I've always wondered how Hollywood movies are scaled down to SD frame sizes for DVD without losing quality. Everything is shot in 4K, and sometimes even digitally with aspect ratios etc coming into play. So how do they do it and keep everything so sharp and crisp?

Every time I render my 1080p final master down to SD it looks like total crap. I don't know why I even bother with HD when everything has to be mastered for SD anyway.

Also, a lot of TV programming is shot in HD and downscaled to SD for broadcast. How do THEY keep everything so crisp? What do they do? Its a mystery to me.

Anyone care to enlighten me?

because when you convert, you choose a profile and run it on all the movie.
in hollywood, each scene is analyzed and encoded according its specification.
you can do that with a multipass encoder using variable bit rate (VBR) and calculate the bitrate to
use the maximum disk space. (if you burn DVD-R, don't forget that is usually 4.5Gig while most hollywood movies are burned on dual layer disk, offering more than 7gig).
additionally , your source are probably 4:2:0 interlaced when hollywood source are at least 4:2:2 or even 4:4:4 progressive
but if you really want, you can make DVd as goog as what hollywood produce.

I guess they just have lots of money/pro software to play with that your average prosumer can't afford.

I've always wondered how Hollywood movies are scaled down to SD frame sizes for DVD without losing quality. Everything is shot in 4K, and sometimes even digitally with aspect ratios etc coming into play. So how do they do it and keep everything so sharp and crisp?

Every time I render my 1080p final master down to SD it looks like total crap. I don't know why I even bother with HD when everything has to be mastered for SD anyway.

Also, a lot of TV programming is shot in HD and downscaled to SD for broadcast. How do THEY keep everything so crisp? What do they do? Its a mystery to me.

Anyone care to enlighten me?

VC100 | Teranex (http://teranex.com/products/vc100)

CineVision - Overview | Sonic Professional Products (http://www.roxio.com/enu/products/professional/cinevision/overview.html)

Along with software (which is proprietary and not for sale to the public), it's a specific job title and I believe they actually treat compression per scene much like a colorist would. A buddy is the manager at the audio restoration studio for Universal and he pointed out the building across the lot that did DVD mastering and compression. He said that's where the mad scientists worked and performed their voodoo rituals!

Most cheap DVD authoring software will provide 'average' quality if you simply let it convert to mpeg-2 in auto set button. To obtain best quality, you must change the compression details. Try to keep your movie to 60-minutes or less so that you can up the bitrate to maximum.

Most DVD authoring will auto-choose to convert to Mpeg at around 5000-7000 kbps which results in low quality.
Also, even if you press the "High Quality" button it will generally keep it at 70% maxiumum quality and 8,000 max kbps to prevent the possibilty of not fitting all the information onto the DVD disc.

It then provides you with:
Video data rate: Variable (Max. 8000 kbps)

Most hollywood feature films are compressed to mpeg-2 at 9,500kbps or higher. I think they are able to this due to using dual-layer and having sound on seperate streams.

However, I've found that I can get very close to this figure on a normal single-layer DVD-R disc by adjusting the compression settings to around 9,200kbps up to a maxium of around 9,400.

You need to also maintain this bit rate throughout the mpeg conversion, so do not leave it on variable rate (because the bit rate could drop well below highest quality during the conversion). I used to do a 2-pass conversion, but now prefer to leave it in constant highest bit rate well above 9000. (When you leave it set at constant data rate, the software will normally block out the 'variable' rate box option anyway).

So for example, the settings for highest quality PAL DVD on your normal home-PC DVD authoring software would be:

MPEG files (*mpeg;*m2t)
24 bits, 720 x 576, 25 fps
Lower Field First
(DVD-PAL), 16:9
Video data rate: 9,400 kbps (or highest possible above 9,000 depending on total video length and sound quality).
digital Audio, 448, or 256 minimum.

(A similar option can be done for NTSC) I hope this helps.

Most cheap DVD authoring software will provide 'average' quality if you simply let it convert to mpeg-2 in auto set button. To obtain best quality, you must change the compression details. Try to keep your movie to 60-minutes or less so that you can up the bitrate to maximum.

Most DVD authoring will auto-choose to convert to Mpeg at around 5000-7000 kbps which results in low quality.
Also, even if you press the "High Quality" button it will generally keep it at 70% maxiumum quality and 8,000 max kbps to prevent the possibilty of not fitting all the information onto the DVD disc.

It then provides you with:
Video data rate: Variable (Max. 8000 kbps)

Most hollywood feature films are compressed to mpeg-2 at 9,500kbps or higher. I think they are able to this due to using dual-layer and having sound on seperate streams.

However, I've found that I can get very close to this figure on a normal single-layer DVD-R disc by adjusting the compression settings to around 9,200kbps up to a maxium of around 9,400.

