Shrek is a landmark film in many ways, not the least of which is that it is one of the earliest films to be transferred to DVD direct from the digital source files, making the video transfer one of the very best you are ever likely to see until the advent of even higher definition video formats than DVD. We wanted to "peek under the hood" at the process of how this transfer was taken from the original computer animation files and put onto DVD, and were privileged to be given the opportunity to question Mark Atkinson from DreamWorks about the nuts and bolts of getting this movie onto this spectacular 2 disc set.
AK: What is your position at DreamWorks and what does this role involve?
MA: We don't exactly have titles here (it's a title-less company), but my basic area of responsibility is video mastering, so pretty much everything from the transfer of the movie from film to video or in the case of Shrek from digital files to video. [My responsibility extends] pretty much throughout the life of the film beyond the initial theatrical release, so when it plays on television, when it goes out on DVD, when it goes out on VHS, [and] when it plays on airlines, all the versions that are created - my department is responsible.
AK: What do you think of DVD-Video as a format?
MA: I personally am a huge fan of the format. I have a rather extensive collection of my own and think it is brilliant. I am a huge fan of the format and have been since its inception, following the development of the format through the years of positioning between the Toshiba/Time Warner camp and the Sony/Philips camp and all the conflict that was going on before they finally decided upon a format and I have been a fan ever since it finally came out.
AK: After the success of the initial release of Shrek, why re-release the film as a 2 disc Special Edition?
MA: That is more of a marketing call and not a technical call so I am not able to answer that one as it is not my area of responsibility.
AK: Were any of the assets from the previous release used for the new 2 disc version?
MA: The new 2 disc version more-or-less mirrors what was released here in the States, so it basically includes the DTS audio which was not previously not available internationally. It also includes the pan and scan version of the film and one or two of the bonus pieces.
AK: You referred to the 4:3 version of the film as pan and scan - is it actually pan and scan or a re-framed version?
MA: It is essentially a pan and scan version. When they animate a film like Shrek or essentially any animated film with the exception of something that might be 2.35 aspect ratio, typically the animation is done at 1.66:1 and then for theatrical they project at 1.85 so they crop off the top and bottom of the animation - it is a non-essential area of the animation and is just a little bit of overspill which is primarily there these days to allow us to do an easier pan and scan version for the video release. So we actually do pan and scan to get it down to 1.33:1 but not a huge amount as we start from 1.66:1.
AK: Can you briefly explain how the image is passed down the production chain from the original single rendered frames to the final MPEG2 encoded files?
MA: It begins with the animation process where they are rendering everything at approximately 2k [Ed. 2048 pixels per line of image]. It is done in a proprietary format, which allows for it to be up-rezzed or down-rezzed as needed - it's a vector-based animation format. Those original files are converted to sequenced TIFF and we then convert those to high definition video. Basically they give us a bunch of drives, a couple of terabytes of data storage in a truck and we take it to our video facility and convert it all to high definition video and then we colour correct it using high definition monitors and a DA VINCI 2K high definition colour corrector. We do this at a lab here in town called Modern Video with a colourist named Skip Kimbell. We sit with the director and spend about two weeks going through and colour balancing everything from the original digital files so that it looks at its optimum best on a high grade telecine monitor. And once we have everything dialed in to the satisfaction of the director we record it to high definition D5 at 24 frames [per second]. That is then downconverted to standard PAL 16x9 and then that tape (that is usually a digital Betacam), is put through the MPEG2 compression. The audio lives separately and it is conformed and streamed to the DTS or Dolby Digital files and it all marries together in the authoring.
AK: From the D5 tape how is it converted to the D1 version?
MA: It is done using a separate device to create PAL and NTSC D1 separately using either the Teranex system, which is a format converter for high to standard definition, or there is also a Panasonic universal downconverter that is used. The results from either version are just about the same.
AK: During the conversion, take the NTSC version for example, when going from a 24 frame master downconverting it to an interlaced format with 3:2 pulldown and then encoding it and presumably removing the 3:2 pulldown and flagging it as a progressive stream - do any of these conversions leave any noticeable artefacts?
MA: None that I have ever been able to spot. We typically will put things up side-by-side to verify that there are no artefacts. When you get down to it, adding 3:2 pulldown and pulling it out again is just taking out fields of video and the devices that do that do it pretty cleanly. I am yet to see a major artefact but once in a while you will see something - a device will spit up a spare frame or a spare field - but that is why we do so many QC steps along the way. Pretty much after every stage we do a QC pass, a quality control pass, and we will hire an operator to sit there and in real time watch the whole film from beginning to end, checking all the audio and watching all the video and noting any problems. As they do these conversions, it is not just "set it up and let it run", it is always an ongoing process, stop and go, looking for any issues.
