I've been working with video for 9 years as a hobby and side business. For the last four+ years, I've been working primarily with DV and MPEG2. I've owned many, many types of encoders both RT HW and software. Right now, I use ProCoder from Canopus (not to mention their totally awesome DV Storm Pro card). I'm a quality fanatic and I always strive to produce the best looking MPEG I can, but I've never been able to produce anything that compares to what TriStar/Columbia, DVCC, Superbit, etc. produce. My experience with HW encoders tells me that software encoders are a bit better and lots of research is the reason I use ProCoder. What are these studios using to produce this magnificent MPEG2. . .not to mention the fact that they do it at relatively low bitrates. I've use bitrate viewer to analyze lots of stuff from my commercial DVDs, but cannot find an answer as to what they used to produce it.

Does anyone out there actually work at one of these places or has knowledge of the system they use to produce such crisp looking video? I had thought that perhaps they author all this in much higher resolutions and then down convert it to 720x480 during the build, but I don't think this would have been possible when DVDs first appeared due to technology limitations. Of course, most studios master in high definition now and do down conversions and I can see a slight gain from doing this, but I can't imagine getting the kind of clarity I see in a movie such as Lord of the Rings or Starship Troopers.

Anyone?

part of the answer is they start with quality material (35 or 70mm film) that has nothing to do with DV.
another thing they probably do (and we can, but never spend the time to) is to encode scene by scene, with parameters adjusted for best rendering each time

It is a combination of things and yes, they simply have the best
encoders as well. Keep in mind that early DVD's (some discs still
do, sigh) looked absolutely worse compared to what is out today.

One of the tricks they have is that they can fine tune the encoding
on a scene-to-scene (or even a seperate picture) basis if they
so want to. I don't know that much specifics but I presume it is
quite the manual labor.

It surprised that you find your discs so much more worse. Since
you have worked for so long I assume you are familair with things
like multipass VBR encoding.

Also keep in mind that all Hollwood encodes (that look good) are
done in true anamorphic at the highest source quality as possible.
A lot of people do not have true anamorphic here and we have
to cope with at least compressing the footage twice. And ofcourse
things like higher color space resolution etc. will help as well.

As it's been said you need to shoot on 35mm film, do your telecine on a C-Reality, Shadow or Spirit and make the encode from HD D5 source at 24p - or at least Digital Betacam.

There are options but the main list of big boy encoders are:

Sony Vizario
Sonic SD-2000
CinemaCraft Encoder Pro
Spruce MPX3000
Optibase (maybe)

No others spring to mind really... other than the ones that come with the Warner Bros. Toshiba system or the Panasonic.

Yeah there is lots of segment re-encoding done for sure. Many many passes. CinemaCraft SP allows 9-passes (not segment re-encodes though) and it's regarded as the 'best' software encoder by many but it really depends on your source footage and the look you want.

Jake

Make sure to read up on the available literature.

One book that has been recommended to me is Ben Waggoner's "Compression for Great Digital Video." I know a guy that took his class and was very impressed. I think Waggoner is one of the "go to" consultants that Studios use when they are going to encode a film to DVD. So, it appears he knows what he's doing.

I haven't read the book yet, but I'd like to. Cost is around $35, I think.

Regards,

Kyle "Doc" Mitchell

Worth a read maybe and it's free:

http://www.cinemacraft.com/eng/download.html

and get the CCE SP user manual

Jake

Noise has a lot to do with it. I've had some wow-moments using Grain Surgery Noise reduction for After Effects. Seems that the encoders are very sensitive to noise, meaning that they spend a lot of energy encoding noise, when they should be encoding actual picture details.

DNVC in bitvice does a great job of clearing noise up.

Jake