smart-renderer produces shortdistortions with titles

[Jerry Jones]

Yes, although I would point out that Pinnacle's Studio Ultimate 11 Plus sees the very same DVD-RAM disc and refers to the very same file as a "Title."

So I wonder why Corel's (Ulead's) software has been designed to distinguish DVD-RAM disc contents as "programs" while Pinnacle software makes no distinction between DVD-RAM or DVD-R or whatever and instead refers to all such contents as "Titles." In that respect, the Pinnacle program seems easier.

Unfortunately, the Pinnacle program can't seem to successfully grab it, but that's another issue.

Jerry Jones
http://www.jonesgroup.net

That's most likely the difference. So maybe that's why you see Program, and I see Title for DVD's?

[Jerry Jones]

I'm not sure if that's exactly correct, George.

We're not engineers, obviously.

But to me, it seems the MainConcept SDK for MPEG smart rendering is negatively impacting both the Corel (Ulead) and Magix programs in roughly the same way.

Users of the Magix software here are reporting that the distortions appear at the beginning of titles while Corel's (Ulead's) distortions appear at the end of titles.

So this suggests that each software developer -- working with the same MainConcept SDK -- might be implementing the smart render feature a little bit differently, but with the same fatal flaw.

My impression is that the MainConcept approach may be to cut on any I, P, or B frame in a long GOP sequence and then it may be creating a new "group of pictures" by inserting a new I-frame at the end (or beginning) of the cut, which sometimes matches up properly with adjacent frames, but sometimes doesn't; this is just a theory that may explain why some edits show the distortion and why some edits do not.

I think Apple's Final Cut Pro should be evaluated by Magix, Corel, MainConcept, etc. to find out if that company has managed to come up with a similar, but better way of creating the new GOP at the edit points; it's possible the Apple method might result in perfectly smooth video motion at the cut points.

If so, then perhaps MainConcept and the rest could modify their SDK accordingly.

I could be wrong, but my understanding of the Womble approach is that it constrains the user to make a cut only on the nearest I-Frame in a long GOP sequence. It may not allow for cuts on any frame, but the method results in cuts that do not render out as distorted.

Jerry Jones
http://www.jonesgroup.net

EDIT: I did not read all 90+ posts in this thread (forgive me if I am repeating something already mentioned). But to get back on topic -- are we saying that SmartRendering would work better if the program could identify the BEFORE and AFTER I-Frames of pieces of timeline that require re-rendering, and automatically turn SmartRender off for those sections of the timeline (making sure to always re-render from I-Frame to I-Frame)

Regards,
George

[GeorgeW]

I'm not sure if that's exactly correct, George.

We're not engineers, obviously.

Agreed, we're not engineers. Sorry if I was "thinking out loud" and was simply speculating what *might* help to resolve this issue (of course, the engineers would have to try to implement/test it, or there might be other reasons why it has no chance of working).

I got the idea from what seemed to produce positive results in this comment by the OP:

In VS I did the following to fix the problem:
-switched to max. zoom on the timeline that the cursor will jump frame by frame
-enlarge the title by so many frames that the last frame of the title will meet an I-frame with the movie on the video track.
If I render the modified project but leave smart-renderer enabled then the distortions have disappeared and the resulting video is excellent.
Unfortunately VS can not tell you which frame of an mpeg2 file is I,B or P frame so an external program(like virtualdub) is required to see this.
This is defenitely a bug in VS10, I hope some project leaders at ulead will read this and apply a fix(a patch) in a future release.
Regards
Sektionschef

Regards,
George

[Jerry Jones]

I've been doing some more surprising tests here in the studio.

Man, I'm surprised.

When I was a consultant for Ulead, I would hang out on the Pinnacle forum to read comments and try to get a feel for why some individuals had rejected MediaStudio Pro in favor of Liquid.

One of the comments that I remember reading was from a guy who had tried both editors and he insisted the output quality of the Pinnacle Liquid software was better than the output quality of Ulead MediaStudio Pro.

