|Page (1) of 1 - 06/14/07||email article||print page|
Getting Intimate with CineForm IntermediateIn Part 1 of this interview, CineForm CEO David Taylor discusses the workflow of CineForm Intermediate
Digital intermediate workflow and Lossless compression seem to have become the buzz words of the month post NAB07 where a number of companies announced either new lossless intermediate workflow products or software support for existing intermediate solutions. For California-based developer CineForm, the industry attention must have been bitter sweet on one hand confirmation and acknowledgement of what CineForm has been touting loudly for quite some time, and on the other hand a newly populated stiff arena of competing products.
DMN contributing writer Mike Jones spoke in detail with CineForm CEO David Taylor to get the nitty-gritty on the workflow pros and cons of CineForm intermediate and how he sees CineForm in the context of a flood of lossless online production formats now filing onto the market and into the NLE.
MJ: CineForm has been around for a number of years now and seems to be achieving a state of maturity and a level of market acceptance. What's the journey been like for CineForm as a company? Did the market and other developers take a lot of convincing of the merits of a lossless HD intermediate workflow, or did you have results that spoke for themselves?
David Taylor: First of all, for purposes of accuracy, our codec is visually lossless, not arithmetically lossless. Our architecture is designed to deliver arithmetically lossless coding, but we havent turned this on yet. What this means in practical terms is the sensitivity of the human visual system (viewing conditions dependent) cannot perceive visual differences that measure in the range of 40+ dB and up (this is the measurement of source-to-compressed signal to noise ratio using the same material). CineForm codecs start in the high 40s and move up well into the 50+ dB range. I have included a graph of a recent comparative test we performed on our new CineForm 444 format versus the respected HDCam SR recording format. You can get full details on our website. Notice that CineForm visual quality is very constant due to the variable bitrate nature of our codec, while HDCam SR varies its PSNR performance due to the constant bitrate nature of its coding.
Now, back to your questions. . . There is no doubt that CineForm has had to ?prove ourselves to the filmmaking community. Projects like ?Dust to Glory that we were involved with 2-1/2 years ago certainly helped get us noticed. D2G was the first feature to ever use a compressed Digital Intermediate workflow throughout, with final film prints created from the compressed source material on the Adobe Premiere Pro timeline. The visual quality of the result was phenomenal.
The reason this project was important to CineForm is a couple reasons: First, we are a relatively small company that didnt have previous credibility playing with the ?big boys for feature film workflows, and second, conventional wisdom was (and still is to some extent) that you cannot use compression throughout post. The reason for this perception is valid that is, camera recording formats from which this judgment was based are NOT suitable for a multi-generation post workflow. Camera formats were designed with certain parameters to enable recording to tape, but the design parameters for a codec that can stand up to the rigors of a multi-gen post workflow are very different.
CineForm has focused on the latter, and has succeeded, but like anything else there are early adopters and later adopters. As image resolutions exceed HD and tend upward to 2K and 4K there is no question that compressed post workflows will be part of mainstream operations that the ?latter adopters will embrace at some point.
In terms of our ?journey and the perception of our company, I think two latter endeavours have been helpful in raising both awareness of what were doing, and also credibility. These two activities are first the porting from Windows to Mac OS. After all, perhaps 2/3 of independent film work is done on a Mac. Second is the introduction of our CineForm 444 format. Because CineForm 444 visual fidelity exceeds that of HDCam SR, we have gathered additional notice within the professional community because of the high respect paid to SR as a recording format. What this means for us in the middle portion of the market is that people recognize that we are a format at the high end becoming respected by professionals, so that opens further doors for us in the middle market.
MJ: You mentioned the move to a Mac version as of key importance. For a long time CineForm, both the codec and the software system, have been a Windows-only solution using the AVI wrapper. Recently you've moved to produce a MOV file format wrapper. Has this been an easy cross over? Or were there challenges involved in making CineForm universal?
DT: The work to port to Mac OS has been incredibly challenging more so than we imagined! After all, its ?just software, right? We have wanted to port to the Mac for quite some time, but the old Power PC processors didnt have the performance of Intel processors, and it also had a totally different instruction set, so prior to Apple announcing support for Intel processors, a port to Mac OS was impractical. As a small factoid, we have tens of man-years of code optimization invested in the Intel architecture to wring out high performance, so its not easy to port that to a new processor.
For the most part, the port of the codec itself was straightforward, but QuickTime and AVI are very different wrappers, on top of which we have to consider various operating system dependencies. So getting all of this ?just right took more time than we anticipated. But were now on the Mac. In fact, weve been shipping a beta version for a few months now which at this point works great on the Mac. In fact, CineForm QuickTime files can be moved interchangeably between platforms and played on either.
