Which is better- HDV transcode or re-capture?

Sorry, I mean that second clip only has two shots back to back. Supposedly.

strypes Wrote:
-------------------------------------------------------
> Haha.. It's 11pm here in sunny singapore.. err...
> not so sunny now... where the hell are you, harry?

I'm in the sunny climes of the 323 area code in the glorious Hollywood Hills, California, home to brilliant intellects like Michael Horton and myself.

Harry.

Wow, Jude!

Thanks for the examples of exciting and innovative film editing. I will forward this stuff to David Fincher.

Can't imagine what you're complaining about. Apparently you imported some fairly standard equestrian footage into FCP and it has automatically generated this groundbreaking avant-garde and eye-catching approach.

Always ahead of the curve in your part of the world, huh? Pushing the envelope. Edgy.

Far out.

Best

Harry.

Harry Bromley-Davenport.

another scenario that had it's tyres kicked was whether there was any real difference between Jeff's method (HDV in ProRes timeline then Export QT Movie w/ Current Settings) or editing in a native HDV timeline (with renders set to ProRes) and then rendering out as QT Movie w/ Settings set to Pro Res (thereby also avoiding the so called HDV conform process)

the advantage of the latter would be the supposed better response in the timeline as the basic editing would all be native but still retaining the advantage of the ProRes rendering

in Jeff's case, the argument was moot as he (correct me if I'm wrong Jeff) specifically works in a ProRes HQ timeline ... as far as I know, the ProRes render option in an HDV timeline is only basic ProRes and not the HQ variety

Well, that's it then. I'm going into real estate.

Gerard writes-
[But naturally, you'd want to be able to do a batch capture]

Any recapture issues, timecode stuff, with HDV? No seat time doing that.

Much thanks to all. Jeff, that was a nice rundown on HDV essentials. Harry, I like your confidence in the workflow. Jude, I worry about blue frames and other myosynclastic burps.

[And why on earth are we all awake so early? ]

My fault! Now we're all vampires.... nyhahah...

My aim is that an editing timeline on a 3Ghz octopus stays realtime as long as possible, that the image doesn't suddenly fall apart or go blue or green just to test if I'm awake; doesn't require lengthy interruption for rendering when lower third titles are added; when special filters or color correction are added; when photos are imported and images are zoomed or rotated; when superimposed shots are stacked over eachother for interesting montage effects; and when audio track grows beyond 8 levels.

In my tests, native HDV falls apart pretty quickly under this load.

I really like the idea of an unrendered ProRes 1080 realtime.

My ultimate aim is best quality at large projection from this stuff.

- Loren
Today's FCP keytip:

Play from In to Out with Shift - \ (Backslash) !

Final Cut Studio 2 KeyGuide? Power Pack.
Now available at KeyGuide Central.
www.neotrondesign.com

Hang on, hang on ... whoah there!. You don't get away with using words like myosynclastic on this forum. There are chidren in the room.

Besides which I can't find it in ANY dictionary. I think you made it up in a fit of mendicular topagarianism.

You are a fraud.

You may even be virtual.

Explain yourself.

Harry

Harry Bromley-Davenport.

Oh, I am definitely a fraud.

But I do experience a Myosynclastic Jerk once in a while-- it's that sudden brain flinch while you're falling asleep after a hectic day-- I think it's a sudden discharge of photochemical energy built up during demanding edit sessions. Sort of like an internal lightning bolt but not lethal.

I don't know if everybody experiences it. But I suspect so; I see an awful lot of it represented in modern horror movies.

Here's an HDV finishing question-- anyone here actually experience a recapture of HDV sources to a Digital Intermediate for film finish?

If so, how's that look projected film 50 feet wide?

- Loren
Today's FCP keytip:

Play from In to Out with Shift - \ (Backslash) !

Final Cut Studio 2 KeyGuide? Power Pack.
Now available at KeyGuide Central.
www.neotrondesign.com

Sure you don't mean myoclonic jerk? A twitch as a result of the brain thinking certain body parts are dying and need to be awoken?


www.derekmok.com

Loren, stupid question: Why would you recapture HDV to go to DI?

At the risk of repeating myself, once the footage has hit the tape, what you get out of it in the first real-time pass through Firewire is all that will ever be there. Everything after that is data.

