HD to SD downconversion

barfnuts9000 wrote on 11/29/2012, 1:45 AM
I know that the topic of having multiple types of target media for the same project (HD BluRay and SD DVD) is a bit of a touchy subject for some people -- in terms of how you can reliably get optimal video quality results for both BluRay and SD-DVD media when shooting in interlaced formats..

When I shot a wedding for a friend a few months ago, I was oblivious to the fact that there was such a painstaking process for shooting for multiple formats. I assumed (wrongly) that if you shot in HD, then downgrading to an inferior format would be a piece of cake. Seems that I thought wrong.

Editing and rendering the master format of the native video (XDCAM 1920x1080i @ 29.97) was not a problem. Unfortunately, it seems that the majority of the target "customers" that I will be sending this video to have still not jumped into the 21st century and purchased BluRay players, so when it came to the time to render the DVD copy of the project, I was a bit stumped as to why the video quality had taken such a horrific downturn since the master format was far superior.

Aside from the issues of choosing the proper field order for the SD render (original footage is upper field, DVD default is lower field), I found that motion artifacts and poor video quality came along with it. Trying both upper and lower field orders produced the obvious disgusting botched interlacing issues, so I just rendered the project to progressive MPEG-2.

After researching on the internet for possible suggestions or solutions, it seems that a successful tried-and-true process has yet to be found.

After looking through several options that didn't involve installing several different 3rd party programs, the only process that seemed to work well for me involved rendering each frame into a seperate TIFF file in the native format (essentially removing the interlacing, it seems), then re-rendering it to MPEG-2 Widescreen/Architect. The results were amazing compared to the stock Vegas downsampling. The problem is, I don't have a trillion GB to render my entire project to TIFF, then to SD-DVD.

So, my question is, is there any sort of plugin, or 3rd party program that can take the master render of my HD file, and put it through a "temporary TIFF intermediate" process, where it will render the full HD interlaced frame to a temp file, then automatically render it to an MPEG-2 file in the same process? Forgive me if that request/suggestion sounds preposterous as this is my first foray into this type of situation. It just seems that it would be the easiest way to go for someone in my situation who has limited HD space and wants optimal results.

Any suggestions/pointers would be appreciated!

Thanks,
Travis

Comments

Kevin R wrote on 11/29/2012, 4:39 AM
I know this doesn't help you.... but having faced this previously I now shoot everything at 30p.
john_dennis wrote on 11/29/2012, 8:15 AM
Start reading here.

The subject has been covered many times here. Use search on this forum.
vtxrocketeer wrote on 11/29/2012, 8:20 AM
Travis, can you shoot progressive? 30p? 24p? Strongly recommend it for ease of down-rezzing. Here's my workflow:

1. Shoot at 24p (23.976p, ok?). Edit at 1920x1080 24p. Render from Vegas to 1920x1080 24p avi master. (I use Cineform to keep the avi visually lossless but manageable in size. Lagarith works perfect, too.)

2. Render master avi from (1) in VirtualDub using resize filter, e.g., to 720 x 405, using Lanczos3 algorithm (important!). This gives a 720 x 405 24p avi (also Cineform, in my case).

3. Render 720 x 405 24p avi master in Vegas to DVD-compliant MPEG-2. (I assume you don't need details here.)

I've used this workflow for many years and the DVD results are stunning. I can't remember when I last shot interlaced, so for that piece of the puzzle, you'll have to take advice from someone else or my advice to shoot progressive.

HTH.
musicvid10 wrote on 11/29/2012, 9:53 AM
The field order makes no difference when rendering a DVD. Vegas gets it right whichever you choose.

Downsampling from 1080i to SD throws away 85% of your video data. There is no magic, just math.

That being said, you must get your bitrates right.
9,500,000 - 6,500,000 - 2,000,000 2-Pass VBR is about right for ninety minutes.
Use a bitrate calculator for longer times.

24p is not a DVD spec, so it will get telecined, which seems counterintuitive.
john_dennis wrote on 11/29/2012, 2:14 PM
If you decide to proceed with rendering to a still image sequence you could use PNG. It will save ~20% on disk space from TIFF. I rendered to png and tiff and put each back on the Vegas timeline, changed the top track composite mode to difference and could see no difference.

