, source time code order, and either rendered or played out as video with the original source time codes, allowing a conform to take place on a finishing system, such as a Smoke or Flame, using the original EDL.

    3. Object based tracking allows very quick and accurate tracking of areas in the image. For instance, the colorist can draw an ellipse around someone’s head, and the system will then determine all trackable points within that shape, track them all, and average them to achieve a smooth but accurate track. This allows for much faster tracking than is available with point trackers.

    4. Resolve reads Red’s R3D files directly and allows their use for grading without a previous transcode step. It supports some other native file formats as well, and on the Mac version, supports Quicktime with use of Mac specific codecs, such as ProRes.

    5. Node based operation allows for some interesting ordering of grading layers and operations.

    6. The Mac version is the cheapest professional grading program currently available, by a wide margin.

    7. The most recent version supports Red’s Red Rocket accelerator card.

    CONS:

    1. Although the system has basic conforming tools, Resolve’s timeline is a single layer timeline, which means that any edits that involve multiple video tracks – very common today – will have to be either corrected separately, or the video tracks flattened if possible. Most professional facilities using Resolve use a different platform for conforming, often Autodesk Smoke or Assimilate Scratch.

    2. The Linux version, often advertised as a $50K solution, is only sold in a turnkey configuration and is considerably costlier than that oft-quoted price point.

    3. The Mac version is expansion and thus feature limited due to the lack of available PCIe slots in a Mac Pro computer.

    4. The lack of multiple tracks in the timeline limits the ability to do basic shot fixes that involve any paint type work.

    Autodesk’s Lustre

    Lustre is sold as an Autodesk Systems Product, in the same family as Smoke, Flame, and Inferno. It is still available on both Linux and Microsoft Windows, although most professional installations are on the Linux operating system. It is currently available only as a turnkey product.

    PROS:

    1. As an Autodesk Systems product, it inherits some technology from the Flame and Smoke code base, including excellent keyers, a very fast and accurate tracker, flexible GMasks, and a familiar looking timeline (at least for Smoke users).

    2. Since Lustre was designed to be used as both a video grader and a film grader (for digital intermediate work), it includes both types of controls – a “standard” lift, gamma, and gain configuration for video sources, and a film-centric exposure/contrast arrangement for use when working with Cineon log based sources. While not unique (Baselight has a similar design), it is very useful as it allows for film work that retains more of a “filmic” quality.

    3. Good interoperability with Autodesk’s other systems products makes it practical to do conforming on, say, a Smoke system, grading on Lustre from the same timeline, and revisions to flow bidirectionally as needed.

    4. Lustre implements an object based tracker, very similar to Resolve, in addition to the traditional point based tracker.

    CONS:

    1. The conforming tools are relatively weak, as Lustre is designed to be used in concert with a Smoke system for conforming. A professional facility using Lustre will really need a companion Smoke system to be used effectively.

    2. Lustre does not support many file formats. It basically expects everything it imports to be in the form of a DPX file, thus requiring transcoding of some file formats that other systems can use directly, such as R3d. Autodesk has published some effective workflows for this, but the fact remains that native file formats are generally not supported in the product.

    Filmlight’s Baselight

    Baselight is a Linux based system that has been used primarily by digital intermediate facilities, but has also found a home in some commercial houses and more recently for television series use. It is very scaleable, and is available in configurations ranging from a basic system appropriate for television use, all the way up to a version capable of working in uncompressed 4K resolution in real time.

    PROS:

    1. Baselight’s color toolset is quite deep and includes both video and film style grading tools that can be mixed and matched on any shot as desired. This allows for very filmic style grading, but with the nice “cleanup” capabilities of video style tools, such as easily balanced blacks.

    2. Baselight’s timeline can include a virtually unlimited number of tracks, and supports direct import of both Avid AAF and Final Cut Pro XML files, including direct import of some effects such as transforms. In addition to being a world class color grader, Baselight can also be used as a very capable high end conforming system, and its multitrack nature allows retention of the original editing transitions, which in turn allows the colorist direct access to any individual image in a multitrack stack. This is a big advantage when dealing with multilayer dissolve sequences.

    3. The multitrack timeline allows for “fix it” type effects involving multiple sources to be performed in real time as part of the color grading session. For instance, simple object removal using a matte and a repositioned second layer is easily accomplished.

    4. It has a well developed 3D stereoscopic toolset that allows for automatic color matching of left and right eyes, as well as grading of either eye separately or both eyes together. All Baselight hardware systems can now support 3D directly.

    5. Baselight Kompressor is a Mac based import/export node that allows real time use of Quicktime codecs that are normally unavailable on the Linux platform, such as ProRes. Material can be both imported and exported in those formats.

    6. Multiple timelines can be loaded simultaneously and accessed by multiple cursors that can be either independently positioned, or ganged. This allows quick and simple comparisons of either different timelines or different parts of the same timeline.

    7. Baselight has extremely wide native file format support. It can directly access MXF (including all flavors of DNxHD), R3d, Arri Raw, Phantom Raw, H.264, and numerous other file formats. This allows for conforming projects that involve multiple formats without any transcoding prior to the conform, and color grading directly from the original files. For the next release, due imminently, Filmlight has announced Red Rocket support, and additional native file format support that includes AVC Intra and XDCam.

    CONS:

    1. Tracking is still implemented as a point based tracker. Although accurate, it is not as versatile nor as fast as an object based tracker.

    2. There is no paint tool.

    3. There are no software based video style scopes, as there are in some of the competing products (there is a real time histogram).

    4. All Baselight systems are sold turnkey and are somewhat more expensive than some of the systems described here (not all, though).

    That’s it for this post. I’ll be adding a Part Two some time later in the week, in which we’ll look at the remaining systems (Film Master, Pablo, Scratch, and Speedgrade).

     

    8 responses to “Grading The Graders, Part 1” RSS icon

    • Great article! I’ve been waiting for the right someone to write this very article for a while now. It’s great to read informative post production blogs that aren’t just a bunch of opinions & hot air. Looking forward to Part 2.

    • Mike! This is good stuff you’re writing about here — don’t leave us hanging!

      Regards,
      Neal

    • Hello Mike, nice article! Just a minor note: the “very fast and accurate tracker” in Lustre is original development from the original Lustre dev team, and not Autodesk/Discreet technology. Not that it matter much 😉

    • Lustre
      pros:
      Have advanced gallery, users and project manager.
      cons:
      In secondary lustre can’t desaturat and it big problem.
      For me than keyer work only with source or primary.

    • Very good article. I didn’t see the FilmMaster review and was disappointed in all the things left out on Lustre, especially relating to it’s 3D capabilities, open architecture for Plugins, file format support via a gateway and color management flexibility. But all in all, pretty accurate in that these systems are more similar than different and the single biggest differing factor is the skill of the operator.

    • Please evaluate SGO Mistika (http://www.sgo.es/es/) Many large studios have been choose this. As I know is very competitive for 3DS too.

    • Regarding the 2K plus “Resizing of the image is not possible”.

      If you are using a Clipster (realtime resizing) as your source, you can use the 2k to create a pan/tilt/zoom list .

    • That is true. However, it should also be pointed out that in that configuration, it’s not the 2K that’s doing the resizing, it’s the Clipster. The 2K, as I said in the article, does not have any resizing or scaling capabilities on its own. In fact, DaVinci themselves tried to address this with the Splice product, which was intended to bring some of the Resolve’s capabilities (primarily its nonlinear nature and the Transformer board, to do exactly what you’re describing) to the 2K. To my knowledge, not many Splices were built or sold, and the product itself was killed (along with the 2K) when Blackmagic bought the company.

    Leave a reply