HD to SD Downconversion

Erik Phairas · December 21st, 2008, 10:10 AM

all the bigshots that post here and none of them actually work for the companies that produce professional DVDs like the new Batman or whatever? Be interesting to read what they do that is so different.

Bill Ravens · December 21st, 2008, 10:17 AM

ahhhh...at last this whole dowrezzing thread is saying things that make sense to me. The fact of the matter is that commercial burning software costs 10's of thousands of dollars. Why, if it can be had for under $100? Well, the answer is because of the sophistication of the algorithms that do the encoding, and the computational power needed to drive that software. Ask Spielberg. Decoding is deriguer. But, the real effort and technology is in the encoding. And that's where most indies lose their quality....because you will never get Spielberg quality out of your quad core, running Vista64, or your MAC, whatever it is. We're getting closer to Spielberg's capabilities, but, I don't think Dreamworks is standing still, either.

Matt Davis · December 21st, 2008, 10:18 AM

Quote:

Originally Posted by Bill Ravens

Does anyone else experience line twitter? Speak up or forever hold your flicker.

I did a while back in January (where did the year go?!) when doing down converts in the NLE. If I went out and shot things with strong lines just off horizontal, it was horrible. When I bought my own EX1 I turned down the detail and it's been less of a problem ever since.

My main issue now is with clothing containing fine checks - bit of diffusion to calm down moire now and again. Most of the time, I get away with standard settings - but to Compressor users, don't forget to turn on Frame Controls to enable the better scaler. And as Alister says, play with the Anti Aliasing control.

Perrone Ford · December 21st, 2008, 10:59 AM

Quote:

Originally Posted by Erik Phairas

all the bigshots that post here and none of them actually work for the companies that produce professional DVDs like the new Batman or whatever? Be interesting to read what they do that is so different.

FotoKem - DVD Mastering

Call them and ask.

But I can guarantee you, when you start with 70mm or even 35mm, telecine on a Spirit, grade on a Davinci, and Mpeg2 encode on real encoders, the results are NOT going to be mirrored by some lackey like me on Vegas with DVD Architect!

In case you are wondering how to get Hollywood level authoring, contact Sonic and ask about their Scenarist product. You can get the BluRay version too. When I looked into it, I think it was $27k though, so that wasn't going to happen at my office!

Bill Ravens · December 21st, 2008, 11:07 AM

Perrone...

Interesting you mention Fotokem. I've done some work with them with film transfers. While they may have the market share and technology, there are QC issues. Guess nothing is ever perfect. But, then, again, I'm a small niche indie. They serve the networks. I really can't and don't expect the same level of attention that they give to a weekly prime time series.

Erik Phairas · December 21st, 2008, 11:39 AM

I'm sure there are reasons why that software cost so much but it brings to mind something that happened the other day...

I was buying plasti-gauges. They are used when you install rods into an engine. I was upgrading rods on my supercharged ATV.(had to mention that..LOL). anyway the gauges are just thin plastic thread, you put it between the rod bearings and the crank shaft, torque it all down and remove everything. You then measure how much the thread squashed and it tells you your clearance.

It's THREAD, 12 inches long... so for that 12 inch thread, it cost me about 3 bucks (there were 2). THREAD.. looks like piece dental floss. 3 bucks...

So now I am thinking about making plasti-gauges for a living.. LOL

Sorry back on topic.

David Schmerin · December 21st, 2008, 12:06 PM

Quote:

Originally Posted by Perrone Ford

FotoKem - DVD Mastering

Call them and ask.

But I can guarantee you, when you start with 70mm or even 35mm, telecine on a Spirit, grade on a Davinci, and Mpeg2 encode on real encoders, the results are NOT going to be mirrored by some lackey like me on Vegas with DVD Architect!

In case you are wondering how to get Hollywood level authoring, contact Sonic and ask about their Scenarist product. You can get the BluRay version too. When I looked into it, I think it was $27k though, so that wasn't going to happen at my office!

Back in the day before Scenarist was owned by Sonic, it cost closer to $40k and only ran on NT. Combined with my Pioneer DVR-S201 and an OptiVision V-Stor MPEG-II Encoding system, and I was the only person in the MidWest who could produce Hollywood quality DVDs for many a year.