Because I'm just looking for someting to do, I rendered 15 seconds of 1440x1080-60i AVCHD to DVD using the default widescreen template except for changing the quality slider to 31. This was done using Vegas Pro 9.0e

Download Vegas Only Render from Here

Using the same settings, I rendered from the PNG still image sequence that I had previously exported from Vegas Pro.

Download Render from PNG Image Sequence from here.

This was all about interlace artifacts. I didn't do anything to colors. You decide which is better. I've got a lot more disk space than I have brains...
bill-kranz wrote on 12/1/2012, 5:21 AM
John:

Hi. My QT install of V 7.7.3 would not play your clips under latest version of Firefox.
Any idea what might be wrong?

Also, I had never heard of rendering a video clip as a still image. What all are the pro's and con's
of that method?

Thanks,
Bill
john_dennis wrote on 12/1/2012, 10:02 AM
The files are MPEG-2 for DVD Architect. You may have to download them and play them locally.

I'm not a big advocate of exporting still image sequences, I'm just curious and trying to keep from getting bored. I would describe it as the "nuclear option" for tasks like frame rate conversion, working in other applications like batch processing in Photoshop or breaking the deinterlace and resize tasks into separate steps.

Flexibility is a pro.

File size is a definite con.

If you do time lapse with still cameras, importing still image sequence is required.

Mindmatter wrote on 10/4/2013, 6:07 AM
sory to warm up the old thread - vtxrocketeer, could you give us a few specs as to the exact settings for the cineform and lagarith renders?
Thanks a lot!

AMD Ryzen 9 5900X, 12x 3.7 GHz
32 GB DDR4-3200 MHz (2x16GB), Dual-Channel
NVIDIA GeForce RTX 3070, 8GB GDDR6, HDMI, DP, studio drivers
ASUS PRIME B550M-K, AMD B550, AM4, mATX
7.1 (8-chanel) Surround-Sound, Digital Audio, onboard
Samsung 970 EVO Plus 250GB, NVMe M.2 PCIe x4 SSD
be quiet! System Power 9 700W CM, 80+ Bronze, modular
2x WD red 6TB
2x Samsung 2TB SSD

vtxrocketeer wrote on 10/4/2013, 8:44 AM
could you give us a few specs as to the exact settings for the cineform and lagarith renders?

Do you mean codec settings? I use a legacy Cineform product, NeoScene, that gives me access to only basic quality settings from Low to FilmScan2. Unless I'm doing a lot of generational renders or compositing, I'll just leave the quality setting at "High," which is more than plenty and an excellent balance between quality and file size.

I haven't used Lagarith in a long time, but I do recall a few settings that allowed an alpha channel or not, among others. Frankly, Cineform has worked so well for me that I've just stuck with it and not bothered with Lagarith.

If your question was getting at other settings, let me know. Since HD-->SD workflows are discussed a lot here, some folks can't resist a chance to jump in and tell you to do a search, so feel free to PM me.
johnmeyer wrote on 10/4/2013, 9:32 AM
Something buried in those huge, old, long posts is the upper and lower field do matter and that Vegas (at least the older versions) degrades the image slightly if you change field order from HD's default upper field first to DVD's default lower field first. It turns out that DVDs don't care if they are upper or lower field first, and neither do any of the players. Therefore, always keep the field order of your DVD encode the same as the field order of your interlaced footage, which for HD is upper field first. I think you will find that in the newer versions of Vegas, the MPEG-2 templates have been changed to upper field first, probably because that was an easier fix than actually going into the Vegas plumbing and wrangling all the issues that various people in these forums have been investigating for years. As a result, while the absolute best workflow still involves using Handbrake and MeGUI, but if you at least keep the field order consistent, you'll get a jump in quality over what you'd get if you didn't pay attention to field order.

And, just for the record, do not try to eliminate field order by rendering progressive, and do not set deinterlace to "none" in the project properties. Many people who have gone down the path you are now walking have made those mistakes.


musicvid10 wrote on 10/4/2013, 10:40 AM
I recall reading those posts with interest, and the rationale seemed logical; however, I looked closely for, and never saw any graphical proof offered demonstrating a difference from changing field order in Vegas. Since I was unable to see such a difference in my own tests (starting with highly-detailed HD source), I tend not to accept anecdotal evidence, since as we know, most subjective visual analysis is overshadowed by the expectations of the observer (aka pixel-peep syndrome).
Even some PSNR / SSIM numbers would have been helpful there . . .
johnmeyer wrote on 10/4/2013, 12:54 PM
I looked closely for, and never saw any graphical proof offered demonstrating a difference from changing field order in Vegas. Since I was unable to see such a difference in my own tests (starting with highly-detailed HD source), I tend not to accept anecdotal evidence, since as we know, most subjective visual analysis is overshadowed by the expectations of the observer (aka pixel-peep syndrome). It is true that I never posted the results of my upper/lower field HD to SD MPEG-2 render tests, but I did document my tests about 2/3 of the way down in this thread two years ago:

