HDV --> SD DVD Workflow?

[hhansard]

Dont want to sound stupid or naive, but probably going to come across that way

I have shot [underwater] dv for years, and moved to hdv in 06. Since I have an end-to-end blu-ray setup with a broadcast scalier, my hdv footage looks great at home.

I was asked to take some of my hdv footage and burn an sd dvd for distribution. I took an edit of 1440x1080 and exported in PP-CS3 [movie] [Sorenson] as 720x480dv, no recompression highest quality (26G for a 30minute vid), etc, etc.. and imported it into a new PP3 d1 project. Burned the DVD again highest quality

I looked at the results on an sd 4:3 monitor terrible much worse then my old native dv footage. I expected to loose quality, but assumed that the scaling algorithm would smooth, and I would end-up with something acceptable --- but it does not look like it it looks like pixels are just dropped with no interpellation at all. I mean if you never saw the original footage, you might let it pass, but having seen the original footage, you can tell that the compression has killed it. I know that this is like a 2 (maybe 3) generation dupe, but I have access to the original pixels and would have assumed with minimal recompression/expansion the results would be as good as 1gen dv but I cant seem to get there

Can anyone point me to a good workflow [or some settings] to take hdv footage and cut a decent quality sd dvd using the production suite??

Thanks in advance,

Hugh

[_Wesley_C_]

>Premiere Pro CS3 by itself will never create decent DVDs from HDV material: edited, unedited, whatever...

Very subjective and arguable statement.

I have made many better-than-decent DVDs from HDV footage in PPro CS3 (no plugins) and have been told by many non-technical viewers that they think the video is 'beautiful' and 'so clear' and 'so much better' than what they had seen before (home-made SD DVDs for similar events).

The workflow is simple and has also been outlined in this thread by others, and the ultimate result is more time for editing/shooting and less time tweaking in and out of third party apps.

[Dan Isaacs]

> Very subjective and arguable statement.

Sure it is. Better/worse comparisions are ALWAYS subjective, but less so when you provide evidence:

http://invertedhorn.axspace.com/ppro_resize.html

It is also subjective to say that DVD looks nicer than VHS, but I think most would agree with that statement. Heck, it's subjective to even say that HD is better than SD. To each his own.

There are inherent technical limitions to the way in which PPro handles these conversions. These are factual, whether you believe that they produce "worse" results in the end or not.

> ...many non-technical viewers that they think the video is 'beautiful' and 'so clear' and 'so much better' than what they had seen before...

Now THAT is truly subjective.

> the ultimate result is more time for editing/shooting and less time tweaking in and out of third party apps

I can only speak for my own method, which adds maybe 5 minutes in overall "tweaking time" once you've established the basic process, so... there's a trade-off, I suppose. Quality vs. time required. I can spare 5 minutes. Maybe your schedule is very tight, so that's your choice :)

Again, it seems the purpose of this thread (from its inception) is to find solutions for those who are unhappy with the "standard" PPro/AME/Encore methods. If you're happy with your results, I certainly won't try to change your mind.

[_Wesley_C_]

Dan - I think your efforts serve yourself and many others well, but I would only suggest that statements such as that which I quoted...

"Premiere Pro CS3 by itself will never create decent DVDs from HDV material: edited, unedited, whatever..."

...is way too definitive/dogmatic for such a subjective issue.

It is not a 5 minute time versus quality issue for most people, it is hours and hours of getting it worked out to the degree that you have, let alone following all of the esoteric instructions provided. When dozens of people say that a HDV-sourced DVD looks wonderful, fantastic, incredible, wow, etc etc... yes it is subjective (I really don't think I said it wasn't), but it is also proof that it is GOOD (good is = to or > decent).

I do more photography than video myself (though I do a LOT of video) and I can tell you there is a whole segment of the photographic universe that pixel peeps every photo they take. Bottom line is that you can always find fault, but put a nicely composed, well focused (etc. etc.) photo in front of the intended audience and they will not be pixel peeping on their computer screens, they will be oohing and aahhing over a very nice print just like they do with video.

Perfectionism has its place to those that choose to achieve it, but most people will never notice or appreciate that you have achieved it.