Bottom line is to produce Hollywood style DVDs takes the right Tools and the right Talent. Nothing has changed. As Ford says, some lackey with Vegas and DVD Architect simply can not produce the same level of product as a Master DVD Author with professional tools.

Sadly it would seem that in today's world, manufactures have adopted a line from The Kinks song, "Low Budget Man" where they sing, "Quality Costs but Quality Wastes"

Speaking for the Minority of 1
David Schmerin
Royalty Free Stock Footage And Video

Bill Ravens · December 21st, 2008, 12:19 PM

Quote:

Originally Posted by David Schmerin

some lackey with Vegas and DVD Architect simply can not produce the same level of product as a Master DVD Author with professional tools.
David Schmerin
Royalty Free Stock Footage And Video

LOL...still laughing over that line. And in order to be an equal opportunity insult to ALL editors, let's be sure to include, besides vegas, FCP, Avid, Premiere.No one holds a patent on incompetence or incompatiblity.

Perrone Ford · December 21st, 2008, 12:25 PM

Quote:

Originally Posted by Bill Ravens

LOL...still laughing over that line. And in order to be an equal opportunity insult to ALL editors, let's be sure to include, besides vegas, FCP, Avid, Premiere.No one holds a patent on incompetence or incompatiblity.

Hey, I was just speaking for myself. But if you want to paint everyone with the same broad brush.. ;)

Mitchell Lewis · December 21st, 2008, 04:09 PM

So back to the topic....

I'm starting to understand the importance of this now. Let's see if I can dumb it down with a couple of potential examples:

EXAMPLE 1: I create a television commercial in 1080P. It uses lots of text that has a black outline around each character to make it more readable. I purposely choose an outline thickness that's attractive but not too thick. My finished commercial looks great in 1080P. Then I scale it down to 480 to burn to DVD. This effectively shrinks the width of my outlines by more then half. When I view the finished commercial on DVD the outlines are so thin you can barely see them. The outlines also twitter (I call it interlace "flicker" but what ever) really bad and are very distracting.

So how can a better quality encoder help with this problem?

EXAMPLE 2: I've noticed that when I'm editing a standard 720x480 DV project and I import a high resolution digital photo (something like 3000x2800), the photo looks funny. It has a lot more detail than my DV video and all the fine lines twitter. Is this sort of another example of bad encoding?

I'm just trying to rap my brain around how I'm going to produce television commercials with our new EX3 (when it gets here) and deliver them to the local broadcasters in SD. This looks to be an unexpected problem. I've been reading this thread carefully. I'm hoping that Compressor will hold the best solution for me. But I'm not so sure of that.

Bill Ravens · December 21st, 2008, 04:33 PM

Mitchell..

pretty classical problem. I suppose it's somewhat counter-intuitive, but, adding a small amount of blur to the text, before you downrez, will help your situation. Trial and error is the best way to find the right amount of gaussian blur.

Perrone Ford · December 21st, 2008, 04:57 PM

Quote:

Originally Posted by Mitchell Lewis

So back to the topic....

I'm starting to understand the importance of this now. Let's see if I can dumb it down with a couple of potential examples:

EXAMPLE 1: I create a television commercial in 1080P. It uses lots of text that has a black outline around each character to make it more readable. I purposely choose an outline thickness that's attractive but not too thick. My finished commercial looks great in 1080P. Then I scale it down to 480 to burn to DVD. This effectively shrinks the width of my outlines by more then half. When I view the finished commercial on DVD the outlines are so thin you can barely see them. The outlines also twitter (I call it interlace "flicker" but what ever) really bad and are very distracting.

So how can a better quality encoder help with this problem?

It can't.

Remember my workflow? Where you cut the size of video you want up front, instead of trying to do it at the end of the workflow? Guess what? That workflow path eliminates this problem by not forcing you to rescale after you've laid in your graphics. And the lovely thing is that since the text and titles are not raster-based, you don't even have to do anything other than lay the video in underneath of the right size and re-render. I developed this workflow precisely because I have to deliver to both DVD and the web and I got tired of my stuff looking like garbage after rescales. So with HD, I can now cut HD, SD, or web with a minimum of fuss. Get a request in the morning, and have a new deliverable that afternoon or the next morning.