I guess I basically do a DI on my jobs, or at least on the ones I care about. It's a ghetto DI, admittedly; I use Color to grade my shots and create a look. I'm finishing a short show right now that incorporates interview footage shot over a span of months. Most of the interviews were shot pretty contrasty, but one was really flat, so I used Color to relight it. The results weren't what you'd get from a professional colorist on a Davinci, but they were pretty good. The basic workflow for doing a true DI would be the same, though there might have to be some data conversion to get the footage into the color corrector. You can convert 8-bit 4:2:0 HDV footage to 12-bit log DPXs if you want; they just won't get any prettier along the way.

Oh, Jeez, based on a search of that term, Derek, I have mad cow disease.

No, I can't find it either, yet I know the term from a doctor. Myclonic is also accurate.

Thus far I know I'm going into real estate followed by spastic death.

In glorious HDV.

- Loren
Today's FCP keytip:

Play from In to Out with Shift - \ (Backslash) !

Final Cut Studio 2 KeyGuide? Power Pack.
Now available at KeyGuide Central.
www.neotrondesign.com

Jeff writes-
[once the footage has hit the tape, what you get out of it in the first real-time pass through Firewire is all that will ever be there. Everything after that is data. ]

Okay. The most popular workflow a couple years ago was HDV three-wire PrPbY component out to DVCProHD, and from there into Avid or FCP. I've never seen that projected big, just TV.

But isn't that better than Firewire? What's the secret sauce these days, assuming you're mastering a DI to film out?

- Loren
Today's FCP keytip:

Play from In to Out with Shift - \ (Backslash) !

Final Cut Studio 2 KeyGuide? Power Pack.
Now available at KeyGuide Central.
www.neotrondesign.com

Hmm. I wouldn't use that signal path myself, Loren. I can understand the rationale behind it ? most HDV cameras lack SDI out so YPrPb is the best anybody could get, and until recently DVCPRO HD was way easier to work with than HDV, and capturing DVCPRO HD in real time is more time-effective than capturing HDV in real time and then transcoding.

But ? geez, man. A D-to-A conversion, a downscale and another A-D conversion? Why shoot HDV at all? Just go buy an HVX200 or whatever.

The thing to remember about HDV is that (at least in 60Hz land, where I live) it's 1440x1080 and 4:2:0 at a nominal 25 Mbps. DVCPRO HD is 4:2:2 at a nominal 100 Mbps, which is great, but it's also only 1280x1080.

Due to the 4:2:0 chroma subsampling, HDV's chroma resolution is effectively 720x540. It's subsampled by half horizontally and vertically. (It's not really 720x540, because if I remember correctly the Pr and Pb components are interlaced together, but whatever, just go with me on this.)

When you stretch your HDV footage out to full raster, as happens when you play out of the component outputs on your camera or deck, you're resampling the 1440 horizontal luma pixels to a slightly blurry 1920, and you're resampling the 720 chroma pixels to a somewhat more blurry 960.

DVCPRO then resamples the luma down to 1280. It's not even resampling the original 1440; it's resampling the already slightly blurry resampled output from the camera or deck. It's the equivalent of blowing up a photograph in Photoshop, then scaling it down to a smaller size than you started with. You'll get poorer results that way than by just scaling it down to start with, because of mathematical averaging of pixels.

The way it treats chroma is even worse. Those original 720 samples get stretched out to 960, then sampled all the way back down again to a mere six hundred and forty. With all the interpolation and averaging that's been done, how many of those samples accurately represent the scene you filmed? Damn few. The problems are especially visible in areas of high chroma contrast, like a bright blue shirt against a skin tone.

And none of that takes into account the softness and averaging introduced by going from a digital signal on tape to an analog signal on the wire and back to a digital signal again in your I/O board.

Sure, DVCPRO HD was easier to cut than HDV back in the day. But the cost is a third of your luma resolution and about eighty percent of your chroma resolution.

Think back to the days when serial-digital first became a thing. It was pretty amazing, you know? You could take footage out of your NLE and into a D5, and capture it back in and end up with exactly the same pixels, down to the last bit. As long as you were laying off to a lossless recorder, like a D-5 or an uncompressed DDR, it was a totally lossless round trip.