WC

[Dan Isaacs]

> yes it is subjective ... but it is also proof that it is GOOD

That is not proof.

> it is hours and hours of getting it worked out to the degree that you have

OK, but if it takes a few hours to substantially increase the quality of every such conversion you perform in perpetuity, then it seems worth it to me. I have spent "hours and hours" on things I consider far less fruitful.

I suppose I have a different attitude about this stuff: I am a professional (this is my job) and I've been doing this for quite a long time. Learning new things and striving to improve the quality of my work is a good part of what I do and what I'm paid for. This does involve some trial-and-error and therefore time. There is no way around this.

> let alone following all of the esoteric instructions

If there is something that needs clarification please let me know. I have tried to do this for others all along as they've run into problems or areas of confusion. I may be "dogmatic" but let it not be said that I am unhelpful.

[Dan Isaacs]

> Bottom line is that you can always find fault ... the intended audience and they will not be pixel peeping

Pixel-peeping? There's nothing nitpicky about this. PPro does a terrible job at resizing interlaced content. I believe the example I provided demonstrates this quite clearly. It is not a subtle difference: No "pixel-peeping" required.

The photography analogy is not really relevant. When viewing a print from a high-resolution digital image, you will never be able to see the original pixels -- not even with a loupe. The same cannot be said for digital video: 720x480 on a large TV will reveal all kinds of horrors.

In truth, I provided a very mundane example. It is just a comparison of still frames (from a segemnt of video in which no action is taking place). Motion video looks even worse when scaled down in Premiere. Its effect on titling and graphics is horrific.

Some people may not notice the difference, true enough. These are probably the same people who would be unable to notice the difference between HD and SD... or YouTube vs. DVD... or lossy JPEGs from a digital point-and-shoot vs. TIFFs from a Nikon D100.

[_Wesley_C_]

>Pixel-peeping? There's nothing nitpicky about this. PPro does a terrible job at resizing interlaced content. I believe the example I provided demonstrates this quite clearly. It is not a subtle difference: No "pixel-peeping" required.

>The photography analogy is not really relevant. When viewing a print from a high-resolution digital image, you will never be able to see the original pixels -- not even with a loupe. The same cannot be said for digital video: 720x480 on a large TV will reveal all kinds of horrors.

I guess I see the analogy as relevant, or I wouldn't have used it.

>No pixel peeping required

Looking at still frames from video at 200% and comparing pixels on a computer screen sure seems like pixel peeping to me!

When I am watching a well produced, entertaining, or for whatever other reason I am into the subject/purpose of the video, I do not pause and zoom to 200%, nor do I scan it with my eyes for the small imperfections. I enjoy the video and I also notice how much better it looks than it used to... with "used to" meaning when I "used to" use SD video source to produce DVDs.

>If you're happy with your results, I certainly won't try to change your mind.

I thought we were letting our discussion term when I didn't respond to your last post, but I guess not...

BTW, Nikon sucks, so bad example... ;-)

[Jeff Bellune]

1. Apply ReVision:FieldsKit to the clip to deinterlace.

2. Nest sequence containing the effected (sic) clip into a new sequence.

3. Scale the nested sequence.

Does that produce better results?

EDIT: I didn't go back and read all 180+ posts, so if this solution has been offered and accepted/refused, I apologize.

[Dan Isaacs]

I don't know, Jeff. I haven't tried it. I am certain that FieldKit provides better deinterlacing than PPro (how could it not?).

The only problem I see with this suggestion is that you're scaling it with PPro after deinterlacing it. Several things come to mind:

1.) deinterlacing -> scaling is the correct method, but does Premiere still interpret your nested sequence as "interlaced"? If so, the same problems would likely result.

2.) This method would only output progressive content (I think?). The goal of such conversions (in most cases) would be to maintain both the spatial and temporal resolution of the original. There is the "reinterlacer" from FieldsKit, but this would add extra time and the "new" fields would be completely interpolated/estimated -- kind of like doubling the frame count with "Pixel Motion" using the Time Warp filter. So I think... I haven't used it so I can't say for sure.