MIx of Interlaced clips (lower and upper) question

I guess since I didn't post the results, you could characterize them as anecdotal, but I did actually do tests and compare the results, so I think it was a little more than just trying to recall what happened and then ascribing it to something I suspected of being the cause of the problem.

So, as I said back then, I saw a significant difference between the field order when rendering from HD to MPEG-2 which was then used to produce a DVD.

VidMus wrote on 10/4/2013, 12:58 PM
@vtxrocketeer

I shoot at 60p because 24p looks just plain awful when panning and/or when people move.

How does one get good quality with 24p?

I am using 2 Sony HDR-PJ710V Cameras.

Great cameras!

At this point, I gave up on 24p...

As for this thread, shooting progressive makes a huge difference when converting to DVD!

No more interlaced for me!!!

Be sure and set the resample for all of the events to off or none or whatever the word is that my old brain will not remember. I am not on my Vegas computer so I cannot take a look and see what it is.

"What a drag it is getting old!"
vtxrocketeer wrote on 10/4/2013, 2:09 PM



Hmm, you're not going to deliver at 60p on Blu-ray or DVD, so at some point you will embrace a frame rate of 30 or 24, right? What then?

Anyway, I shoot a lot of stage productions, so 24p is beautiful. In the past when I've shot action at 24p, I'll boost my shutter speed so the action stays crisp (the "Saving Private Ryan" look, if you will).

I'm careful to not pan fast, at least at the longer end of a lens. By the same token, following a fast moving subject kept in frame can yield a slightly blurred background, which looks pleasing if pulled off right.

I now shoot on a Sony NEX-FS700 and I'm in love with 120p and 240p. I get super slow motion for jaw-dropping action shots. Sped up 10x in Vegas, my 240p footage yields real time on a 24p timeline.

Finally, I should have emphasized above that "my" HD-->SD workflow assumes (and probably works the best with only) progressive footage, which I've shot exclusively now for over 5 years. Much more learned folks than I have worked miracles with interlaced footage and the Handbrake/MeGUI workflow.
musicvid10 wrote on 10/4/2013, 3:09 PM
Sorry John, I had no idea those observations originally came from you. I'm pretty certain what I read came in a post from another user, with no supporting background. Now that I know, I'm far more likely to trust your observations, "anecdotal" or otherwise, because I know you, unlike many, go to some lengths to be objective. It's enough to make me curious about retesting with my 1080i source again.
Best as always.
;?)
Mindmatter wrote on 10/4/2013, 3:27 PM
Thanks vtxrocketeer.
I saw that Cineform seems to be included in the V12 render options, at least I don't recall buying it.
My footage is all in HD 25p, so should I really all the same choose interlaced as johnmeyer highly recommends?

I see several Cineform options: low, medium, high, filmscan1, filmscan2, keying

and: YUV422, RGB444, RGBA4444, default isYUV422

checkbox on, saying interleave every ... seconds.

So your recommendation is "high"?

AMD Ryzen 9 5900X, 12x 3.7 GHz
32 GB DDR4-3200 MHz (2x16GB), Dual-Channel
NVIDIA GeForce RTX 3070, 8GB GDDR6, HDMI, DP, studio drivers
ASUS PRIME B550M-K, AMD B550, AM4, mATX
7.1 (8-chanel) Surround-Sound, Digital Audio, onboard
Samsung 970 EVO Plus 250GB, NVMe M.2 PCIe x4 SSD
be quiet! System Power 9 700W CM, 80+ Bronze, modular
2x WD red 6TB
2x Samsung 2TB SSD

Kimberly wrote on 10/4/2013, 4:54 PM
@vidmus:

[I]I shoot at 60p because 24p looks just plain awful when panning and/or when people move.[/I]

I have the Sony cx760v. Do you have a suggested 60p workflow for:

NTSC Blu-ray
NTSC DVD
PAL DVD

I've been doing the 24p vxtrocketeer workflow for the last year with good results, but there is the motion issue with 24p.