You need to also maintain this bit rate throughout the mpeg conversion, so do not leave it on variable rate (because the bit rate could drop well below highest quality during the conversion). I used to do a 2-pass conversion, but now prefer to leave it in constant highest bit rate well above 9000. (When you leave it set at constant data rate, the software will normally block out the 'variable' rate box option anyway).

So for example, the settings for highest quality PAL DVD on your normal home-PC DVD authoring software would be:

MPEG files (*mpeg;*m2t)
24 bits, 720 x 576, 25 fps
Lower Field First
(DVD-PAL), 16:9
Video data rate: 9,400 kbps (or highest possible above 9,000 depending on total video length and sound quality).
digital Audio, 448, or 256 minimum.

(A similar option can be done for NTSC) I hope this helps.

Excellent bit of info there mate, pretty much the settings i've been using however, there's no chance i could achieve anything above 8200kbps bit rate with a 90 minute wedding DVD :(.

Be careful here...

The maximum permissible bit rate TOTAL for a DVD is 10.08 Mbps. And the maximum permissible VIDEO only rate is 9.8 Mbps. This has nothing to do with dual layer or anything else. It is in the codified standards of DVD creation as set forth by MPEG.

If you encode that high, older units will have a DEVIL of a time playing it back. I remember having trouble playing back the second Matrix movie in my older player. It would stutter and freeze at the same point each time. It worked in my newer units, and prompted a replacement of my older player.

If you encode at 9800 Mbps, you will leave yourself barely enough room to put mp3 quality audio on the disc. So if sound is important to you, you're gong to need to back off the video bit-rate so that you can have higher quality sound. If you intend to place more than one video or audio track that could play at the same time (like allowing a commentary track), you'll again need to leave room in the bit budget for it.

The fact of the matter is that excellent DVD quality comes from several places

1. Well shot material. Most people here aren't shooting on RED, or 35mm, or the things Hollywood does. And YES it makes a vast difference in the quality.

2. Most people aren't using excellent tools to get that HD or greater resolution down to SD. This is where the vast majority of the quality loss occurs.

3. Most people are not using first class mpeg2 encoders.


The combination of these three factors, all but ensures soft transfers.

Though I didn't give much technical info in my first response, I think this has taken a turn in the wrong direction. I will second Perrone's advice on the high bitrate and will guarantee it's not high bitrate that makes Hollywood DVDs look so good. The last company I worked for hired a software engineer to write an algorithm and software for DVD encoding. This took almost 6 months to get it sorted out with 4 video editors (me being the senior editor) and a few other specialists from the software end of the biz. Our main goal was quality vs speed of encode as this company through-puts an enormous amount of video each day and all the commercially available encoders were not up to the task. Problem is they don't scale up to handle serious volume. And by writing our own software, we got dramatically better results and still in an automated way. However, as I mentioned before, scene by scene and variable bitrates are the answer. And those tasks were automated in software. Most of the dvds were 2 hours on a single layer and looked fantastic.

Theres a boatload of threads about workflow and my contributions to those threads all come from advice given to me by the software guys who "rolled their own" method. Scaling is the first part of the problem and because I'm an FCP guy, this is especially apparent. Unfortunately, the solution requires a double encode. Problem with mpeg2 is it's not an efficient codec and any noise or artifacting or complex visuals (like water or snow) in the original video "freaks it out". Hence the need to treat those scenes separately. Since I don't have time to do that, the solution for me was to let the H264 codec do the overall 1st pass and scale it during that pass as well. During compression, the artifacting is handled and effectively smoothed during that pass. When moving to the next step, mpeg2, the codec doesn't have to work nearly as hard and becomes much more efficient using VBR. Keep in mind that this is with QT and compressor NOT the proprietary software.

I have stuffed 2 hours and 30 min on a single layer at 3.7-4.5mbps which looked pretty darn good. Not perfect but WAY better than any other method I tried. And I tried a bunch of them. Had I gone in to the scenes that bothered me and tweaked the encode to suite them, I could have made it 100%. Took a long time to double encode but all the DVDs that went out, not one complaint on the quality, in fact they raved about how good it looked compared to the previous company and the gig is already on my calendar for early next year!

Thanks for chiming in Robert. Nice to hear someone speak on the level of work required to get very good results. Most people are looking for the "easy answer". And in those cases, quality will be compromised. When I was interested in getting my encodes to look more solid, I always did the downscale in virtualdub using the Lanczos rescaler. There are other excellent methods (including bicubic spline for things that might shimmer or cause moire or aliason) but Lanvzos worked well for me 99% of the time.

After that downscale to a lossless codec, I'd do my encodes in DVD Architect or Sony Vegas. I knew the mpeg2 encoder wasn't awesome, but it was better than most. Now that I have Sorenson Squeeze, things are a bit better. And Avid's downscaler is first rate, so I no longer have to jump to virtualdub.

