Render project to DV first or straight ot MPG?

[kcwdad]

What generates best end DVD quality result (where neither time nor disk space is a concern):

1) Rendering MSP8 project (w/ DV source) straight to MPG2, or
2) Rendering project to DV, then converting DV to MPG2?

Thanks!

[Helge]

Theoretical there should be no difference, if you use identical settings for mpeg coding and the msp8 encoder. I ususally use the first way, except if I think I could reuse the DV output. Using the 2) method can have the following advantages:
a) may be faster, if you have a lot of tracks and / or timelines and want to do a two pass render
b) you could use non-ulead mpeg encoders (e.g. if you own a very expensive high quality multi pass encoder)
c) if you not only want to use the output as final result but also as further input for an other video (e.g. generate trailer, first play ...)

But:
I noticed that rendering the same input two times the same way (with smart rendering turned off) did not result in binary identical videos.

[kcwdad]

Thanks--very helpful answer.

[Gorf]

Generally speaking, if your DV file would need a lot of rendering (PiP, filters, transitions etc) you're better off going straight to MPEG.

[Devil]

There is a slight ambiguity in the question.

If you "capture" MPEG-2 from your DV source (transcoding), then the quality will definitely be poorer. If you then edit it on the the timeline, then the quality will be worse again, if any re-encoding is needed.

Also, if you wish to convert from NTSC to PAL or vice versa, then this should be done in DV AVI and not MPEG for best results.

[sjj1805]

Bearing in mind the comments given by Devil. think of Video production is a 3 stage process and treat each one separately.

Stage 1. Capture.
Getting it onto your Hard Drive. Preferably in a non lossy editable format
such as DV (avi).
Stage 2. Edit.
Make your cuts, joins, Titles, PIP's, Additional sound etc. Save as a project file so that you can return to it later for further alterations.
From this create a DVD compliant MPEG2 File ready for stage 3.
Stage 3. Author.
Create your DVD Menus and burn to disc.

[Jerry Jones]

In answer to the first question in the thread, rendering directly to MPEG-2 would yield higher quality output -- for some but not all frames.

The reason has to do with digital signal processing.

Anytime you apply...

1. a title;

2. a transition;

3. a filter.

...you may be happy to learn that -- on the timeline of Ulead MediaStudio Pro -- all three are applied in UNCOMPRESSED space.

A lot of people think that as soon as you apply a title or a transition or a filter, that DV compression takes place.

This is incorrect.

Strictly speaking, even the process of rendering has nothing to do with compression. It's merely a mathematical computation performed by your processor.

However, each frame rendered is then fed to your codec for compression at the moment when you write the timeline to a file or export to tape.

What about non-changed segments of your timeline?

Those frames are simply COPIED as data -- with no quality loss -- when you write the new DV .avi file or -- if you go directly to MPEG-2 -- they are subjected to MPEG compression.

This assumes you have Smart Render enabled.

This is why Smart Render technology is so great.

It essentially means that DV video on your timeline that isn't changed by the editing process is protected from any further DV compression.

If you render and then compress a title to the DV .avi file format first, you are subjecting that rendered segment to one pass of DV compression.

If you instead render directly to MPEG-2, you avoid the DV compression and you render the UNCOMPRESSED title, the UNCOMPRESSED transition directly to DVD-ready MPEG-2 and you spare the DV video (already compressed the first time in the camcorder) from another round of DV compression.

This is why rendering your timeline directly to MPEG-2 results in higher quality for some frames.

Now, if you're talking about frames that don't require rendering, e.g., DV video frames with no changes applied in the timeline, then it doesn't make any difference whether you write out a DV .avi file or render directly to MPEG-2 because those frames -- in the end -- will be subjected to one pass of MPEG-2 compression.

There is also a great deal of misunderstanding about MPEG-2 editing.

MPEG-2 editing can work very well.

Smart Render -- again -- is the key.

If you begin with very high quality MPEG-2 files from a good standard definition DVD camcorder or HDV files from a good HDV camcorder, you can spare most MPEG frames on your timeline from compression.

The Ulead software -- when editing MPEG-2 -- caps off each edited segment with newly-created I-Frames that sandwich the I, B, and P frames to form a valid "Group of Pictures" (GOP).

The size of the GOP -- and the small number of frames affected -- can depend on the type of camcorder you're using.

When you're editing MPEG-2, you can apply titles, transitions and filters.

And make no mistake.

When you apply these items in the timeline, you are doing so in UNCOMPRESSED space.

Is compression coming?

Sure, when you write out a finished HDV or standard definition MPEG-2 file, your edited GOPs will be fed to your compressor and subjected to one pass of MPEG-2 compression.

Your unchanged MPEG-2 source frames will not undergo ANY further compression!

