How are people rendering?

TeeJay wrote on 5/19/2005, 3:52 AM
Hi all,

I have done a few menu based projects and have always rendered each individual clip as an MPEG2, then built my menu in DVDA2 which obviously creates my end DVD.

I have just now wondered whether I am doing this the right way. Is this "double handling"? Does DVDA re-render my clips or does it realise that they have already been rendered and just let em through?

Should I be rendering from Vegas as AVI (or something else) and letting DVDA do the rest?

I must say that my projects have generally come out pretty good, but I am always on the quest for better quality.

Love to hear some thoughts on this.

Regards,

TJ

Comments

Paul_Holmes wrote on 5/19/2005, 4:42 AM
I think I get the cleanest output by rendering my final movie to AVI, then I render two files, the DVD Architech mpg stream, then the AC3, using the same name for both. I then use those in DVD Architect and there's no re-rendering, just uses the files.

P.S. I like rendering the AC3 out in Vegas because if you pull in an mpg file that's been PCM encoded it will then create an AC3 from the PCM which is going from a compressed file to another compressed file.

There are times when I've pulled in a short AVI to DVD Architect and let it render it to MPG and AC3, but I like things to go quickly when I'm creating the DVD, so usually the files are all prepped and ready to go.
cheroxy wrote on 5/19/2005, 4:53 AM
PH, You say you get cleaner videos by rendering to AVI first? I always assumed that by rendering to AVI from compressed DV and then rendering the AVI to MPEG that you would get twice the artifacts that you would get from a single render.

Do you know why you get a cleaner vid?
rs170a wrote on 5/19/2005, 6:37 AM
I'm with Cheroxy on this. I always render direct to mpeg-2 & AC-3 from the Vegas timeline. A lot of my projects involve stills and, by doing it this way, you avoid having the image quality of them (they start in Vegas at 4:4:4) changed to DV (4:1:1) and then to mpeg-2 (4:2:0).
I use a bitrate calculator if my project is more than 60 minutes but, fortunately, this is rare for me.

Mike
Paul_Holmes wrote on 5/19/2005, 7:28 AM
Maybe in the case of photos brought into the timeline it would make a difference -- you've taught me something new. I didn't realize the chroma sampling was 4:4:4 in that case.

However, from what I understand you can render many generations of DV in Vegas without any appreciable degradation (I think SPOT has said up to approximately 10 or 12). The way I edit is to do first cuts, color correction, add effects, music, then render, then do some more editing, add titles, more music, whatever, then rerender, etc. etc. Usually I've got about 4 renders deep from the original by the time I'm done. (Not the whole project, but pieces of it (usually there's some intial color-correction, curves, etc on the first to 2nd generation).

So, at least in the case of DV I don't think you're losing anything appreciable if you render to an avi and then to mpg. Theoretically, yes, but visually no (except possibly in the case of photos). I'm working on a photo movie currently so I'll have to do some tests to see if your method makes a visible difference to my eyes on the final product.
JohnnyRoy wrote on 5/19/2005, 8:01 AM
> So, at least in the case of DV I don't think you're losing anything appreciable if you render to an avi and then to mpg.

Yes you are losing something and I guess the key word is what “appreciable” means to you. You are confusing the degradation of re-rendering DV with the loss of color information of going from 4:4:4 to 4:1:1. They are different things.

Vegas operates internally at 4:4:4. When you use generated media to make a title, or color background, that title or background is 4:4:4. When you render directly to MPEG2 that title goes from 4:4:4 to 4:2:0. When you take that same title and render to DV first and then MPEG2, it does from 4:4:4 to 4:1:1 to 4:2:0. The problem is you lost information going down to 4:1:1 so going up to 4:2:0 is not going to generate extra information. It’s lost forever. I use titles as an example because it is the most visible area of improvement in going from Vegas to MPEG2 rather than Vegas to DV to MPEG2. The same is true for color correcting or any other effects.

You can argue as to whether or not the difference is visibly “appreciable”, but you will get the highest color quality going direct to MPEG2 from the Vegas timeline because you are going direct from 4:4:4 to 4:2:0.

~jr
Paul_Holmes wrote on 5/19/2005, 8:04 AM
Well, as I said, I've learned something new. So from now on I will edit my DV, but titles will come last and be directly rendered to MPG. Also, any slide show or photo movies I do will be done that way also.