AK: What are the actual MPEG encoders that are used for the transfer?
MA: Well usually for the MPEG encoding we are using the Sony encoder at the facility we use for authoring and compression. I personally think that the Sony encoder does a great job with the compression. It is done usually in what we call a three-pass process where the machine goes in and analyses what the best bit rate is going to be for particular scenes as it is all variable bit rate compression. They then review that, do a QC and make some adjustments, they run it again, do more QC, make more adjustments, going back and forth tweaking everything to get it all at its optimum level. I think in reality it turns out to be five or six passes before they are done with it and satisfied with the compression. At the end of it all, I or someone from my department will go in and look at the compression to make sure that it looks perfect or as good as it can be.
AK: To your knowledge is this process similar for all major computer animated titles at this time?
MA: Every studio is pretty much using the same process; we are all using pretty much the same authoring and compression facilities. There is only a handful of really top notch authoring and compression facilities out there and most of the major studios are using them. So really all the computer animated movies that came out; Sony with Final Fantasy, Disney with the Toy Story stuff, DreamWorks with Shrek and Jimmy Neutron was Paramount. I think we have all pretty much used the same facilities and they have the same processes in place as far as encoding goes.
AK: With the current advances in computing power would you expect future MPEG encodes to be done via software directly from the original source material?
MA: The source files are not optimized for the colour space of the television monitor. Usually they are rendering the animation to be projected on a movie screen and so the colour space that they are rendering in is different for what you would want for TV. I doubt that this is something that could happen - the multiple passes we go through are involved in first getting it to the proper resolution and also to get it into the right colour space and have the director review it all.
AK: You mentioned the colour space that the film is created in - what sort of colour depth is used for the rendering process?
MA: I think that they are typically 16-bit, and I think that is pretty much across-the-board on film these days, but I am not an animator so I cannot tell you specifically what it is.
AK: I noticed that the Kodak Cineon Lightning film recorder used to create the film prints utilizes a 10-bit log colour depth, so why render using 16 bit colour?
MA: When you get down to it, the high definition video is all 10 bit as well on the D5, so to be honest I think that it is because this is the system that they brought into DreamWorks but it might be an attempt to future-proof, looking down the road so they can still use some of the gear and some of the files that they can later re-purpose.
AK: When you approve the colour on the 10 bit HD D5 masters does the conversion down to 8 bit for DVD ever produce any problems?
MA: Yes, we did have some problems with banding. What would happen [was that] when you removed some of the colours, you would get bands where it would go from nice even tonal progression to bands of colour. That was a bit of a challenge - what we ended up having to do was to apply some dithering to that in the original animation to compensate for it and in some cases we had to remove some of the grads because there was just no way of making them go away. But at the end of the day, it looked pretty darn good.
AK: What authoring system was used to create the Shrek DVD?
MA: On Shrek we used the Daikin Scenarist.
AK: The totally rock-steady image of Shrek is almost unsettling in its clarity - is there ever any need to deliberately introduce artefacts to prove a more life-like look?
MA: Not really, not unless it is something that the director wants and I have to say after sitting with Andrew [Adamson] on the colour correction he loves the way it looks. He was delighted - he wanted everything to go out. Initially, we were colour correcting this for digital cinema as well and he wanted to do theatrically only digital cinema as he thought it looked so much better than the film presentation. So he was very happy with how it turned out and if he is happy I am happy.
AK: Was the same audio master used for both the theatrical and DVD releases of the film?
MA: Yes, it is the same 5.1 mix we used for both. The only difference is for the PAL market everything runs at 25 fps so you are going to hear a little bit of pitch adjustment on that.
AK: Is the transfer pitch corrected?
MA: We typically try not to do that only because pitch correction can add some digital stepping. So unless a director specifically asks us to do it we avoid it, as do most studios.
AK: Was the same audio master used for both Dolby Digital and dts soundtracks and was any different pre-processing applied for either mix?
MA: Yes the same mix was used for both and no pre-processing was applied.
AK: Are there any other areas in relation to the transfer that you wish to comment upon?
MA: It is certainly one of my favourites; I think it looks beautiful. I was really happy with the way it came out and so much of that is because they did such an incredible job animating the film - the detail is amazing.
AK: Thank you very much for your time.
MA: My pleasure.
Thank you to Rebecca Chaires, Meagan Breakwell and Katie John for their assistance in making this interview possible.