Those frames -- probably the vast majority of your timeline -- will simply be COPIED AS DATA as your new HDV or standard definition file is being written to your hard disk or when your timeline is exported to tape.

MPEG-2 (or HDV) frames that are touched by edits will undergo one pass of MPEG compression during export to tape or during the writing of a new file on your hard disk.

Because they've already been compressed once by the camcorder, video frames in the edited segments will be compressed again and -- in this respect -- these are the frames that will undergo more compression than their DV counterparts.

During the rendering of any title, transition or filter-applied segment, there will be a small amount of quality loss due to the limited precision of the mathematics utilized to compute the content of each new frame, but again, this is only a very small quality loss.

Some people think it's better to transcode to uncompressed and then edit.

Well, think again.

The process of transcoding results in a small amount of quality loss.

Unlike the native editing of MPEG-2 and DV, transcoding to uncompressed affects every single frame... and you have to do it TWICE.

So rendering to uncompressed and then editing and then rendering back to the output format is no magic solution for avoiding quality loss.

I would only use that method if I were editing MPEG-4 files from the Sanyo HD1a, for example.

Jerry Jones
http://www.jonesgroup.net

[Jerry Jones]

I would add that it's a good idea to render out a DV .avi version of your timeline if you can't keep the original project file and all of your original source media due to hard disk space limitations.

Keeping a DV .avi copy is a good way of preserving a relatively high quality copy of your project.

Jerry Jones
http://www.jonesgroup.net

[Jerry Jones]

Also, if I've stated anything that's incorrect, I'm sure H.T. will jump in.

But I think I've written about this topic correctly because I've been discussing this topic in detail with an engineer who specializes in digital signal processing.

One thing that I would like to know -- if H.T. knows the answer -- is whether Ulead MediaStudio Pro applies titles, transitions and filters in 4:4:4 color space or in a lesser color space.

In this respect, Apple's Final Cut Pro is interesting because that application can natively edit HDV -- like Ulead MediaStudio Pro 8 -- but that application allows for application of titles, transitions and filters in uncompressed 4:4:4 color space.

You can read about native HDV editing.

There's an excellent PDF PowerPoint presentation utilized by APPLE to explain the process of native editing.

By the way, Brett Halle is the senior director of pro video applications ***engineering*** at APPLE.

Halle is not a "marketing guy" out to impose a "marketing ploy" on a gullible public.

Halle gives the following PowerPoint presentation about the advantages of the native long GOP workflow over transcoding to intermediate formats:

http://tinyurl.com/y6hyzt

This is a good PowerPoint presentation.

Page 9 explains GOPs, for those not familiar with that topic.

Page 20 shows how APPLE's engineers measured PSNR (peak signal to noise ratio) for both *native* and *transcode* workflows.

(The native workflow approach wins.)

I would suggest that, if anything, the "marketing hype" has been the weakness of the arguments in favor of transcoding HDV to intermediate formats for editing.

I wish Ulead could hire a good artist to diagram how native editing works.

Right now, all Ulead has really done to explain the subject is here:

http://tinyurl.com/mevw2

But native editing is now adopted by Ulead, Apple, Adobe.

Avid supposedly also does native editing, but in a slightly different way.

Only Sony seems to be hanging on to the transcode method for now.

Jerry Jones
http://www.jonesgroup.net

[neonbob]

Getting it onto your Hard Drive. Preferably in a non lossy editable format
such as DV (avi).

Contrary to popular belief... DV IS lossy... Probably more so than mpeg because of Smart render.

On another board we did a hands on experiment and rendered 10 generations of mpeg vs. DV... the mpeg faired better because Smartrender took hold and did not completely render the mpeg over and over.

If DV is what is in your cam then it is best to stay DV for as long as possible, so long as you do not plan on multiple renders. However if you originate as mpeg, then you should stay in mpeg. One should keep in mind that there is a definate loss involded by just simply converting, and therefore should be done as little as possible.[/img]

[Devil]

I agree DV is lossy, as is any compressed system. I disagree wildly with your premiss that MPEG-2 is less lossy. You are comparing apples with guavas.

If you wish to compare the "lossiness" of different systems, you must do so under identical conditions. In your case you choose to use Smart Render for the MPEG but not for the DV. This is grossly unfair. For such a comparison to be valid, you must use Smart Render for both or neither.

Donkey's years ago, I did a DV re-rendering to the 20th generation (without smart rendering). I forget the actual figures - it was done over 10 years ago when I bought my first mini-DV - but I made a comparison of each generation. I can assure you that, viewed carefully on a TV, there was no discernible difference up to the 5th or 6th generation. By about the 10th generation, some small artefacts were beginning to become visible on certain scenes. By the 20th generation, I agree that degradation of the image quality was very visible.