I suppose the only other way would be to render uncompressed, but that makes for a huge file, doesn't it.
craftech wrote on 5/19/2005, 8:35 AM
Actually I raised this question over a week ago.
For the life of me, I can't see a difference in the quality of the stills OR the video.

John
Paul_Holmes wrote on 5/19/2005, 9:09 AM
To say the least that was both an informative and "entertaining" thread, John. After reading JohnnyRoy's comments I had my eyes opened and John Cline's "tutorial" helped cement the facts into my thick brain.


BillyBoy wrote on 5/19/2005, 9:31 AM
In my opinion too much is made between actual and preceived loss of color information due to compression schemes. While of course factually once you remove information its gone forever, if or not that VISUALLY impacts what one sees is a totally different matter.

The reason is very simple. The human eye is far more sensitive to changes in luminance (brightness) levels then it is to changes in hue. This has been demonstrated in countless scientific experiments down through the decades.

So while "purists" may make a lot of noise for example about 4:2:0 and 4:1:1 what one SEES may or may not suffer that much a impact, depending on source material.

For YUV color space It boils down to this. Video has three components for every pixel. The luminance (called the "Y" value) carries the brightness level (what human eyes are very sensitive to changes in) and the color information carried by "U" and "V". These combined make up what's called YUV color space.

For 4:4:4, nothing is thrown away. Result "pure" color, huge file size.

For 4:2:2, you can save about a third in file size by averaging adjacent "U" and "V" values in every 4 pixel block. Result smaller file size, some loss of color information, image is less "pure" simply meaning there is a shift due to averging.

For 4:1:1 (DV) the powers that be decided they wanted still smaller files sizes. So instead of averaging using 2 values from adjacent pixels they bumped it up to averging 4 pixels. Result: Still smaller file size, but further loss in "pure" color information.

Remember what I said in the beginning?

The human eye is far more sensitive to changes in luminance (brightness) levels then it is to changes in hue. As you see the U and V portion can take a big hit, not the Y. So yes, while higher compression will alter "actual" color information due to averging it, your eyes may not be able to detect the change, or probably more accurately the change often will be minor thatnobody is going to know.

So in summary sure, loss of color information can be bad, but how bad is really relative. Quality is preceived. So depending on source you may see a change, then again often times you really won't.
John_Cline wrote on 5/19/2005, 9:37 AM
I'm into quality and I sweat the details. It's a curse. I have never completed a project with the words, "Oh, hell, it's good enough." unless I knew that it really was good as it could be (given any time or budgets restraints.) While it is true that the human eye is more sensitive to luminance than chroma, I just can't let anything go out of here knowing that it isn't of the highest technical quality that it can be under the circumstances.

I had done a render test going from the 4:4:4 Vegas timeline directly to 4:2:0 MPEG and then a test going from 4:4:4 to 4:1:1 DV then to 4:2:0 MPEG2. I uploaded the results in a 477k .ZIP file to my web site and posted the link in the thread to which Craftech referred to in the message above. I noticed about 50 people have downloaded the file, but no one has commented on the results. I think it pretty clearly shows that Vegas straight to MPEG2 is the way to go. Does anyone that has looked at the results have any comments?

Colorspace Conversion Test Results

(I also included the 4:4:4 test images in the file for anyone wanting to do their own experiments.)

Video and audio production involves a lot of "smoke and mirrors" but intentionally not doing something technical using the best method possible, particularly if it doesn't take any more time, is just being lazy. It's taking pride in one's own work whether anyone, other than myself, can see it or appreciate it.

John
BillyBoy wrote on 5/19/2005, 9:44 AM
I downloaded the file, but didn't comment because you fudged the results using a simplistic test using only geometric shapes having high contrasting colors which OF COURSE will impact the results.

Why not do a "real world" test where a actual photograph of somebody is used where the changes in hue will be far less dramatic.
slacy wrote on 5/19/2005, 10:11 AM
I would argue slightly by saying that quality is not really perceived. In Cline's test, the quality of one is clearly superior to the other, whether it's a real-world test or not. What is perceived is the *value* of minor increases in quality. And in that sense, BB, you're right, it's entirely subjective.

Here's the thing, though. If you know that rendering straight to 4:2:0 yields superior results, no matter how subtle, and no matter the perception of the increase in quality, why not make that your workflow? If you can avoid intermediate DV renders, then it seems like the smart thing to do.
John_Cline wrote on 5/19/2005, 10:53 AM
Yes, those test images were designed to be worst case examples. Chroma subsampling errors will be most visible on the edges of high contrast, fairly well saturated, diagonal objects. Like text, for example.