But the fact remains, Bitrate is a piece of the puzzle no quesiton, but not the biggest piece. Even average rescaling and encoding looks pretty amazing when starting with well shot 4K or 2K material.

Look at us on a Saturday night talking about encoding! Ha ha!
The gig I referenced as my "quest" for the solution with my given tools was shot on my hmc150 and was a bit noisy even though the lighting was good. I recently did a similar shoot with my XF300 and the difference down the line was staggering even though both were 1080 30p in the beginning. In my estimation, the noise I think was the crux of the issue. When downscaling, those bits of noise or artifacting get smaller and actually more complex for the encoders. The XF just provides a much cleaner image to start with (and for the price it had better!)
Working in 2k or 4k would be awesome! Maybe someday the Scarlet will become real! Sigh!

Here's a good method I've borrowed form Anton using TmpgEnc...
TmpgEnc 4 Xpress Tutorial for downscaling Edius HD to SD (http://www.videoproductions.com.au/html/t4-hd-sd.html)
Many artifacts dealt with handily!

As Perrone advised, try to maintain a balance of good sound and highest quality video compressed onto your DVD... and obviously that starts from the ground up; so keep your filming and sound recording at the best levels that you can obtain in any given situation and equipment used.

Regarding producing actual DVDs, I never try to squeeze 2-hours of movie onto a DVD disc, and ALWAYS try to stick to the 1-hour max rule, because this allows you to include both sound and video at highest SD levels throughout the DVD. With 2-hour movies I always prefer to make them into 2-disc DVD sets with 1-hour on each disc. On the subject of bit rates, as mentioned, I normally compress at 9000-9,400 for a 60-minute SD movie (inclusive of authoring buttons & images) and not above that level.

On a different note, it has become very frustrating that Blue-ray has not caught on fast enough worldwide, especially across Europe. I had hoped that sales of Blue-Ray players and discs would have reached a level high enough to make it worthwhile to produce them in large volumes during 2009 and this year, but the simple facts are that the much higher costs of production and lower returns in total numbers of sales of Blue-ray means that even during 2010, the need to Down-rez HD video to SD for normal DVD production is still a frequent occurrence.

On a different note, it has become very frustrating that Blue-ray has not cought on fast enough worldwide, especially across Europe. I had hoped that sales of Blue-Ray players and discs would have reached a level high enough to make it worthwhile to produce them in large volumes during 2009 and this year, but the simple facts are that the much higer costs of production and lower returns in total numbers of sales of Blue-ray means that even during 2010, the need to Down-rez HD video to SD for normal DVD production is still a frequent occurrence.

In the midst of a worldwide economic recession, what else could be expected? I am still creating HD masters for EVERYTHING though. I can always downscale at delivery.

I don't think that economic recession is the main cause, but more to do with the massive rise in streaming videos via internet. There is no doubt in my mind that the future of TV and films is the World-Wide-Web.

I don't think that economic recession is the main cause, but more to do with the massive rise in streaming videos via internet. There is no doubt in my mind that the future of TV and films is the World-Wide-Web.

I'll buy that, but we aren't there yet. The US has some of the best broadband connectivity in the world, and honestly, less than a third of our population can stream movies. It's going to take some time before the WWW replaces a set top box. Heck I *MAKE* movies and I still prefer BluRay and DVD. The quality from streaming is 3-4 Mbps at BEST. Youtube quality may look ok for some, but on a large display in the living room, I really want something better.

...Regarding producing actual DVDs, I never try to squeeze 2-hours of movie onto a DVD disc, and ALWAYS try to stick to the 1-hour max rule, because this allows you to include both sound and video at highest SD levels throughout the DVD. With 2-hour movies I always prefer to make them into 2-disc DVD sets with 1-hour on each disc. On the subject of bit rates, as mentioned, I normally compress at 9000-9,400 for a 60-minute SD movie (inclusive of authoring buttons & images) and not above that level...

Tony why don't you use a double layer DVD so you can have one copy for your customers? That's what I do and the quality is great, besides double layer DVDs are not that expensive.

Stelios

Stelios, the reason is marketing and sales. The cost of producing a double-layer with onbodies in DVD packs and seal-wrapped is similar to that package as a 2 disc DVD set. The advantage is that I can advertise the movie as a special 2-disc pack as opposed to a single disc, and this attracts higher sales.

Thanks to everyone for shedding light on this matter. Are there standalone MPEG-2 compressors that would let me do the second pass (H.264-MPEG-2) rather than open a new project in Vegas? Also, are there ways to tailor how certain scenes are rendered in Vegas or Premiere? Sometimes it's only a matter of a few seconds that get pixelated, and I'd like to target that area only.

J.