Quote:

Originally Posted by Mitchell Lewis

EXAMPLE 2: I've noticed that when I'm editing a standard 720x480 DV project and I import a high resolution digital photo (something like 3000x2800), the photo looks funny. It has a lot more detail than my DV video and all the fine lines twitter. Is this sort of another example of bad encoding?

First, there is no way you should be dropping files that size on a DV timeline. That's like dropping RED 2K footage on VHS. Rescale the photo to DV size and then bring it in. Otherwise you're going to have to add some blur to it so that it drops in more realistically.

Quote:

Originally Posted by Mitchell Lewis

I'm just trying to rap my brain around how I'm going to produce television commercials with our new EX3 (when it gets here) and deliver them to the local broadcasters in SD. This looks to be an unexpected problem. I've been reading this thread carefully. I'm hoping that Compressor will hold the best solution for me. But I'm not so sure of that.

Follow the workflow I gave. Scale the video BEFORE bringing it into the NLE and you'll have no problems. This way, you can edit, grade, etc., on the SD version and deliver it. And should you need an HD version later, you can simply do a "replace video" from the timeline and suddenly you have an HD version with all the titles and graphics in exactly the same place and proportion. Assuming you cut a 16:9 SD. If you're cutting 4:3 SD versions then you'll have a bit more work.

Mitchell Lewis · December 21st, 2008, 10:23 PM

Hmmm......I agree that your workflow is the solution. But I'm not thrilled with the idea of shooting all this glorious HD footage and instantly converting it to SD. If that's the workflow, why did we buy a fancy HD edit room monitor. I'm not saying your wrong, I'm just thinking out loud here....

Also, do you think this is how the big production houses do it? We get Beta SP dubs from the big post houses for commercials like Sylvan Learning Center. I've always noticed how in their spots (Sylvan) they can get away with things I'd never be able to do with DV footage. Thinks like very narrow color type on a white background. When we do this it twitters something awful. Sometimes even the color the text will "rainbow". But get these national spots on Beta, capture them into FCP with our standard DV codec and they hold up just fine. It's weird!

I guess I was always hoping in the back of my mind that by shooting and editing in HD I would now be able to get away with doing things like this and the image would hold up. Even when we scaled it down to HD.

BTW, we do most of our editing in Adobe AfterEffects. We only capture and play out the final rendered piece in FCP. Swapping SD for HD footage would be a little more difficult for the majority of our projects because they are very graphic intensive.

I'm sorry, I just realized I've taken this thread more into a FCP category. Not XDCAM EX. My bad. :)

Perrone Ford · December 21st, 2008, 10:53 PM

Mitchell, I used to wonder for YEARS how the big boys did it. I have friends here in broadcast. Production manager at the local PBS station, camera and edit folks at the City's TV station, the local AP affiliate, and used to have friends at the local Fox and NBC affiliates. In every case, when I told them that I was now HD end to end, including deliverables, they looked at me in amazement.

I had dinner with my friend at the city last Wednesday. I asked what cams they were using. They have a couple of HDCams, but mostly DigiBeta and miniDV. The friend I have at the local AP affiliate cuts on Avid. They have Digibeta cams. The local PBS affiliate has DVCam. I swap tapes with them regularly because of our broadcast agreements.

In reading the past 3 years of ASC, I was constantly struck by the workflow.

Shoot film, 2K DI, cut proxies for dailies, rushes, and rough cuts and grades. Take all that into a post house for work on the DaVinci, Get picture lock and send out to sound. Screen that, then master out to film. So I looked at my workflow when I went to HD. After beating my head against the wall trying to edit and color 1080p, I cut uncompressed or Cineform proxies at DV size or 540p and went to work. When I was happy, I did media replacements and let it render of the weekend or longer if necessary. I have been totally pleased with the results.