It was a "real world" video that started me on the investigation of chroma subsampling artifacts in the first place. I was working on a video taken in the jungles of Honduras. There was a lot of detail in the image because of the abundant foliage in the scene, but the people in the video were wearing bright, red shirts and any time they had their arms at an angle, I could CLEARLY see artifacts. It turned out to be chroma subsampling in the 4:1:1 DV video. The video was shot with a Sony DSR-570 camera with good glass and 2/3" chips, so the DV video looked as good as it possbly could. It was definitely the 4:1:1 chroma sampling that was the problem. Once I saw it for the first time, I can now see it everywhere. My life was a lot simpler before I became aware of this. Once I edited the "red shirt" video and it went to 4:2:0 DVD, the problem became ever more noticable. Even "regular" people saw it at that point. The answer would have been, "don't wear a red shirt" and "don't wave your arms around" but they did and they were, it was about as "real world" as it could be and there was no chance of a reshoot.

While there is nothing I can do about 4:1:1 DV footage, there are things I can do to ensure that these artifacts are minimized on stuff over which I have control. One of those things is rendering straight to MPEG2 from the Vegas timeline when there are conditions in the video which will trigger those artifacts. By the way, the test image with the three diagonal color bars looks just like part of one of my client's logos, there is nothing more "real world" than having to make that look good on video and being forbidden to change it in any way.

John
BillyBoy wrote on 5/19/2005, 11:54 AM
Indeed, you used worst case examples, but didn't admit as much till now. Many times in this forum things are held out to be something other than what they really are and skewed to reflect one's viewpoint or forgone conclusion, which is disingenuous. As I said, if your test image didn't include text, didn't have high contrast edges, wasn't on a diagonal, or as saturated the differences would not have been as striking as you presented them to be.
Chienworks wrote on 5/19/2005, 12:39 PM
Surely though this indicates that there is a difference. And since rendering directly to MPEG from the timeline keeps from compounding the problem, is faster than a two-phase render process, and also is simpler, it indicates that direct to MPEG renders are a better way to go. Whether the difference is noticeable or not doesn't affect the other two benefits, and the first benefit is present in subtle ways. There is no reason* to argue against it.

* Well, i can come up with one reason. If you normally render to several different output formats and you have used a lot of time-intensive effects or compositing, you can save your self a lot of time by first rendering to the least-compressed format you will use (usually DV .avi), and then use this file for the other encodes. This avoids processing all the effects and compositings for every render. Since that will only happen once it will speed up the overall process. But, if you're simply making a DVD and no other output formats then this doesn't apply.
John_Cline wrote on 5/19/2005, 12:43 PM
I didn't feel the need to "admit" anything as there was nothing to "admit." I didn't "rig" the test or skew it in ANY WAY. The results ARE the results. I wanted to demonstrate something and I chose images that would clearly show what I was talking about. Yes, the artifacts can be more or less noticable depending on the source material, but they are ALWAYS there regardless.

If anyone, other than me, had posted the results of the same test, you would have most likely remained silent. Your last post was squarely aimed at accusing me of being less than honest. THAT makes it personal. You have been warned over and over and OVER again by the forum moderators to keep the discussion on topic without resorting to thinly disguised (or blatant) personal attacks. Leave me ALONE.

John
BillyBoy wrote on 5/19/2005, 1:05 PM
Anyone reading the other thread clearly sees YOU are the one that always tries to make things personal saying things like even I should see the difference. I then challenged you to prove what you said. You accepted the challenge, but as we find out now you deliberatley skewed the results to try to win the argument by designing the "test" to make it look like you were right. Its a silly game you keep playing.

I have NEVER been warned directly by the moderators about anything. Yes, they have deleted some of my posts, as they have yours. Again, you attempt to distort the facts to suit your purpose or you try to play the victim.

You have repeatedly resorted to vicious personal attacks on me calling me racist and scum multiple times. You deliberately created a web site using my nick name "BillyBoy" using some crude condom "humor". That obviously was a deliberate personal attack. That thread was deleted by the moderators as have many, many of your other posts where you attacked me. So John, I guess you don't pay attention too well for everything you accuse me of doing you've done worse yourself and repeatedly, including calling for me to be banned from this forum several times.