That's how HDV-over-Firewire is. Yes, your HDV camera compresses the ever-lovin' @#$%& ? pardon my French ? out of your footage when it records it to tape. But once it's on the tape, transferring it to your computer over Firewire, and even back out to HDV and back in again (as long as you don't make any edits along the way) is a totally lossless round trip. If your tape is a sponge saturated with footagey goodness, capturing it over Firewire wrings the sponge totally dry, giving you every single drop that was in it. No D-A or A-D passes, no resampling, no averaging, no interpolation. It's as lossless as copying a Quicktime to your framestore from a Firewire drive.

The tragedy is that you can't make HDV look any better. Ever. Once it's on the tape, you're done. This is dramatically unlike film, where there's so much useful information crammed onto the negative that you can practically keep scanning until all your RAIDs are full and you'll never get it all. (Well, sort of. Reasonable people can argue over whether the term "useful" applies to finer and finer digital depictions of grain.)

Oh, one little slightly-off-topic side note: I've done a lot of work with the 24F format shot on the XL H1. You wanna know Canon's dirty little secret? It's this: 24F is just 23.976p with plain old 2:3 pulldown added.

The first time I tried shooting tethered, coming out of the HDSDI spigot into my Kona BOB and doing a "capture now" from a non-controllable device, I discovered that my monitor wouldn't show a picture if I had my Kona set to 23.976p. So I tried toggling it over to 29.97i, and it captured fine. The footage came in as uncompressed 4:2:2 HD, full raster, with 2:3 pulldown. Since removing pulldown from uncompressed footage is child's play, I was good to go.

As I understand it, Canon does do something funny with its HDV tapes when you record 24F to them. Word on the street is that a tape recorded in 24F will only play back on a Canon camera that supports playing back 24F tapes. Sony's HDV decks, for example, aren't supposed to be able to play those tapes at all, even as just straight 29.97i with pulldown. I've never actually tried it myself, though. Maybe I'm wrong about that bit.

Wow, Harry, that was a pretty rude and hurtful response to my post demonstrating problems I've had with HDV. Thanks for that.

In no way did I mean that post to be taken seriously.

I can't believe that you took it at face value ... are you having me on?

Please, Jude, let me quote my post:

"(FCP) has automatically generated this groundbreaking avant-garde and eye-catching approach. "

I find it difficult to believe that you read that statement as anything other than an attempt at levity.

I'm truly sorry, and personally appalled and ashamed, that my words were written in such a way as to hurt your feelings. I assure you that it was meant in jest.

But I have hurt you and I didn't mean to do so.

Please accept my apology, Jude. I greatly value the friends I make here on this board and I consider you one of them.

Can we please have a virtual drink together and put my foolishness aside?

Sincerely,

Harry.

Harry Bromley-Davenport.

Sorry Harry, I must be feeling overly sensitive today. Thought you were having a dig.

Jude,

Having had ten minutes of horrified guilt at having offended you, I can now see that it wasn't tactful of me to make fun of the serious problem with which you are currently dealing.

I think that, were I in your position, I too would be annoyed and feel that my issue was being slighted.

One of my true pleasures in life is that, late at night, I can browse this board where the best advice, the most intelligent and educated people and also some of the most amusing wits can be found.

Perhaps, I think, I was trying too hard to be as sardonic and amusing as the others who contribute to these pages, and who are usually much funnier than me. Forgive me. I'm English and therefore somewhat competitive about humor ... or humour, as I would spell it.

Now please - accept my apology.

I have fallen on my sword right up to the hilt and am bleeding to death here.

Best wishes,

sincerely,

Harry.

Harry Bromley-Davenport.

Harry - get off the sword, you're bleeding on the shag pile. I can see that I misconstrued your intention. It's one of the buggers about not being able to see people's faces in this place.

All is forgiven, and I hope you'll forgive me jumping to the wrong conclusions.

Good.

R.I.P.

Harry

Harry Bromley-Davenport.

>Any recapture issues, timecode stuff, with HDV?

If you capture Prores via firewire, you can't, apparently. The transcoding seems to affect tc accuracy.

> But once it's on the tape, transferring it to your computer over Firewire, and even back
>out to HDV and back in again (as long as you don't make any edits along the way) is a
>totally lossless round trip.

Capturing via firewire basically sends whatever digital information that is stored on tape into the machine rewraps that from dv stream into an .mov wrapper. SDI is digital information, but it only sends uncompressed video (which also means that the footage is decompressed). And this also means that the footage needs to be captured uncompressed or you'll be introducing recompression into the picture, same goes for Components but Components are worse as the footage goes through a DA-AD conversion.