Tonight, I spent some time doing codec testing. Basically laid a bunch of test patterns on a timeline, and rendered them into every major codec on my machine. Then did differences against the uncompressed originals. The results I got back taught me a LOT. I did this same test 2 years ago with DV codecs. But this one was with 720p. In every single instance, the quicktime codecs came back indistinguishable from the source. PNG Adaptive, PNG Paeth, QT Lossless. The Sony AVC encoder was clean, but increased both color saturation, and contast. Windows Media loss quite a lot of detail, but it's colors and luma was on the money. The lagarith codec that I thought I was going to use to replace working uncompressed disappointed me a LOT.

So it looks as though I'll be cutting and mastering with Quicktimes in Vegas. At least until Cineform gets 64bit support.

So now that I've taken the thread in a COMPLETELY different direction, let's bring it back. The choice of codec is absolutely CRITICAL in maintaining visual quality as we go through the workflow. Those who discount this and simply use automagic settings and opt for the simple and easy workflows may well find that they are losing quality even when they think they are being careful.

Trust, but verify. Drop some test patterns onto your timeline, some ramps, gradients, color bars, etc. Render it, and then drop that render back into your timeline underneath the original. Then either setup a difference composite, or A/B them with the scopes up. The Waveform, Vectorscope, and Histograms will tell you plenty. And have a look at the RGB info too. You'll be amazed where your chroma info goes with some of these codecs.

Bill Ravens · December 22nd, 2008, 08:30 AM

Wonderful, terrific post Perrone. I've done similarly. Running the following codecs thru AVID, using SMPTE bars as a baseline:
WMP11
Cineform CFHD
Mainconcept MPEG2
Apple H.264
Bitjazz' SHEER
DNxHD
AVID MPEG4
Canopus HQ
Quicktime Reference(embedded DNxHD)

For my money, Cineform is the most stable, both in quality and color rendition, followed by SHEER. The SHEER codec is right up there, but, takes a bit more attention when applying. Both codecs are cross platform, but my testing was done on a PC platform. Exporting to a MAC v10 gave the same result.

When working with FOTOKEM on a film transfer, we always work with DNx175. I've intercut super 16mm with Panny P2 footage...needing to match color and dumb down the panny footage to match.

I should note that the QTRef is a real timesaver, as it's equivalent to frame serving to the next application, be it AE or a transcoder like Procoder. Also, note that the above testing was done to determine the best inter-application workflow stream. The rules are different if I never leave Avid from capture to master. In that case I'd stick with DNxHD.

I have used Vegas, extensively in the past. Vegas seems to have a real bone with Quicktime. If using QT within vegas works for you, I'm, honestly, quite amazed.

Best to you, thanx for confirming my own tests.

December 21st, 2008, 10:10 AM	#46
Erik Phairas Major Player Join Date: Sep 2008 Location: Las Vegas, NV Posts: 628	all the bigshots that post here and none of them actually work for the companies that produce professional DVDs like the new Batman or whatever? Be interesting to read what they do that is so different. __________________ EX3, Q6600 Quad core PC - Vista 64, Vegas 8.1 64bit, SR11 b-cam

December 21st, 2008, 11:39 AM	#51
Erik Phairas Major Player Join Date: Sep 2008 Location: Las Vegas, NV Posts: 628	I'm sure there are reasons why that software cost so much but it brings to mind something that happened the other day... I was buying plasti-gauges. They are used when you install rods into an engine. I was upgrading rods on my supercharged ATV.(had to mention that..LOL). anyway the gauges are just thin plastic thread, you put it between the rod bearings and the crank shaft, torque it all down and remove everything. You then measure how much the thread squashed and it tells you your clearance. It's THREAD, 12 inches long... so for that 12 inch thread, it cost me about 3 bucks (there were 2). THREAD.. looks like piece dental floss. 3 bucks... So now I am thinking about making plasti-gauges for a living.. LOL Sorry back on topic. __________________ EX3, Q6600 Quad core PC - Vista 64, Vegas 8.1 64bit, SR11 b-cam