Today, I've been doing some more comparisons between the Pinnacle Studio Ultimate 11 output with the output of both VideoStudio and MediaStudio Pro.

VideoStudio and MediaStudio Pro seem to have identical render engines so their output quality is practically the same.

But I am really astonished to see how much better the Pinnacle render engine performs on virtually *all* tests.

This just shocks me; but it seems to confirm what that fellow said back in 2003.

First, I took a Panasonic VDR-M30 DVD camcorder long GOP MPEG-2 standard definition file...

NTSC drop frame (29.97 fps)
MPEG files
24 bits, 704 x 480, 29.97 fps
Upper Field First
(DVD-NTSC), 4:3
Video data rate: Variable (Max. 8800 kbps)
Audio data rate: 256 kbps

Then I transcoded this file in Pinnacle Studio Ultimate 11 to be a type 2 DV .avi file.

Then I transcoded the same Panasonic file in Ulead's software to be both type 2 and type 1 DV .avi files.

Then I did a visual comparison of the playback and the Pinnacle DV .avi file was far better than those from either Corel (Ulead) NLE.

Next, I took the Pinnacle DV .avi file and I converted it back to MPEG-2 at two different data rates... 8500 and 9800... using Pinnacle Studio Ultimate 11.

Next, I took the Pinnacle DV .avi file and I converted it back to MPEG-2 at 9800 in the Corel (Ulead) software.

The Pinnacle MPEG-2 output was superior at both data rates to the Corel output.

In the past, I found it difficult to believe that the output from one non-linear video editor could be so much better than the output of another non-linear video editor.

I was clearly mistaken.

The difference is quite stark here.

I'm concerned about this for Corel's sake.

I think I can prove to anybody at the company who is willing to do some simple tests that the output of competing software is somewhat better.

Kind of depressing to learn this.

Jerry Jones
http://www.jonesgroup.net

[etech6355]

Jerry,
Did you happen to have inserted "Advance=1" into the VS11+ uvs.ini file?
Curious, and actively running the program with this setting activated?

Referencing the -VR disk from the pana recorder it sounds like the ulead software is reading them correctly, the other software is attempting to join all the separate mpeg2 files on the disk (maybe why it's failing).

[Jerry Jones]

Negative.

I did not enable the "Advanced" encoding options.

You mentioned to me once -- as I recall -- that doing so will break VBR encoding in VideoStudio.

Why does the Pinnacle output look so much better?

Do they not rely on MainConcept for MPEG? Apparently they have their own custom MPEG technology.

Do they not rely on Microsoft for their DV codec? Apparently, they have their own DV codec.

The other thing that is impressive about the Pinnacle Studio Ultimate software is the effects.

There are some things that I have learned how to do in Pinnacle Studio Ultimate in a few minutes that I would have struggled to do in Corel MediaStudio Pro 8.

Pinnacle Studio Ultimate's titler is excellent.

The DVD menus -- which at first I thought were not as good as those in Ulead -- are actually pretty good; the reason they didn't look so good at first is due to the fact the previews out to my external monitor from Pinnacle Studio Ultimate are not as accurate as the previews out to my expernal monitor using Corel VideoStudio 11 or Corel MediaStudio Pro 8.

Previewing is where the Corel programs still seem to have a significant edge; the Corel previews are more accurate.

The Pinnacle previews look at little soft by comparison and it's a bit misleading because after you render out the timeline and then look at the actual results you learn the Pinnacle output is actually a lot better.

Jerry Jones
http://www.jonesgroup.net

Jerry,
Did you happen to have inserted "Advance=1" into the VS11+ uvs.ini file?
Curious, and actively running the program with this setting activated?

[Ken Berry]

The following comment is thrown in for what it is worth. I have refrained from getting involved in this thread previously since its main thrust had appeared to be mainly about HD video from mini DVDs... But now that it has wandered into SD video as well, maybe my observation may -- or may not -- have some relevance.

I had noticed similar slight blips after transitions were SmartRendered by VS9 (as it was what I was using at the time). This was two or three years ago. In fact it was getting so regular an occurrence that I abandoned VS9 for a Cyberlink competitor product for a while.

