HDV to ProRes

[svar]

Hey
I used the search, but could not find any answers to this:

I was wondering if I sould start to make ProRes files out of my HDV footage.
Yes, I know I dont get any better quality, but as far as I know HDV is a lossy format and ProRes is not, and then you work with Motion and effects in general its better to work with a lossless format to keep the quality.
(as far as I know, HDV loose quality for eatch render step, ProRes dosent.)

But is it best to import footage from cam into ProRes direct, or use compressor (or any other tool) on the HDV files after import?

:)

[IndyFX]

Yes, yes, yes and yes.

I just capture clips in native HDV (HDV in mov) and then use compressor to batch convert the lot. (this is typical workflow for professional treatment of HDV content)

Typically I also resize the movies (in compressor) to 1440 X 810 with square pixels Besides being the true (actual) horizontal resolution of HDV, this gives me clipping room for framing, stabilization and/or effects for eventual conversion to 720p (1280 X 720). HDV makes really nice 720 or kind of soft 1080, take your pick.

Question; why did you post this in the PC (non Mac) forum?

[CycleWriter]

Question; why did you post this in the PC (non Mac) forum?

My bad. It was originally in the Questions, Problems...subforum, but as it is a software issue I moved it not knowing it was a Mac program. I am so ashamed.:hv20-smilie15::hv20-smilie84:

[Rumpelgeist]

Typically I also resize the movies (in compressor) to 1440 X 810 with square pixels Besides being the true (actual) horizontal resolution of HDV, this gives me clipping room for framing, stabilization and/or effects for eventual conversion to 720p (1280 X 720). HDV makes really nice 720 or kind of soft 1080, take your pick.
If you care so much about quality loss (or, preservation of quality), why do you resize to 1440x810? What is so magical about 810 lines? is it because 1440 * 9 / 16 = 810? So what? Why do you care for square pixels, what is wrong with rectangular ones? If you eventually will be converting to 720p, why not resizing to 720p right away, skipping one conversion?

Or, if you shoot interlaced, why not resizing to 1440x540 with 0.6667 PAR, keeping ALL information intact? Sure, another resizing to 720p will add some extra pixels vertically, but with interlaced you don't have more than 540 vertical pixels anyway, so resizing to 810 lines does not make sense to me. Please, elaborate.

[svar]

Hi and thanks :)

I tried to find some more info about the subject and found out that ProRes IS a lossy format in teory, but it seems not to in practis.

And it seems like people use 1 of 2 solutions:

1: Capture from cam to ProRes over FW (taking huge space on HD) and work in a ProRes timeline, export masters as ProRes

2: Capture from cam to HDV over FW (making small files) and work with ProRes as render codex (not shure if they work in a ProRes or HDV timeline), and export masters as ProRes

As far as I understand, ProRes is for HDV,XDCAM HD/EX only, for they who use cams with SDI, they ude codexes thats comes with the SDI cards, like the ones from Blackmagic.

Some people out there seems to like nr 2 best, cause you get small files for storage and the avanges of ProRes in the render prosess (and it is in the render prosses of HDV source you realy need the ProRes power as far as I know).
And some other people dont care about space and use nr 1 no matter what.
I guess I have to try for myself, but as far as I see, using HDV as render codex is "stupid" then you have FCS2.

Source:
http://library.creativecow.net/articles/poisson_chris/hdv-prores.php
http://discussions.apple.com/thread.jspa?threadID=1656290&tstart=0

[IndyFX]

Aramis,
I shoot almost everything in 24p, (as does most of the professional world (or 25p), outside news gathering, and yes I am aware of your one man crusade to convert the world to interlaced video, if you believe it, good for you)
So... yes, 810v pixels is not "magical" it simply maintains aspect while using the actual 1440 horizontal pixels present in HDV in a square PAR. This allow keeping all the original horizontal resolution while 1440X810 can be reframed, offset or stabilized without up-rezing (in a 720 target format).