This is yet another classic example where you start out saying one thing, get caught, then move the goal posts and pretend you didn't really mean what you said.

John_Cline wrote on 5/19/2005, 1:13 PM
First, www.billyboy.com is a REAL web site. Your namesake is the #1 selling condom in Germany. I had absolutley nothing to do with it in any way, shape or form. Just another example of having your facts completely and entirely wrong. You would have known this if you had done a simple "whois" search to see who owns the domain name and when it was registered.

Secondly, I didn't fake any results whatsoever. I provided the test images, do the test yourself. By the way, Billy, that is the EXACT same test using the EXACT same images that I used to prove to Pinnacle that their DV codec was flawed. It was a good enough test for them. In fact, I was publicly thanked by the president of Pinnacle for bringing it to their attention and they spent a lot of resources, both in money and manpower, fixing the problem.

Third, you absolutely earned the racist scum label and everyone but you knows it. By the way, I saw the video you made "poking fun" of me and Spot, it was borderline racist, too.

Fourth, Don't be responsible for getting this thread killed, Shut the **** up.

BB: "including calling for me to be banned from this forum several times."

I'd still like to see that happen. Nothing would make me happier.
Chienworks wrote on 5/19/2005, 1:31 PM
I don't think John created that website. It's an actual legit company in Germany.

[whois.joker.com]
domain: billyboy.com
status: lock
organization: Mapa GmbH
email: hostmaster@ision.net
address: Industriestr. 21-25
city: Zeven
postal-code: 27404
country: DE
Jimmy_W wrote on 5/19/2005, 1:50 PM
?????
BillyBoy wrote on 5/19/2005, 2:08 PM
My point is simply this. Cline among others say they have a right to "correct" my opinions they don't agree with, as he often has tried to do when external monitors or HD comes up, so why shouldn't I point it out when someone tries to pass off "testing" claimed to be done in a nonpartisan way, when in fact they now admit they did slant the test to come out to a conclussion that was predetermined because of the material they used to do the "testing". That is being disingenuous and misleading because this forum isn't suppose to be some place to try to win a pissing contest, rather to honestly address issues.

I'm no longer surprised that the same people frequently try to make it look like I say one thing when I've actually said something else.

The issue isn't will 4:2:1 have more color information than 4:2:0, that's obvious and just a Staw man argument. The real question is will anyone looking at the results "see" a difference. The answer is no, not always. Sometimes for sure, other times no. So skewing "test" results by using an extreme example and not a real world one is misleading.

Again here "real world" means a typical project, not the specific project Cline was working on in his attempt to justify how he did testing. If he would have used test material that didn't have text, or sharp contrast, etc., the test results may have been very different. Again it boils down to the source material which is the important part. No one answer is always the best solution.

John_Cline wrote on 5/19/2005, 2:31 PM
Billy,

Those were "pure" test images. In just the same way as audio equipment is tested using pure tones to measure for distortion. You can't very well do it by using heavy metal music, you wouldn't know if the distortion was in the equipment or the source material.

Like I already said, using a "real world" image would have masked the artifacts. While you may not have seen them, they would have been there nevertheless.

Whether it's HD, pro monitors or now, chroma subsampling, you just never hesitate at the chance to forcefully and defiantly demonstrate your ignorance. Why is that?
Spot|DSE wrote on 5/19/2005, 2:32 PM
For my own edification, could you please point out specifically how the test result is "skewed?"
When it comes to issues of titles, still graphics, bright reds, high contrast, horizontal lines, masking,or keying, the colorspace makes a tremendous difference.
If you are editing uncle Bob's vacation shot with a palm cam in the backyard, maybe no one will notice. But if you have motion, generated media, bright contrast, borderline illegal colors, bugs, lower thirds, high rez stills....it's quite noticable.

If you have a project that needs to go to several sources, the best way to maintain integrity is to render to 4:2:2 YUV, then use that as your main source file. This is how we store *most* client data, because it also is a smaller file that can be up or down sampled intelligently.

Back to point, I'd like to understand for myself, how the test/footage used, is skewed?
donwidener wrote on 5/19/2005, 2:33 PM
Paul,

Instead of rendering as dv or avi, will you now render to DVD3 compliant mpeg2 w/o audio, render the AC3 file as a separate file, then combine them in DVA3?

John and Spot, is that what you reccomend?