We also have developed a re-wrapping tool to re-wrap either AVI or MOV files into the other wrapper without disturbing the compression layer underneath. This is being received enthusiastically by our customers as they now can literally use the same files on both platforms when this workflow is necessary. We still have some work to do however for codec performance optimization. Our visual quality is identical on both platforms, but we still have some additional performance to wring out of our Mac OS threading in the next month or two.
MJ: It seems that after a long period of pushing the 'edit native' line for formats such as HDV a number of major NLE developers have changed tune and are now singing the praises of HD intermediates. We're seeing a crop of new codec products appear, most recently Apple's ProRes 4:2:2. This would seem to be a direct competitor to CineForm and in the past third party developers have often found it difficult to compete with a native Apple made product. How is CineForm different? What do you see as its advantages?
DT: In HD there are a myriad of different camera capture formats, and it seems more keep coming out all the time. So it makes sense for others to adopt the compressed DI strategy that we pioneered. There are a number of differences in what Apple has announced with ProRes and what CineForm wants to enable on the Mac.
First of all, our desire is to enable a cross-platform workflow to enable efficiencies that were previously unachievable. Most customers are ?forced to make a decision about their NLE and their operating system. Until recently it has been Premiere Pro/Windows versus Final Cut/Mac OS for instance. Obviously this is an over-simplification as there are other solutions including Sony Vegas plus higher end systems. But regardless, at HD resolution it has been almost impossible to move files between Mac and PC unless they were uncompressed, and uncompressed files are HUGE. For instance, if youre using an HVX200, once you ingest files onto a Mac they are re-wrapped in a manner that is incompatible with the PC. So you have a camera that theoretically ought to be common to both PCs and Macs but you cant move the files between them. This is the kind of compatibility we render obsolete. Second, with our CineForm 444 format we enable a 12-bit RGB workflow that is not addressed by ProRes which is a YUV 4:2:2 codec. Third, we are enabling a CineForm RAW workflow on Mac OS that will ultimately support all Bayer pattern cameras of which there are numerous available now, and more that will come to market. Because you never render into a RAW format, you only render OUT of RAW, you need an RGB codec as a ?partner for the RAW workflow. We believe CineForm 444 is an excellent choice for this.
Finally, remember that Adobe is moving support back to Mac OS this summer. We intend to support Adobe on Mac OS as we do on the PC, although the resulting CineForm products will be slightly different for Premiere Pro on the Mac as they are on Windows.
MJ: Apart from lossless compression, one of the key traits you market on is the huge boost to real-time performance with CineForm within various NLE's. Having used it myself it really has to be seen to be believed. Can you explain how this works? Is it to do with the way the codec is written or is it a plug-in architecture that gives the boost?
DT: The performance boost we get is based on a number of things. First of all, the manner in which Wavelets perform their compression has some interesting characteristics that can be exploited in post-production. One of these is its multi-resolution decoding capability. That means, from (say) a 2K source image you can dynamically choose to decode to 1K for display purposes. This is very powerful, especially if you can dynamically choose between full-resolution and preview resolution decoding. When you decode to half-horizontal resolution it takes roughly 40% of the CPU load as decoding the full-resolution image. If your video processing pipeline understands this, it can make decisions about what resolution is appropriate to decode to at any particular point in the timeline. So in this sense there is a symbiotic (if you will) relationship between the video processing pipeline and the codec that we exploit to its fullest.
Second, weve optimized our codec for performance in many dimensions. So for instance we can acquire anything from HDV to 12-bit CineForm 444 to CineForm RAW in real-time in software on a commonly-available Intel-based PC no specialized hardware required. As another example, CineForm RAW is approximately 5x to 7x faster than JPEG 2000 (another Wavelet codec) in software because we made design choices in our algorithms to exploit software performance on the PC. So considering use of J2K in a post environment would be difficult because of its poor software performance. Yes, CPUs continue to get faster, but so do the resolutions of the images were asked to process in cameras and in post.
Part Two will appear next week
Mike Jones is a digital media producer, author, educator from Sydney, Australia. He has a diverse background across all areas of media production including film, video, TV, journalism, photography, music and on-line projects. Mike is the author of three books and more than 200 published essays, articles and reviews covering all aspects of cinematic form, technology and culture. Mike is currently Head of Technological Arts at the International Film School Sydney (www.ifss.edu.au), has an online home at www.mikejones.net and can be found profusely blogging for DMN at www.digitalbasin.net
Related Keywords:Digital intermediate , CineForm, compression, codecs, NLE, video editing workflows