It is always better to hold resolution as long as possible (so simply scaling initially to 720 is not a good idea particularly when you will likely need that buffer in some shots), However given the choice of maintaining the original 1080 vertical pixels by using non-square PAR or just getting a clean (square) 1440X810 I choose the latter.

[IndyFX]

Mixed questions (and no "one size fits all" answers) My replies in Bold:


Hi and thanks :)

I tried to find some more info about the subject and found out that ProRes IS a lossy format in teory, but it seems not to in practis.

ProRes HQ is "visually lossless" the standard ProRes is smaller (more compact) by nearly half but is slightly lossy. However, be aware that either is an order of magnitude away fron the losses incurred in a codec like HDV

And it seems like people use 1 of 2 solutions:

1: Capture from cam to ProRes over FW (taking huge space on HD) and work in a ProRes timeline, export masters as ProRes

2: Capture from cam to HDV over FW (making small files) and work with ProRes as render codex (not shure if they work in a ProRes or HDV timeline), and export masters as ProRes

As far as I understand, ProRes is for HDV,XDCAM HD/EX only, for they who use cams with SDI, they ude codexes thats comes with the SDI cards, like the ones from Blackmagic.

Generally, only interframe codecs need to be transcoded. If you have a starting format that is lightly compressed and intra-frame (no temporal compression) many choose to just keep everything in that format. However transcoding EVERYTHING to ProRes HQ is SOP for many shops now.

Some people out there seems to like nr 2 best, cause you get small files for storage and the avanges of ProRes in the render prosess (and it is in the render prosses of HDV source you realy need the ProRes power as far as I know).
And some other people dont care about space and use nr 1 no matter what.
I guess I have to try for myself, but as far as I see, using HDV as render codex is "stupid" then you have FCS2.

Yes HDV as a edit format will cause losses when ever intermediate rendering is required. However, I would not advise using a format for clips (HDV) that is not the edit format (prores), this will require you to render everything for preview. (whereas prores clips will play un-rendered in a prores timeline)

Source:
http://library.creativecow.net/articles/poisson_chris/hdv-prores.php
http://discussions.apple.com/thread.jspa?threadID=1656290&tstart=0

[svar]

IndyFX: So what you say is that keep all in ProRes is not so bad if you have the HD space, Im right?

[IndyFX]

IndyFX: So what you say is that keep all in ProRes is not so bad if you have the HD space, Im right?

Yup that is basically it, if you have the space, keeping everything in ProRes will give you little or no degradation and an easy workflow (no render editing & previewing) which is why it (encode everything in ProRes) has become the de-facto standard for a lot of shops.

[Rumpelgeist]

Aramis,
I shoot almost everything in 24p, (as does most of the professional world (or 25p), outside news gathering, and yes I am aware of your one man crusade to convert the world to interlaced video, if you believe it, good for you)
I don't know how you can be aware of something that does not exist. I hate interlaced video. You must have confused me with someone else.


So... yes, 810v pixels is not "magical" it simply maintains aspect while using the actual 1440 horizontal pixels present in HDV in a square PAR. This allow keeping all the original horizontal resolution while 1440X810 can be reframed, offset or stabilized without up-rezing (in a 720 target format).

It is always better to hold resolution as long as possible (so simply scaling initially to 720 is not a good idea particularly when you will likely need that buffer in some shots), However given the choice of maintaining the original 1080 vertical pixels by using non-square PAR or just getting a clean (square) 1440X810 I choose the latter.
I got it for the first time that you wanted to achieve 1.0 PAR, and I asked you why, and you haven't given a clear answer. What is so... dare I use this word once again... magical about square pixels? Is it just your own perception of "clean" and "good", or the software you are using cannot work properly with non-square pixels?

By resizing 1080 lines to 810 you lose information. You seem to be ok with it. Is it just because the HV resolves only about 600-650 lines, and the A1 resolves about 800 lines, so in your practice, the remapping of 1080 lines into 810 lines works fine?