3.) The tools I suggest are free and fast. I don't know how fast FieldsKit is, but it is not free. It's not that expensive either, however.

4.) There are still the colorspace conversion issues. I know ProCoder has a 709->601 filter (and may even do a decent job on HD->SD conversions in its own right, I don't know) but it is expensive.

If you have FieldsKit and have done these types of conversions I'd love to see your results. I am always interested in different methods of doing anything. I'd also like to know how long a sequence takes to render with FieldsKit applied.

[Jeff Bellune]

FieldsKit renders slowly.

One notion behind a nested sequence is to force a render order, much like a pre-comp in AE. The clip inside the nested sequence should be rendered (thereby deinterlacing the clip) and then the Motion effect in the parent sequence should be applied to the (rendered) nested sequence.

Since you can't alter the field order of, or deinterlace, a nested sequence, the problem of the sequence being "interlaced" should not be an issue. The parent sequence should only see the clip the way that the nested sequence hands it off - deinterlaced.

Using FieldsKit in PPro would eliminate the round trip out and back in to PPro that your workflow (if I understand it correctly) requires.

If you have the time and feel so inclined, please try the demo of FieldsKit and see if it produces results that meet with your approval, including render time and output quality.

I'm not trying to get you to change your workflow, I'm just providing alternative methods for others. It would be enlightening for all to see if the alternative method works well in the frame of reference of your critical eye.

[Dan Isaacs]

> The parent sequence should only see the clip the way that the nested sequence hands it off - deinterlaced

Right, but I'm still not sure. If you're project is setting is "interlaced", PPro may still scale it as if it were interlaced. Which brings me to...

> If you have the time and feel so inclined, please try the demo of FieldsKit

I will. What are the limitations of the demo?

> ...the round trip out and back in to PPro

Actually, I never bring it back into PPro. I use other apps to encode, then drop the converted files into Encore or whatever. If you wanted to, you could use my workflow to create an uncomressed .avi from VirtualDub and let PPro or Encore do the encoding. That's a viable option.

Try out the DebugMode frameserver plugin for PPro if you haven't already. It will open a whole new world transcoding possibilities -- whether you use FieldsKit, Cineform, ProCoder, etc. or not. It will circumvent the need (in most cases) for large .avis of your sequences which is especially useful with HD.

I'd love to change my workflow: Give me something that's cost-effective, fast, provides great results and complete control and I'm there!

[Jeff Bellune]

>Actually, I never bring it back into PPro. I use other apps to encode, then drop the converted files into Encore or whatever

Then how do you scale the deinterlaced video if you never bring it back into PPro? Don't you ever do PIP/Split-Screen effects? Or is the only scaling you ever do the downconversion from HD to SD?

>Try out the DebugMode frameserver plugin for PPro if you haven't already. It will open a whole new world transcoding possibilities -- whether you use FieldsKit, Cineform, ProCoder, etc. or not. It will circumvent the need (in most cases) for large .avis of your sequences which is especially useful with HD.

Obviously you haven't seen my tutorial in the Encore tutorials section. :)

EDIT:

>What are the limitations of the demo?

Watermarked output with an "X" across the frame.

[Dan Isaacs]

> Don't you ever do PIP/Split-Screen effects?

(Oh, OK. Sorry. I guess I missed your point)

Truthfully, I don't use them too often. When I do, sometimes I will pre-render it via AviSynth if quality is critical. Other times I just let PPro handle it -- quality be damned.

I often work with animation myself, so sometimes I'll render stuff at 60fps, scale it then weave it. If memory serves me, AfterEffects will do this if you output interlaced 30fps from 60fps sources.

In any case, I don't ever export from Premiere and import back in that I can think of.

As I mentioned (quite early on in this thread) most of the editing I personally do is SD progressive-in/progressive-out for PC/web playback. Sometimes I make an interlaced version for DVD distribution.

In this thread I'm really only focused on the HD->SD thing, however.

> Obviously you haven't seen my tutorial in the Encore tutorials section

Nope. I'll check it out though. Thanks, Jeff.