I have not done a similar systematic trial with MPEG-2, but I do remember that I edited a version of DVD that was burnt at 6000 kbit/s CBR, converted to MPEG-2, a lower-third added and reburnt at 6000 kbit/s CBR (IOW first generation re-render) and there was already a very visible deterioration in quality, visible on any TV.

And Smart Render does work on DV, as well.

[neonbob]

I agree DV is lossy, as is any compressed system. I disagree wildly with your premiss that MPEG-2 is less lossy. You are comparing apples with guavas.

If you wish to compare the "lossiness" of different systems, you must do so under identical conditions. In your case you choose to use Smart Render for the MPEG but not for the DV. This is grossly unfair. For such a comparison to be valid, you must use Smart Render for both or neither.

Donkey's years ago, I did a DV re-rendering to the 20th generation (without smart rendering). I forget the actual figures - it was done over 10 years ago when I bought my first mini-DV - but I made a comparison of each generation. I can assure you that, viewed carefully on a TV, there was no discernible difference up to the 5th or 6th generation. By about the 10th generation, some small artefacts were beginning to become visible on certain scenes. By the 20th generation, I agree that degradation of the image quality was very visible.

I have not done a similar systematic trial with MPEG-2, but I do remember that I edited a version of DVD that was burnt at 6000 kbit/s CBR, converted to MPEG-2, a lower-third added and reburnt at 6000 kbit/s CBR (IOW first generation re-render) and there was already a very visible deterioration in quality, visible on any TV.

And Smart Render does work on DV, as well.

Yes... I will admit that is a slightly unfair comparison.... but then on the other hand so is yours towards the bottom of your post... you can not do comparisons with dv and mpeg at 6000 under too many circumstances. If it was possible to drop the bitrate of dv to 6000 THEN compare... it might be fair, but as we all know... dv bitrate is fixed. It would however be interesting to compare mpeg at 25000 to dv.

[Devil]

you can not do comparisons with dv and mpeg at 6000 under too many circumstances. If it was possible to drop the bitrate of dv to 6000 THEN compare... it might be fair, but as we all know... dv bitrate is fixed. It would however be interesting to compare mpeg at 25000 to dv.

Yes I can do such comparisons. When shooting in DV, then 6000 kbit/s to burn DVDs or for direct transcoding from DV to MPEG is absolutely normal and re-rendering this will produce a deterioration of quality, much, much more than re-rendering the original DV.

You are being unfair with your comparisons again by introducing non-sequitur 25000 kbit/s which no one, to the best of my knowledge, uses for DV work. The OP specifically mentions DV. In any case, the GOP is probably more important than the bitrate. A 6000 kbit/s GOP of 1 (all I frames) will not be pretty to look at, but will re-render with very little loss. A GOP of 12 or 15, is usual, very acceptable in visual quality, but poor at re-rendering. Double that GOP and you will have severe problems in re-rendering. In any case, Main Level MPEG-2 (non-DVD) has a max of 15000 kbit/s for an SNR/Spatial Scalable Profile or 20000 kbit/s on all layers for a High Scalable Profile (less for fewer layers) For non-scalable profiles in simple and main, the limit is also 15000 kbit/s, so your premiss is not valid. To reach 25000 or more, you're talking about 1280 x 720 or higher capability, mostly progressive (High Level), which is a totally different kettle of fish. In terms of DV, which this thread is all about, you are now comparing apples to bell peppers.

[neonbob]

Yes I can do such comparisons. When shooting in DV, then 6000 kbit/s to burn DVDs or for direct transcoding from DV to MPEG is absolutely normal and re-rendering this will produce a deterioration of quality, much, much more than re-rendering the original DV.

Oh please....

That's NO comparison at all!

DV contains MUCH more information than mpeg at 6000 REGARDLESS to gop.

If you remove $1 from a million do you notice.. maybe, maybe not. But if you remove $1 from $10... it surely WILL be noticed.

There is simply NO comparison to the quality of mpeg at 6000 to that of dv to START with. How can you even THINK of comparing DV to mpeg at dvd compression levels!!?? It may be absolutely normal to shoot in dv and transcode to mpeg 6000.... but that's not AT ALL what we are talking about here. Nor are we talking about doubling gops. And as far as Iframe gop is concerned, I can show you a document that states some pros actually choose to shoot in dv and edit in low compression Iframe only mpeg. (This document by the way does mention that mpeg @ 25000 IS comparable to DV.)

But I think MOST of all what you are not seeing is that comparing mpeg with DV under ANY conditions is the same as comparing "apples to bell peppers". So in order to do even discuss comparisons, you must be able to bend a little. Where that bending point is, I am not sure, but I DO KNOW that it is NOT at dvd compression levels.

Oh and no... this thread is NOT about DV... it's about WHEN to go from DV to mpeg.

[sjj1805]

... this thread is NOT about DV... it's about WHEN to go from DV to mpeg.

I hope the post I just placed Here might go a long way to answering that point!