Have you tried resizing to 1920x1080 instead? This preserves vertical resolution, as well as horizontal, though uprezzing usually introduces artifacts, but you will be down-resizing to 720p anyway. Is this better or worse or about the same compared to your approach?

In any case, if you worked with the original 1440x1080 PAR 1.33 frame, you would not lose any info and not spend time and computer power on resizing.

[Dr. Benway]

What a mess.

[Dana Love]

Actually, it's a pretty cool conversation...though it has the tone of a diagnostic session on House. :)

Aramis, are you saying converting to ProResHQ with a 1440x1080 PAR 1.33 frame would introduce the least artifact, then resize down to 720p to be left with the crispest image?

Indy, you crop during preprocessing, right? I need to dig into my presets and make sure I'm resizing as beneficially as I can. Your post helped me figure out where I was screwed up.

This helps me understand why softness gets introduced late in my workflow. Cool!

And thanks!

[Distortedrumble]

This is definitely one of those threads where I need to go back and reread it when I don't keep turning back to look at the tv.

[Rumpelgeist]

Aramis, are you saying converting to ProResHQ with a 1440x1080 PAR 1.33 frame would introduce the least artifact, then resize down to 720p to be left with the crispest image?
I am not saying, I am asking. I just don't get what is wrong with non-square pixels.

[svar]

btw, here is where I found the discussion about pure proRes or yust in the rending:
http://discussions.apple.com/thread.jspa?threadID=1916139&tstart=0

[Owen]

I have been evaluating FCS recently and ran into this apparent need to render during editing. I do not want to render during the edit process or multiple render under any circumstances. Adobe Premiere only needs to render once, for final output, there is never any worry about multiple rendering steps during editing and no need to use an intermediate renderer like ProRes. Can FCS work this way, and if not why not?

[Owen]

but with interlaced you don't have more than 540 vertical pixels anyway,....

That’s just not true, 1080i is 1080 lines and deinterlaces to 1080p, the resolution you get in motion is dependant on the deinterlacing quality.
The HV cameras don’t have 1080 vertical resolution but it’s more then 540. Converting 1080i to 540p (simple bob deinterlacing) definitely loosed vertical resolution.

[Owen]

though uprezzing usually introduces artifacts, but you will be down-resizing to 720p anyway.

Upscaling will not introduce “artefacts” if it is done properly. In fact upscaling tends to smooth out “artefacts”, not introduce them.
I don’t know how capable FC is at upscaling, but there are solutions that do a very good job.

[Ian-T]

From what I always understood...downrezzing 1080i to 720 smooths out compression artifacts.

[HueyNRolf]

I have been evaluating FCS recently and ran into this apparent need to render during editing. I do not want to render during the edit process or multiple render under any circumstances. Adobe Premiere only needs to render once, for final output, there is never any worry about multiple rendering steps during editing and no need to use an intermediate renderer like ProRes. Can FCS work this way, and if not why not?

Did you set the 'RT' pulldown to 'unlimited'?

[Owen]

Just tried with RT set to “unlimited” and I can now play 1080i Mpeg2 direct from the time line but with many dropped frames, its way to slow to be usable and this is on a quad core Q6600 CPU running at 3Gig. My single core 2Gig laptop running Windows XP and Premiere has no problems with playback of the same video from the time line

[HueyNRolf]

a quad core Q6600 CPU running at 3Gig.

What machine is that?

[Rumpelgeist]

That’s just not true, 1080i is 1080 lines and deinterlaces to 1080p, the resolution you get in motion is dependant on the deinterlacing quality.
I am not going to argue, this is pointless. Sure, motion-adaptive deinterlacing can get the most of interlaced video, using weave for static scenes, bob for random movement, and motion-adaptive region-based deinterlacing for something in the middle. But consumer software does not provide these features, it is either weave, blend or bob, of which only the last one always works.

[Owen]

What machine is that?