The video I was using was originally from a Canon mv430i standard definition 4:3 mini DV camera. I had expanded the timeline view down to the frame-by-frame size, and thus thought that the cuts between scenes that I was making were frame-accurate. I would continue with my editing, moving some of these cut clips about on the timeline, adding transitions etc. However, after SmartRendering, I would notice a momentary blip and it would seem (after repeated viewing) to include a frame or two from a clip which I had either deleted or was now elsewhere in the project.

After coming back to VS after the Cyberlink escapade, I realised that part of my problem may have related to the fact that I had been using the 'virtual' editing capabilities of VS9. In other words, the cuts I was making were only virtual, in the sense that they were recorded in the project file, but were not actual cuts in fact. And it seemed that, despite my thinking that my cuts were frame accurate, somehow or other, a frame from the scene immediately before the cut was being included in the clip in question.

I thus reverted to my original practice (developed with VS7 and 8 ) of always using Clip > Save Trimmed Video after each cut. In other words, creating a *real* new clip, rather than just a virtual one. After that, the blips disappeared whether using SmartRender or not.

Seeing the current discussion about I-frames would seem to hit the nail on the head of my earlier problem...

As I said, FWIW!!!

[Sektionschef]

Hi Ken
You have already posted this hint on page 2 of this thread(see here: http://phpbb.ulead.com.tw/EN/viewtopic. ... 4355#84355) and I had already confirmed here( http://phpbb.ulead.com.tw/EN/viewtopic. ... 4464#84464 ) that this does NOT fix the problem.
It seems that many people do not read this thread from the beginning.
They all think that this problem is only related to HD mpeg2 files but I have originally startet this thread and reported that it occours with SD mpeg2 files.
This is a terrible bug in Videostudio and it is present in perhaps all versions of videostudio, at least from VS9 on until VS11.
Regards
Sektionschef

[Ken Berry]

Ah! But this thread is already so long, and I have made so many posts on other threads since then, that I simply don't remember!! In fact, I would have to say that apart from one of the original stickies, this must be one of the longest threads I have seen on this Board...

[GeorgeW]

@Sektionschef,

I apologize that I am very late to this discussion -- I have seen it mentioned the issue occurs with Transitions, and am wondering if the same issue happens with TEXT clips (instead of Transitions)

I tried to duplicate this issue, and can see some "distortion" at the end of a Transition. I did not see the issue with TEXT clips (although I have not read all 90+ posts in this thread).

An interesting thing in some of my tests is when I placed a TEXT clip underneath a transition (making sure to have about a half second of text BEFORE and AFTER the start/end of the transition, I did not see the issue in the rendered file -- SmartRender was enabled). My TEXT clip contained only a black Underline -- and I placed it to lower left edge (overscan area) so it could not be seen during playback on a TV (hard to see it on the computer as well).

I have not run extensive tests using this procedure, but on a couple of tests it did seem to fix the transition issue. Basically, it's just making the program re-render a little bit more on each side of the Transition (and assuming the problem does not happen with TEXT clips -- which might be a false assumption on my part).

If you do try this procedure, please let us know whether or not it helps (it is not a true fix, just a possible bandaid for the moment).

Regards,
George

[Sektionschef]

Hi George
The distortions are generated with transitions OR with text(titles), but you will not see them with every text or transition element.
I doubt that your procedure will help here.
Please do more tests because the problem is not seen every time.
On page 1 of this thread(see here: http://phpbb.ulead.com.tw/EN/viewtopic. ... 1938#71938) I have posted a way that really works but it requires so much effords that it is unacceptable as a workaround solution.
I don't know of any other way to fix this problem except turning off smartrendering.
Again, this is a terrible bug in VS11(and older versions) that needs to be fixed by Ulead.
Regards
Sektionschef

[bsuratt]

sectionschef,

I tried a similar test as your method by capturing from Vegas7 which gave me separate m2t files for each scene, imported into Womble DVD and, using the I-Frames only mode, trimmed the scenes I wanted and saved each trimmed scene as a new file. Using the Womble file format converter I converted the trimmed clips into Program Stream MPEG format. I then imported the clips into VS11+, added titles, transistions music, etc then output to MPEG file. Then burned this to HD-DVD in MF6+. The resulting video looked clean and free of the "blips" around transistions. So, this would tend to point at the problem being the exact cut point needing to be on an I frame. This would seem to exonerate the MainConcept encoder as the source of the problem and point towerd the editors not handling the cuts correctly as being the cause.