December 21st, 2008, 04:09 PM	#55
Mitchell Lewis Trustee Join Date: Dec 2008 Location: Kennewick, WA Posts: 1,124	So back to the topic.... I'm starting to understand the importance of this now. Let's see if I can dumb it down with a couple of potential examples: EXAMPLE 1: I create a television commercial in 1080P. It uses lots of text that has a black outline around each character to make it more readable. I purposely choose an outline thickness that's attractive but not too thick. My finished commercial looks great in 1080P. Then I scale it down to 480 to burn to DVD. This effectively shrinks the width of my outlines by more then half. When I view the finished commercial on DVD the outlines are so thin you can barely see them. The outlines also twitter (I call it interlace "flicker" but what ever) really bad and are very distracting. So how can a better quality encoder help with this problem? EXAMPLE 2: I've noticed that when I'm editing a standard 720x480 DV project and I import a high resolution digital photo (something like 3000x2800), the photo looks funny. It has a lot more detail than my DV video and all the fine lines twitter. Is this sort of another example of bad encoding? I'm just trying to rap my brain around how I'm going to produce television commercials with our new EX3 (when it gets here) and deliver them to the local broadcasters in SD. This looks to be an unexpected problem. I've been reading this thread carefully. I'm hoping that Compressor will hold the best solution for me. But I'm not so sure of that. __________________ Sony EX3, Canon 5D MkII, Chrosziel Matte Box, Sachtler tripod, Steadicam Flyer, Mac Pro, Apple/Adobe software - 20 years as a local videographer/editor

December 21st, 2008, 10:23 PM	#58
Mitchell Lewis Trustee Join Date: Dec 2008 Location: Kennewick, WA Posts: 1,124	Hmmm......I agree that your workflow is the solution. But I'm not thrilled with the idea of shooting all this glorious HD footage and instantly converting it to SD. If that's the workflow, why did we buy a fancy HD edit room monitor. I'm not saying your wrong, I'm just thinking out loud here.... Also, do you think this is how the big production houses do it? We get Beta SP dubs from the big post houses for commercials like Sylvan Learning Center. I've always noticed how in their spots (Sylvan) they can get away with things I'd never be able to do with DV footage. Thinks like very narrow color type on a white background. When we do this it twitters something awful. Sometimes even the color the text will "rainbow". But get these national spots on Beta, capture them into FCP with our standard DV codec and they hold up just fine. It's weird! I guess I was always hoping in the back of my mind that by shooting and editing in HD I would now be able to get away with doing things like this and the image would hold up. Even when we scaled it down to HD. BTW, we do most of our editing in Adobe AfterEffects. We only capture and play out the final rendered piece in FCP. Swapping SD for HD footage would be a little more difficult for the majority of our projects because they are very graphic intensive. I'm sorry, I just realized I've taken this thread more into a FCP category. Not XDCAM EX. My bad. :) __________________ Sony EX3, Canon 5D MkII, Chrosziel Matte Box, Sachtler tripod, Steadicam Flyer, Mac Pro, Apple/Adobe software - 20 years as a local videographer/editor