Regards,

Kimberly
vtxrocketeer wrote on 10/4/2013, 10:01 PM
@Mindmatter

In this thread are interwoven workflows for interlaced HD-->SD, which goes to the OP's question, and for progressive HD-->SD. Since you acquire in 25p, then you need not worry about (or choose) any interlace settings. In the Vegas project properties, I leave the interlace setting to "none."

I don't know how you're seeing all the Cineform settings above "High" without having purchased a Cineform product. Maybe you installed the GoPro freebie Cineform product at some point . . .? Anyway, yes, I recommend the default YUV422 at "High" and you'll be excellent. Leave the interleave setting alone.

One time for grins I chose RGB444, which was ridiculous since at the time I acquired only HDV at 4:2:0, and a Cineform pop-up box let me know that I hadn't acquired the license for RGB444. I'm not sure why the 444 options weren't greyed out or left out altogether, but it didn't matter.

I like to have Cineform HD and SD masters. These take up HDD space, but I like to have flexibility to use my masters however I see fit. In another thread, I suggested (after trying this) to frameserve directly from the Vegas HD timeline to VirtualDub, with the result of having no HD master (since the output are only frame-served frames), but having an SD master from VirtualDub. I guess this would make sense for someone if they wanted ONLY SD delivery.

In my case, however, I typically render each Cineform master, HD and SD, to DVD Architect compliant MPEG-2's and, at this stage, I'll sometimes add a little sharpening with Sony Deconvolution Kernel. I highly recommend this for DVD delivery as it makes a significant improvement to the softened image after the down-rez step.

Finally, the whole business of using Cineform here is because VirtualDub reads only avi files and it, not Vegas, has the superior Lanczos3 resizing algorithm. Since the workflow involves multigenerational renders, I chose to make Cineform intermediates. Otherwise, i.e., if Vegas did the better resizing, I'd simply edit my acquired footage directly and render/resize all in one render.
Mindmatter wrote on 10/5/2013, 3:17 AM
Thanks again rocketeer!
I don't know either how cineform landed in the Vegas settings, unless I bought it years ago and cannot rermember - I thought it was part of V12 by default...
I'll give my full HD Jazzshow a go following your workflow and will post results some time next week.

AMD Ryzen 9 5900X, 12x 3.7 GHz
32 GB DDR4-3200 MHz (2x16GB), Dual-Channel
NVIDIA GeForce RTX 3070, 8GB GDDR6, HDMI, DP, studio drivers
ASUS PRIME B550M-K, AMD B550, AM4, mATX
7.1 (8-chanel) Surround-Sound, Digital Audio, onboard
Samsung 970 EVO Plus 250GB, NVMe M.2 PCIe x4 SSD
be quiet! System Power 9 700W CM, 80+ Bronze, modular
2x WD red 6TB
2x Samsung 2TB SSD

VidMus wrote on 10/6/2013, 12:38 AM
@Kimberly

I am doing a lot of video work right now so I will post again soon.
Mindmatter wrote on 10/7/2013, 3:31 PM
Ok Cineform riddle solved ...I got a "license expired" bar throughout the render. Must have used a trial at one point.
One more question: I finished rendering my .avi to the 720x405 master in VirtualDub. Now, do I need to respecify that exact aspect for DVDA in Vegas? The standard widescreen template seems to be 720x576?
Also, the markers get lost along the way...unless I forgot to specify something in VirtualDub?
Thanks!

AMD Ryzen 9 5900X, 12x 3.7 GHz
32 GB DDR4-3200 MHz (2x16GB), Dual-Channel
NVIDIA GeForce RTX 3070, 8GB GDDR6, HDMI, DP, studio drivers
ASUS PRIME B550M-K, AMD B550, AM4, mATX
7.1 (8-chanel) Surround-Sound, Digital Audio, onboard
Samsung 970 EVO Plus 250GB, NVMe M.2 PCIe x4 SSD
be quiet! System Power 9 700W CM, 80+ Bronze, modular
2x WD red 6TB
2x Samsung 2TB SSD

vtxrocketeer wrote on 10/7/2013, 4:48 PM
@Mindmatter, you know, I'm thinking of writing all of this out on a page on my website. 'Till then, and when a ton of free time comes my way, here is further advice:

Now, I'm assuming that you rendered your 720x405 avi to a DVD Architect-compliant MPEG-2 within Vegas. Right? RIght?! Right.