[Dan Isaacs]

Hi Jeff --

I did play around with FieldsKit. Is seems like a good deinterlacer, but it is incredibly slow -- even with fairly modest settings. The "blend" option does not provide nearly good enough results considering the time it takes to render. The "Best 5 neighbors" method is quite good, but even slower.

First off, many of its most useful options are not very useful in PPro: Although you can double the framerate in AfterEffects, PPro does not allow sequences of varying framerates in the same project. You can also double the frame count, but Premiere will not allow to increase the clip length to any greater than the original. For similar reasons, I was unable to find any use for Re-Interlace filter (which also seems much slower than it needs to be).

When I nested the deinterlaced sequence in another sequence and scaled it, I was still able to see significant stepping artifacts -- just as if no deinterlacing had been applied. My guess is that PPro assumes the sequence to be interlaced because the project is interlaced. Similarly, exporting the deinterlaced sequence via AME also treats it as if it were interlaced.

Scaled output with FieldsKit does appear to look a "little better" (less stepping)... but this is most likely just a result of the blending caused by the deinterlacer. I was actually able to get less stepping without the deinterlacer just using PPro's Anti-Flicker filter at its maximum settings (at some expense of sharpness).

In my opinion, the FieldsKit deinterlacer is also "tricked" by noise rather easily. Even using the various mask options I was unable to mitigate this to my satisfaction. To be fair, this is a problem for all adaptive deinterlacers. For this reason I usually perform noise reduction prior to deinterlacing, but FieldsKit is so slow by itself that this would become highly impractical.

Another thing I'm not too crazy about is that the FieldsKit deinterlacer only works in RGB mode. In other words, once it is applied your YUV footage is converted to RGB, thereby clipping super whites/blacks. I believe this can also reduce the effectiveness of an adaptive deinterlacer as it is unable to "see" interlacing artifacts in areas with bright highlights. Even though they're clipped-out in the end, these highlight-flickers can provide "hints" that surrounding (non-clipped) pixels may require deinterlacing as well.

From what I can see, FieldsKit does not benefit HDV->DV conversions at all. I even nesting the FieldsKit-processed project in a 1080p project but the results via AME were not better than exporting directly from the 1080i project.

Sorry, Jeff. I did try 🙂 Perhaps better results can be attained but the preview/render performance was so slow that I lost patience after about 2 hours of playing with it.

Now that I know you're a DebugMode FrameServer fan, I think you should give AviSynth a try. It is very useful for post-processing output from Premiere. Check out some of the available plugins (notably SmoothDeinterlacer). I think you'll be quite impressed with both the quality and performance.

-- Dan

[Dan Isaacs]

A link to another page of workflow comparisons (I posted this in a different thread, [ http://www.adobeforums.com/webx/.59b50c10/25 ] but I figured it should be included here also)

http://invertedhorn.axspace.com/ppro_resize2.html

[_james_bruce_]

I recently got my HDV timeline out to DVD and the result looked really good. After many tries, here is my workflow. the main flaw up until this one was that all footage looked soft and slow mo clips jogged and sputtered (just did not go slow).

Total project, 1 hour 4 min.

I had kept each section in a different project so that I would not have one giant project that would take so long to open. Each project was about 4mins.

1. Captured HDV footage from sony z1-u using Premiere CS3 Capture.
2. Split scenes using HDV Split. (optional step)
3. Edited each section using CS3 as a HDV 60i widescreen project.

Here is the crucial step

4. Exported each section to Widescreen ntsc as a dv-avi file.
5 Created a widescreen dv project in encore.
6> Imported each section into the project as an assett, placed on timelines added menus and then rendered out to DVD.

The output of this process looks great and saves the step of sending all footage out to the camera and reacapturing as DV.

Hope that helps

james

[mickkeay]

Interesting James. That workflow wasn't so good for me. In step 4 I assume you mean Export Movie / Microsoft DV AVI as the setting? What were the other settings - did you de-interlace then? Was the HDV footage quality top notch / well lit? I have found that it is the poorer quality footage that suffers in the adobe work flow from HDV to SD - especially if one has used things like color correction on the timeline. Glad it worked for you though.