Of course this workaround is too time consuming to be of any use for real editing needs, but it does prove that if the clips are trimmed properly then it will work.

Sad truth is, we will all probably have to pay for an upgrade before this is properly fixed!

[Jerry Jones]

I'm not so sure this releases MainConcept from any responsibility.

They market MPEG smart render products for Premiere and they also provide the software development kit (SDK) for Corel and Magix and others to develop.

As others have pointed out, this smart render bug is observed in both the Corel and the Magix programs except that the Magix program exhibits the problem at the beginning of titles, etc., while the Corel programs exhibit the problem at the end of titles, etc.

So wouldn't that suggest the software developers may be following the instructions of an SDK, but possibly are implementing it in slightly different fashion?

Jerry Jones
http://www.jonesgroup.net

This would seem to exonerate the MainConcept encoder as the source of the problem and point towerd the editors not handling the cuts correctly as being the cause.

[bsuratt]

Jerry,

My experience with Magix 12 Pro was the same as the others in that the "blip" was sometimes on one side of a transistion, sometimes the other, and sometimes both. You need at least 30 or 40 transistions in a test to see the full picture. (And, in some cases, the first time I viewed the video I did not see the "blip").

If the cuts are done properly prior to being fed to the encoder then everything is OK.

I do not think I would be happy having the encoder decide where to move cuts if that is how they do it. But, I don't know how it is managed.

At the time these products were release to beta there probably was little interest in extensive testing of mpeg smartrendering. Lack of an adequate testing program seems to be the achilles' heel for all of these companies. Testing cuts into profit and takes a lot of time!.

I have written MAGIX about the problem; received a nice reply.

I have written SpeedEdit; received no answer.

Unfortunately no manufacturer at this time seems to be actively looking for an answer.

[sjj1805]

Steve,

Apparently, you didn't get the memo.

This seems to be another example of you not being aware of what's going wrong with the software.

There's a bug in VideoStudio and in MediaStudio Pro, OK?

It's an MPEG-2 smart render bug........

There is no need to be condescending. Interestingly whilst I concede there is a problem, I can show you another test that suggests this has nothing to do with MPEG rendering. So whatever the problem is I guess its back to the drawing board. My own impression is that the fault lies in the software logic that creates the title. It is not a Smart Rendering issue either because in the following scenario there is nothing to be smart rendered - it is a collection of JPG images, Smartsound and a Title.

Take a look at these 3 screen shots.
The first shows a section of timeline comprising of static images taken with a digital camera. I have staggered these along Video 1 and Video 2 purely for the purpose of assisting me to synchronise the pictures with the smart sound on Audio 1. The Item on Video 3 is a title track where I am scrolling the text from right to left across the bottom of the screen.



The next screen shot is the last frame of the penultimate still image.
Note here the time code 0:00:23:03
Note the position of the text
[nt tourism Centre in Turkey]



Now look at the next screen shot which is the first frame of the last still image. Note the timecode 0:00:23:04 - this is the very next frame after the one in the screen shot above.
Note how the text has jumped back to an earlier position:



Therefore despite your unwarranted remarks I stick to my earlier statement that it is pointless converting MPEG2 to DV for the purpose of editing and then back again. The problems I have outlined above are all in DV type 1. Here are my project settings:



By the way if you want to know how I got around this problem OK it shouldn't be there in the first place, but it is.
I rendered out the slide show without the title, then placed the title on that new video. Works perfectly and didn't take long to do.