DvcproHD is a nicer friend in the PAL world, btw... It's 1440x1080. But it's still effectively more compressed than ProRes (14.4MB/s vs 23MB/s), as it's also only 8 bits. And ProRes seems to make more effective use of compression.



www.strypesinpost.com

Sorry, I was helping Jude backchannel, devising assassination plots on Sir Harry... then he dashed it all by being a true gent!

Gerard-

Right, everything you say was backed by Larry Jordan's pithy (I know Ken S. likes those!) Pick Our Brains explanations in last night's Digital Production BuZZ radio show, with Chairman Mike delivering dramatic HDV questions. Outstanding.

[And this also means that the footage needs to be captured uncompressed or you'll be introducing recompression into the picture, same goes for Components but Components are worse as the footage goes through a DA-AD conversion.]

Noted. In case you didn't hear the show, Larry echoed what I always counsel-- get the hell out of HDV ASAP. But he suggested a good capture route would be component to Blackmagic Decklink to ProRes422 SQ. What do you think of that?

HQ would be more useful for HDCAM captures-- apparently overkill for HDV.

So I've learned HDV timecode sucks, it's better to get into a reliable code format-- again, ProRes, and ways to do it, including ways to process existing captures via Compressor.

This is why Larry takes affordable credit card payments for detailed phone time and research for you-- a great secret weapon. Highly recommended. I think this was a nice gift for us all.

www.larryjordan.biz

- Loren
Today's FCP keytip:

Play from In to Out with Shift - \ (Backslash) !

Final Cut Studio 2 KeyGuide? Power Pack.
Now available at KeyGuide Central.
www.neotrondesign.com

Loren wrote:
>then he dashed it all by being a true gent!

Harry wrote:
> R.I.P.

Dying makes a true gent!? Geez! I'll never understand women (urrr... now i have genders all mixed up...)

Ah... General Mike.. Didn't catch the podcast, unfortunately. Too little free seat time these days to run tests.

I dunno, Loren, about the component thing.. if you have decklink extreme HD (I don't know which other card supports HD-SDI off hand) and a good deck, you could capture that as ProRes via HDSDI. Which will be my number one choice, because 1, it's fast, and 2, it's reliable. Second option would be FW capture and a batch transcode. If recapturing is needed, then batch recapturing and batch transcoding. It's slightly slower (or you can retranscode the HDV files if you have them backed up somewhere).

As far as capturing uncompressed goes. I go on ProRes, because you're unlikely to spot the difference, and it doesn't take up as much space.

Graeme Nattress recommends editing native in HDV because it adds compression on top of compression. But I'm that more conservative when it comes to GOP formats, which seems to receive the most amounts of complaints.

Graeme also has an old article on Ken's site about SDI chroma smoothing. Components may also be smoothing out the footage during decompression. So each method may add its own "flavour".
[www.nattress.com]

Then again, I'm thinking of adopting Jeff's workflow whenever I have to face budget basement set ups (limited HD space, no HD-SDI captures, etc).



www.strypesinpost.com

Be careful there. When you say "batch recapture and batch transcode," the hair on the back of my neck stands up. When you run HDV through Compressor to turn it into ProRes, the original timecode track is lost. If you subsequently cut that footage into a Final Cut project, you won't be able to recapture it.

The only way to make this work is to log-and-capture your HDV footage into one bin, keep those clips around, run the media files through Compressor to generate new, no-timecode ProRes clips and bring those into a second bin. If you lose all your media, recapture the original HDV clips in that first bin.

The downside there is that there's no easy way to just recapture the footage in your timeline. Normally, you can select your timeline, do a batch capture, and Final Cut will just grab the footage you need, leaving all the unused media offline. In this two-bin system, you'd have to manually correlate each ProRes clip on the timeline with an HDV clip in bin one. Not the world's greatest workflow.

Yet another advantage of the edit-HDV-into-a-ProRes-timeline approach is that your original media files retain all their timecode and reel number information throughout the whole workflow. They can be re-batched at any time.

>When you run HDV through Compressor to turn it into ProRes, the original timecode
>track is lost.

Ahh... Completely missed this one. Compressor doesn't keep reel names nor timecode information. Then it has to be exported out via fcp, which will then be an overnight process...



www.strypesinpost.com