[Dan Isaacs]

Hi Mick --

> I have found that it is the poorer quality footage that suffers in the adobe work flow from HDV to SD - especially if one has used things like color correction on the timeline

That's interesting. I can't say I really understand why that would be. Can you descibe exactly how they look worse (color, noise, stepping and statial distortion, motion artifacts)? Does the same "bad" footage look "better" without color correction applied?

What I've noticed is that sharper clips are most problematic (such as those with thin diagonal lines). High motion also gets a bit funky looking. Titling (especially animated titling) often fares poorly as well.

Again, It would be great if you (or anyone else) would post some clips and/or frame grabs of your workflow results -- and the exact steps/settings you used.

Cheers,
-- Dan

[mickkeay]

Got me thinking now Dan. The footage I most disliked was thin lines (a tasteless patterned shirt for example) but it was also shot under too warm lighting - so I cooled it on the timeline and lept to the conclusion that that was the problem cos other footage without the correction was OK (well, better). But thinking about it the ok footage was a different scene and less 'liney' but on the poor stuff there were plenty of lines and text in backgroud posters etc (it was a guy talking to camera) and they didnt fare well. I eventually rendered out with debug frameserver and avisynth using one of your scripts (thanks Dan) and then into procoder for SD MPEG. I deleted the adobe only renders but will see if I can re render a minute or so both ways and re-compare

[_David_B_Meed_]

I've read this entire post and installed the frameserver/avisynth/hcencoder setup.

Have an email in to Dan Isaacs requesting the updated scripts.

Video issues are fixed with the scripts listed in this thread and the quality is much better than the PPro export.

Question is, how do I get audio file to go with that video?

I'm using a HDV 1440x1080i preset (source is a Sony HDV camera) for editing in PPro CS3 on a PC.

I'm trying to export the edited video to dvd for folks with SD. (We'll keep the HD edit around for when they upgrade to Bluray.)

I can create the m2v file using what is listed in this thread, but I'm missing any audio file to add to that video file to create my disk in Encore...

Also - what should the settings be in the movie export window - there is a "recompress" button - should that be unchecked.

Thank you

David Meed
david at - meed dot ca

[_Kevin_Haga_]

I also read this entire thread (and many, many others) in an attempt to get the best SD DVD results from my Canon HV20 HDV camcorder.

I have Adobe PP3 (3.2.0 on Windows XP) and have tried multiple workflows, and have spent many, many hours (even on a Quad core PC) creating different samples with widely different resulting DVDs.

Some results to share:

First, I found that the best Adobe-only PP3/AME workflow (as has been discussed on this thread) is to capture your HDV footage in an HDV 1080i project and save the clips; create a new SD 16:9 project; import the HDV clips into the SD workflow; motion scale 45%; do your edits in the SD workflow; encode with AME MPEG2 2pass-VBR, Quality=5, min/avg/max bitrates set at 6/7/8. This results in a DVD that looks good for your average home movies taken with an HDV camcorder and played back on a regular NTSC 4:3 television.

HOWEVER, the above Adobe workflow does NOT result in a DVD that looks too good when played back on my 1080p 50" HDTV... This was what got me searching for something better!!

I read the instructions presented by Dan Isaacs in this thread and installed AviSynth 2.5.7, DebugMode Frameserver, VirtualDub 1.7.8, and I purchased the CCE Basic MPEG encoder ($58 online).

After getting the sample script working that Dan provided, I then got his advanced script working (simply more options) as well. The results from this encoding workflow really blew me away when played back from a regular DVD player (without upscaling) on my 50" 1080p HDTV - WOW!!

The DVD output from Dan's AviSynth workflow wasn't too much different when shown on my regular "tube" 4:3 TV - it was the HDTV that really showcased what this workflow does for you. I wish I had an upscaling DVD player - I expect the results to be stunning when I get one!

Some comments:
1. one good thing about the Adobe PP3/AME workflow is that it fully exploits my Quad core PC, which saves a huge amount of encoding time - it will drive all 4 CPUs for hours at 100% each. This approach is great for encoding movies that you need to get out on DVD fast (relatively speaking). On my system, I calculate that the Adobe-only workflow needs about 12-15 mins to process each minute of video footage (e.g. a 40 min video can take 10 hours to process using 2p-VBR) with PP3/AME. The same footage takes about 30-40 mins per minute of video footage using the AviSynth/CCE workflow (e.g. a 40 min video will take about 20 hours to process, but only uses 30% of my Quad core CPUs).