Hackintosh, with nVidia 8800GT, 4Gig RAM and OSX version 1.5.6

[HueyNRolf]

Hackintosh, with nVidia 8800GT, 4Gig RAM and OSX version 1.5.6

And your FCP version is?

[Owen]

I am not going to argue, this is pointless. Sure, motion-adaptive deinterlacing can get the most of interlaced video, using weave for static scenes, bob for random movement, and motion-adaptive region-based deinterlacing for something in the middle. But consumer software does not provide these features, it is either weave, blend or bob, of which only the last one always works.

That’s why I leave 1080i in its native form and let the display device deal with deinterlacing, that works fine and provides the same resolution and significantly smoother motion then 1080p 24/25.
Unless you need to transfer to film 1080i works great, especially if you need to capture action where 24/25 frames per second does not cope well.

[Owen]

And your FCP version is?

6.0

[HueyNRolf]

You're running a Hackintosh with an old OS and wonder why the latest FCP isn't performing well?
Oy vey!
It's a miracle it runs at all

[Rumpelgeist]

That’s why I leave 1080i in its native form and let the display device deal with deinterlacing, that works fine and provides the same resolution and significantly smoother motion then 1080p 24/25.
Unless you need to transfer to film 1080i works great, especially if you need to capture action where 24/25 frames per second does not cope well.
Indeed, it works if kept as is, at 1080i. But if I need to make a regular DVD out if, it looks like crap. Progressive scales much better, and faster, no need for complex deinterlacing algorithms.

[Owen]

Excurse my error, I should have said 10.5.6. As far as I know 10.5.6 was released in December 2008, I am not aware of a newer version.

I have also used 10.5.5 and 10.5.4.

[HueyNRolf]

Excurse my error, I should have said 10.5.6.

Indeed you should.

[Owen]

Indeed you should.

So now that we have the “facts” sorted out, is what I am seeing normal? Should I be able to play 1080i Mpeg2 video direct from the time line smoothly without rendering first?

I can play clips before they are dropped on the time line just fine and playback in QuickTime and VCL is no problem either, only playback from the timeline is stuttery.

I should add the HD video was not captured on the Mac, and once rendered in FC it does play smoothly from the time line. Could be some video format compatibility issue, but I have never had this kind of problem on Windows with Premiere, all video that can be played on the machine can be edited.

[svar]

According to what info I find, this sould be the right settings for import 25p HDV footage to ProRes:
http://bildr.no/thumb/355802.jpeg (http://bildr.no/view/355802)
(click on picture to make it larger)

Because of the 25p in a 50i stream, I guess its wrong to choose FireWire 25fps in the "device control preset", but instead use a ProRes 25p sequenser and remember to set the "field dominance" to "none" in the clips.

Im on the right track here, or far off?

[svar]

Just captured a tape into ProRes and here are my experience so far:

-almost 60min of tape took 40,83GB

-since the 25p recordings are captured into a 1080 stream, the clip properties say its 1920*1080

-Because of this, the clip and sequence has to be et to "none" at "field dominance, and you have to work in a 1080 timeline (FCS2 Im using)

-because 25P is in a 1080 stream, its hard to tell by the clips themselves if its recorded in 50i or 25p, better keep one setting for the tape, if mixing in same tape, better writhe down the settings.

-Because of the field dominance, then its set right, working with you are working with true 25p, even if the sequence are a 1080 (i guess the Field dominance has to be "upper" for 1080i shots), and then the export settings are set right, true 25p is the result.


So far so good :)

[Rumpelgeist]

-because 25P is in a 1080 stream, its hard to tell by the clips themselves if its recorded in 50i or 25p, better keep one setting for the tape, if mixing in same tape, better writhe down the settings.
Set preview to 100% (hope your NLE allows this), if you see comb artifact when there is motion, then this is 50i, otherwise 25PsF.

[svar]

Aramis: Thanks for the trix :)
But else Im on track here?