Open a new DVDA project and, within properties, select a 16:9 ratio, and choose 720 x 480 (NTSC) or 576 (PAL). DVDA should "know." You will have a 16:9 project and it should look correct without letter box bars.

Turning to the subject of markers, you've got to realize that they will appear (or are available) in DVDA with MPEG-2 media having embedded markers. The avi files and Virtual Dub are of no help in this regard. In fact, you've noticed that your markers are gone after the VD step. You can actually "see" markers on avi files within Vegas, but for reasons I don't know at the moment, these do not magically reappear if you render to a DVDA- compliant MPEG-2. Hence, you've got to get the markers "back."

You need only set markers ONCE for Blu-ray and DVD. How? Read on.

I make a Blu-ray project first. In Vegas just before rendering my HD avi master to a DVDA-compliant MPEG-2, I'll set chapter markers. Often I'll have 10-20. In the Render dialog, I'll tick the check box for saving markers. Dropping the rendered MPEG-2 into DVDA will, of course, yield back the markers.

Just above the DVDA timeline is an icon called "Save Markers." Click to save. Then save the PROJECT by another name for the next step (File | save as).

Next I'll prepare all my DVD media. Here, as you noted, you have no markers. No problem. Within the DVDA project that you "save[d] as" above, change the project type to DVD and change appropriate properties. Then replace all of your relevant media with the DVD media, e.g., replace the video (Blu-ray MPEG-2) with another video (DVD MPEG-2). Same with audio and menus if those matter to you.

Now, here is the major time-saver: just above the DVDA timeline, click the icon called "Load Markers." Boom -- they reappear. But you're not quite done.

For DVD, the chapter markers need to fall on I-frames, and probably some of your markers from the Blu-ray project won't line up exactly with I-frames. These markers have a tiny exclamation point inside of them. Zoom in on these markers one-by-one, click and hold a marker, then you'll see small, black, upside down triangles appear near the marker. Usually three appear; these are I-frames. Just move the marker slightly to snap to one of these I-frames and release the mouse button. Do this for all markers, and you're done.

This is MUCH, MUCH easier and quicker to do than me typing it. It saves a lot of time by having you type markers just once for both projects.

One potential limitation of this workflow is that by slightly adjusting the chapter markers in DVDA to align with I-frames, you are of course changing the chapter start points by several frames. This could be important if you really need frame-accurate chapter start points. In that case, just suck it up by setting your markers for DVD in Vegas and having Vegas create I-frames at each marker. (In my typical projects, chapter start points occur after fades to black, so a few frames don't matter at all to me.)

Hope this makes sense.
Mindmatter wrote on 10/7/2013, 5:09 PM
vtxrocketeer, I really don't know how to thank you for all your trouble and time...!
I'll certainly keep geting back to all your advice.

meanwhile, my first test attempt at your workflow did not yield good results...the (65Gb) file coming out of Virtualdub was extremely pixellated and weird looking. I assume i made mistakes along the way - as a quick test i simply rendered a DVDA mpeg2 straight out of the original project and compared the 2. The Virtualdub avi looks terrible and the Vegas render is actually good.
I'll have another go and see what I misssed along the way. Tbh, I don't find anything particularly bad in the V12 render, I'll have a look an my flatscreen tomorrow.

Night, and thanks again!

AMD Ryzen 9 5900X, 12x 3.7 GHz
32 GB DDR4-3200 MHz (2x16GB), Dual-Channel
NVIDIA GeForce RTX 3070, 8GB GDDR6, HDMI, DP, studio drivers
ASUS PRIME B550M-K, AMD B550, AM4, mATX
7.1 (8-chanel) Surround-Sound, Digital Audio, onboard
Samsung 970 EVO Plus 250GB, NVMe M.2 PCIe x4 SSD
be quiet! System Power 9 700W CM, 80+ Bronze, modular
2x WD red 6TB
2x Samsung 2TB SSD

Arthur.S wrote on 10/8/2013, 6:38 AM
I followed this advice from Rosebud a while back:

DVDA
Go in to internal preferences (Ctrl+ Shift + Options>preferences)
Set "Media markers in video file" to FALSE.
And you should be able to load markers from audio files.

I always make a separate DD audio file anyway, so dropping that into DVDA then hitting "Load markers" is a very quick and easy way to accomplish what you want. You can run your video through anything you want - the markers are taken from the audio file.