2. for the AviSynth/CCE workflow, you can test out the workflow using VirtualDub to preview the AviSynth (.avs) file and test out the various parameters. Bottom line - if you can't see the video in VirtualDub, it isn't going to work in the CCE Basic encoder.

3. I found that installing AviSynth, DebugMode FrameServer, VirtualDub, and CCE Basic ($58) was much easier and more straightforward than I thought it would be. AviSynth uses several plugins for processing the video - these take a bit more care to make sure you have the right versions and they are in the right places (paths). There are many good examples of what parameters to use, so just experiment to see what works best for you. With VirtualDub, you can experiment - it encodes 1 frame for you to preview, which makes it fast to see the results!

4. A side benefit of having DebugMode Frameserver installed is that you can frameserve directly from Adobe PP3 to other encoders without creating an intermediate file. I frameserve directly to Sorenson Squeeze 5.0 for compressing high quality web videos from HDV, and I save 2-3 hours by frameserving instead of creating a huge intermediate file.

5. Playing around with a couple of the settings is required (both in PP3 and in the .avs scripts - I have a Canon HV20 and have found that you want to frameserve from the HDV project with "Upper Field First" set in the Export dialog in PP3, and use the OutputBFF as a parameter in the AviSynth parameters file for Dan's advanced script. Don't oversharpen to much. This will vary by camera, etc - I tried different combinations to get the best result for a DVD playing on my HDTV.

6. With the AviSynth/CCE approach you can frameserve directly from the PP3 HDV workflow - you don't need 2 projects as you would with the Adobe-only workflow (which requires one project for HDV capture; one for SD 16:9 output to the frameserver).

7. to answer an above post - the AviSynth/CCE workflow is best for creating the MPV (MPEG) files. I just Export Video (no audio) to DebugMode Frameserver to get the MPV file from CCE Basic. Then I Export the Audio only from Adobe PP3 (directly out) to get a Wave file (this is quick). Load the .MVP and .WAV files into Adobe Encore to author the DVD - no further compression or encoding will take place, so it just writes the files (plus your menu, etc) to the DVD.

My point is - both of these workflows have their place (speed with passable quality versus high quality), but if you want high quality SD video from your HDV camcorder, take the time to learn AviSynth, etc. - the results are worth the effort by far, and you get the side benefit of learning more tricks (like frameserving).

A big huge thanks to Dan for his contributions to this forum, and to my education!!

Kevin

[mickkeay]

I second that Kevin! Dan is a HDV-SD star! Real solutions with working scripts.

It took me a while to work it out and I know a lot of GUI only newbies have been put off by the scripting side of it - A few camtasia style web videos would probably point many more in the right direction - or a step by step guide - if anyone has the time and appetite?

[_dimas_Indra_]

Hi, i'm new here
I'm a rookie so this is my question
I have MXF files that I shoot using sony XDCAM HD 355 with 60 fps,(slow motion)
but why my aspect ratio is 1440x540 and not 1440x1080?and how can I edit? when I import with XDCAM project on premier cs3.2 my picture is very small...
please help....

thks
JON

[Jim_Simon]

Your question is not really related to this thread, Dimas. You should start a new topic in the right forum.

[Guest]

> MickKeay wrote: A few camtasia style web videos would probably point many more in the right direction - or a step by step guide - if anyone has the time and appetite?

Second that! Would really be apreciated!
Anyway, thanks for all the useful input in this thread, and hopefully we will see more.

[_ian_clubb_]

I too have experienced the problems of HDV to DVD and found the results to be very disappointing. I have downoaded the 15 day demo of Cinform Prospect in an effort to get better results but the initial results have not proved any better than PPCS3 (3.2) ot Matrox RTX2.

I would be very interested in trying Dan's work around but i would need detailed info from Dan as I know nothing about AVISynth.

Dan, can you give some guidance.

Ian