...The US has some of the best broadband connectivity in the world, and honestly, less than a third of our population can stream movies...

Well, it sorta depends on how you define "best" I think. There are a number of organizations that track broadband implementation and usage around the world. The US ranking tends to average around 15-20, and is falling steadily. Here's a specific report (http://www.itu.int/ITU-D/ict/publications/idi/2009/material/IDI2009_w5.pdf) to look at, see page 22. There are other reports, this is just the one that came to hand the quickest for me.

The problem isn't so much broadband penetration in the US, it's more about speed. The bit-rates just aren't there, even in the major cities, to support downloading full blu-ray (1080p24) quality video.

Without either competition in the last mile (to your house), or a government mandate, there's really no reason for the telecom / cable providers to improve the infrastructure. They know from experience that we'll continue to pay high rates for low grade service rather than give up the service all together. So what's their incentive to move from 1960s copper wire (which they already have installed) to fiber optics (which would be expensive and painful to install, and would generate hardly any additional revenue)?

Yet even with the problems of obsolete infrastructure in the US, Netflix downloads now account for 20% of peak 'net activity. (http://blogs.barrons.com/techtraderdaily/2010/10/22/netflix-now-over-20-of-peak-downstream-us-net-traffic/) Amazing -- people really prioritize convenience over quality.

Much as I would like to see it, I doubt I'll have fiber optics to my house in my lifetime. Thank goodness for Netflix, blu-ray, and roof-top antennas!

...Sometimes it's only a matter of a few seconds that get pixelated, and I'd like to target that area only.

J.

Jacques, I find that sometimes, rather than converting directly from full 1080 to mpeg-2, I convert the problem file to a HDV 720P file prior to compressing back to mpeg-2 and this often prevents certain difficult clips from pixelating, especially if the problem area is large areas of black in the scene that need bringing back to solid black. If this doesn't work, then try lowering the Gamma levels on that particular clip during post editing. Sliding the Gamma level from zero to minus until you have pure clean blacks, also works very well in lowering noise levels in night footage.

Yet even with the problems of obsolete infrastructure in the US, Netflix downloads now account for 20% of peak 'net activity. (http://blogs.barrons.com/techtraderdaily/2010/10/22/netflix-now-over-20-of-peak-downstream-us-net-traffic/) Amazing -- people really prioritize convenience over quality.

Who can forget "CD quality" 128-kbps mp3s? ;-)


J.

Jacques, I find that sometimes, rather than converting directly from full 1080 to mpeg-2, I convert the problem file to a HDV 720P file prior to compressing back to mpeg-2 and this often prevents certain difficult clips from pixelating, especially if the problem area is large areas of black in the scene that need bringing back to solid black. If this doesn't work, then try lowering the Gamma levels on that particular clip during post editing. Sliding the Gamma level from zero to minus until you have pure clean blacks, also works very well in lowering noise levels in night footage.

Thanks Tony. I also see the problem during transitions between clips, not just with noisy blacks. I'll also try Robert's suggestion to render to H.264 first, then to MPEG-2.


J.

Hi Jacques,

If you are trying to use Vegas to go one step from HD to an MPEG-2 for DVD you will never achieve the quality you're looking for. The problem lies in two areas, the fist is, much like FCP, Vegas does not scale video very well. As Perrone mentioned the best quality you'll get is taking the HD edit to a lossless HD file, then use an external program to scale it. VDub does a good job and is free.

The second is that the MPEG encoder in Vegas is decent but there are much better ones if you are willing to pay for them. One of the best encoders that won't cost you an arm and a leg is Cinema Craft Encoder SP3. It can do up to 8 passes so it will compress your video pretty much as far as it can go and produces noticeably better results than most other encoders small operations can afford. I has an adaptive quantization matrix or I believe you can adjust the quantization matrix based on segments of the movie so you can adjust it to yield the best quality for each given scene. But, it is $1000 for a program that basically just encodes DVD's. So if it can pay for itself it could be worth it but it is still expensive for most small companies.

I'm not sure about encoding first to H.264 then to MPEG. Remember that H.264 is a very compressed format already. I've tried this route and there is still noticeable recompression issues. It might be the encoders that come with Vegas but I did notice a loss in quality. H.264 was never meant to be an intermediate codec. I would use one of the other lossless intermediate codecs instead.

It does take longer to do the multi-step process of HD render, resize, render to mpeg but it does yield noticeably better results.

Just my experiences and take on things.

-Garrett

Wow, lots of resources and information in this thread!

Can anyone discuss the comparisons in quality and performance of video encoders like Adobe Media Encoder CS5 and ProCoder 3 etc, as compared to TMPGEnc for scaling HD down to SD for DVD.

Also, would the same HD to SD downconvert methods and techniques apply when converting HD source to DV PAL for broadcast?