December 21st, 2008, 10:53 PM	#59
Perrone Ford Inner Circle Join Date: Nov 2006 Location: Tallahassee, FL Posts: 4,100	Mitchell, I used to wonder for YEARS how the big boys did it. I have friends here in broadcast. Production manager at the local PBS station, camera and edit folks at the City's TV station, the local AP affiliate, and used to have friends at the local Fox and NBC affiliates. In every case, when I told them that I was now HD end to end, including deliverables, they looked at me in amazement. I had dinner with my friend at the city last Wednesday. I asked what cams they were using. They have a couple of HDCams, but mostly DigiBeta and miniDV. The friend I have at the local AP affiliate cuts on Avid. They have Digibeta cams. The local PBS affiliate has DVCam. I swap tapes with them regularly because of our broadcast agreements. In reading the past 3 years of ASC, I was constantly struck by the workflow. Shoot film, 2K DI, cut proxies for dailies, rushes, and rough cuts and grades. Take all that into a post house for work on the DaVinci, Get picture lock and send out to sound. Screen that, then master out to film. So I looked at my workflow when I went to HD. After beating my head against the wall trying to edit and color 1080p, I cut uncompressed or Cineform proxies at DV size or 540p and went to work. When I was happy, I did media replacements and let it render of the weekend or longer if necessary. I have been totally pleased with the results. Tonight, I spent some time doing codec testing. Basically laid a bunch of test patterns on a timeline, and rendered them into every major codec on my machine. Then did differences against the uncompressed originals. The results I got back taught me a LOT. I did this same test 2 years ago with DV codecs. But this one was with 720p. In every single instance, the quicktime codecs came back indistinguishable from the source. PNG Adaptive, PNG Paeth, QT Lossless. The Sony AVC encoder was clean, but increased both color saturation, and contast. Windows Media loss quite a lot of detail, but it's colors and luma was on the money. The lagarith codec that I thought I was going to use to replace working uncompressed disappointed me a LOT. So it looks as though I'll be cutting and mastering with Quicktimes in Vegas. At least until Cineform gets 64bit support. So now that I've taken the thread in a COMPLETELY different direction, let's bring it back. The choice of codec is absolutely CRITICAL in maintaining visual quality as we go through the workflow. Those who discount this and simply use automagic settings and opt for the simple and easy workflows may well find that they are losing quality even when they think they are being careful. Trust, but verify. Drop some test patterns onto your timeline, some ramps, gradients, color bars, etc. Render it, and then drop that render back into your timeline underneath the original. Then either setup a difference composite, or A/B them with the scopes up. The Waveform, Vectorscope, and Histograms will tell you plenty. And have a look at the RGB info too. You'll be amazed where your chroma info goes with some of these codecs. __________________ DVX100, PMW-EX1, Canon 550D, FigRig, Dell Octocore, Avid MC4/5, MB Looks, RedCineX, Matrox MX02 mini, GTech RAID, Edirol R-4, Senn. G2 Evo, Countryman, Moles and Lowels.

December 21st, 2008, 10:17 AM	#47
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	ahhhh...at last this whole dowrezzing thread is saying things that make sense to me. The fact of the matter is that commercial burning software costs 10's of thousands of dollars. Why, if it can be had for under $100? Well, the answer is because of the sophistication of the algorithms that do the encoding, and the computational power needed to drive that software. Ask Spielberg. Decoding is deriguer. But, the real effort and technology is in the encoding. And that's where most indies lose their quality....because you will never get Spielberg quality out of your quad core, running Vista64, or your MAC, whatever it is. We're getting closer to Spielberg's capabilities, but, I don't think Dreamworks is standing still, either.

December 21st, 2008, 11:07 AM	#50
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	Perrone... Interesting you mention Fotokem. I've done some work with them with film transfers. While they may have the market share and technology, there are QC issues. Guess nothing is ever perfect. But, then, again, I'm a small niche indie. They serve the networks. I really can't and don't expect the same level of attention that they give to a weekly prime time series.

December 21st, 2008, 04:33 PM	#56
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	Mitchell.. pretty classical problem. I suppose it's somewhat counter-intuitive, but, adding a small amount of blur to the text, before you downrez, will help your situation. Trial and error is the best way to find the right amount of gaussian blur.

December 22nd, 2008, 08:30 AM	#60
Bill Ravens Join Date: Nov 2001 Location: santa fe, nm Posts: 3,264 Images: 10	Wonderful, terrific post Perrone. I've done similarly. Running the following codecs thru AVID, using SMPTE bars as a baseline: WMP11 Cineform CFHD Mainconcept MPEG2 Apple H.264 Bitjazz' SHEER DNxHD AVID MPEG4 Canopus HQ Quicktime Reference(embedded DNxHD) For my money, Cineform is the most stable, both in quality and color rendition, followed by SHEER. The SHEER codec is right up there, but, takes a bit more attention when applying. Both codecs are cross platform, but my testing was done on a PC platform. Exporting to a MAC v10 gave the same result. When working with FOTOKEM on a film transfer, we always work with DNx175. I've intercut super 16mm with Panny P2 footage...needing to match color and dumb down the panny footage to match. I should note that the QTRef is a real timesaver, as it's equivalent to frame serving to the next application, be it AE or a transcoder like Procoder. Also, note that the above testing was done to determine the best inter-application workflow stream. The rules are different if I never leave Avid from capture to master. In that case I'd stick with DNxHD. I have used Vegas, extensively in the past. Vegas seems to have a real bone with Quicktime. If using QT within vegas works for you, I'm, honestly, quite amazed. Best to you, thanx for confirming my own tests.