|
|
Up size - then downsize - is FCP smart?
Hello,
I'm running FCP 7.0. So I'm working in a 1920x1080 timeline with applePro Res 422(LT) as my compressor. Most of my footage was shot in that format.
I'm also bringing in a lot of stock video with resolutions lower than that (sometimes as low as 640x360).
It also turns out that the final cut needs to be delivered on a SD DVD. So my question is:
Can I simply enlarge the lower res footage in my HD timeline, then when I go to export for DVD (using compressor) is FCP smart enough to realize that that footage does not neccessarily need to be blown up 300% and therefor keeps the original resolution information (which matches nearly perfectly for the SD DVD)
Does that make sense? I'd like to keep working in my HD timeline if possible.
Thank you.
ok thanks. Then what should I do? Work in a SD timeline?
> Then what should I do? Work in a SD timeline?
How much stock footage are we talking about?
If it's not a lot, then I might consider staying in HD for now. Just because they say they don't want an HD version doesn't mean they won't, eventually. Because if they do end up wanting an HD deliverable, then your move should be to upres the SD footage, hopefully with hardware conversion. For example, did they ask for web movie files? Lots of clients don't realize that a 720x486 movie file is quite small by web standards. You don't want them complaining when you give them a 640x480 web movie and then they say, "Can you give us 1280x720?" after you throw away the HD timeline.
If you're absolutely certain that they want SD, then go to an SD timeline. Make sure you check to see if they want the HD footage letterboxed or panned-and-scanned. Huge difference. Moving to an SD timeline affects everything and it's not that easy to go back, so make sure it's the right move before you do it.
www.derekmok.com
o.k. if someone can help me identify the SD timeline settings that would help. So I'm going to rebuild the whole thing in SD for export to a Widescreen SD DVD. FYI - I shot all my footage with a canon 60D at 24fps and brought it into FCP by ingesting it as ProRes 422 (LT). I'm also buying a bunch of SD stock. And have some H264 SD movie to put in there as well. I plan to "Export using compressor" when I'm ready to make my DVD.
This is what I'm thinking for the sequence settings - please tell me if I'm wrong anywhere - It's been a while since I've worked in SD for a DVD.
Frame size: 720x480 (NTSC DV (3:2) )
Pixel Aspect Ratio: NTSC - CCIR 601 / DV (720x480) ANAMORPHIC Checked!
Editing Timebase: 23.98
QT Video settings: Compressor: Apple ProRes 422 (LT) (I'm figuring this reduces my rendering time while I'm working in the timeline).
Thanks!
Don't use DV codec, please use ProRes, it is much better
Logan6 Wrote:
-------------------------------------------------------
> o.k. if someone can help me identify the SD
> timeline settings that would help. So I'm going to
> rebuild the whole thing in SD for export to a
> Widescreen SD DVD. FYI - I shot all my footage
> with a canon 60D at 24fps and brought it into FCP
> by ingesting it as ProRes 422 (LT). I'm also
> buying a bunch of SD stock. And have some H264 SD
> movie to put in there as well. I plan to "Export
> using compressor" when I'm ready to make my DVD.
>
> This is what I'm thinking for the sequence
> settings - please tell me if I'm wrong anywhere -
> It's been a while since I've worked in SD for a
> DVD.
>
> Frame size: 720x480 (NTSC DV (3:2) )
> Pixel Aspect Ratio: NTSC - CCIR 601 / DV
> (720x480) ANAMORPHIC Checked!
> Editing Timebase: 23.98
>
> QT Video settings: Compressor: Apple ProRes 422
> (LT) (I'm figuring this reduces my rendering time
> while I'm working in the timeline).
>
> Thanks!
The DV in the frame size and PAR is purely frame size and PAR. The compressor he chose is ProRes LT.
Personally, I'll do everything up in HD and at least in ProRes SQ, since principal footage is in HD, and that also gives you an option to send out an HD sized output.
www.strypesinpost.com
Quote
Logan6
... is FCP smart enough to realize that that footage does not neccessarily need to be blown up 300% and therefor keeps the original
resolution information (which matches nearly perfectly for the SD DVD)
Why think that would be smart, even if FCP could do it?
The key is in your word "nearly". Transforming video to nearly the same resolution is a serious, and therefore visually lossy, transformation. All, or nearly all, the pixels must be changed. Only if the original and final resolutions are exactly the same is there benefit in avoiding the uprez-downrez course.
To do this job right you should not use ProRes (LT) but a codec like ProRes (HQ) which can hold visual quality through multiple generations.
When the codec itself adds no garbage, a considerable uprez is not visually lossy. (In fact, in your particular case of transforming 640x360 to 1920x1080 where 1 pixel becomes 3x3 pixels, if the uprezing is of the crudest "nearest pixel" kind, the 1920x1080 image looks identical to the 640x360 image.)
Generally it is the downrezing that is visually lossy, but when the codec itself adds no garbage, the uprezing followed by the downrezing is no more visually lossy than a single transformation to nearly the same resolution.
Dennis Couzin
Berlin, Germany
You don't want any H264 in your master program-- it's heavily compressed and not an edit codec. This material transcodes very nicely to ProRes SQ.
Ideally you want all footage in the same format and that includes upscaling undersize footage to your program frame size.
The codec used is crucial, otherwise the upscale will soften the image, just as it will with an undersized photo. When it comes to SD formats for your HD timeline, if you're stuck with them, consider uprezzing through a professional service, and letterboxing the widescreen originals for the SD version to avoid stepping on the footage in downrezzing, as discussed above.
There are specialty post houses offering upscale through hardware, using tools like a Teranex, or a Miranda box, which cost thousands for its simplest products, and an experienced user of it should comment here, because upscaling legacy video is a hot topic!
There is also software out there-- I was using Topaz Enhance for some of this work. It's been discontinued. It worked, but was very slow, even on an 8-core. 1 minute= 1 hour! Not 64-bit. Excruciating. Understandable considering what its algorithm was actually doing-- comparing and borrowing additional image data from neighboring frames rather than trying to reinvent pixels in each frame which would soften the upscale. It gave pretty darn good results, at least to 1280-- and also refined inferior SD original transfers from VHS to dramatically sharper DV. Carpet details for instance came back!
I posted a Digital Restoration sample on my YouTube channel (Neotrondesign) showing treatment of a 16mm to SD AVI transfer which was uprezzed to DVCProHD 720p, accomplished in Topaz Enhance plus a lot of dustbusting in Photoshop Extended. I probably botched some of the steps on this first outing. There's too much Sharpen here and there. Check out the file "neo_digrestore.mov" It's a service I offer for those who want to clean up borderline useable stock, but it isn't cheap doing all that custom dustbusting, and if you have time, the dustbusting alone improves it.
- Loren
Today's FCP 7 keytip:
Summon your Video Scopes with Option - 9 !
Your Final Cut Studio KeyGuide? Power Pack
with FCP7 KeyGuide --
now available at KeyGuide Central.
www.neotrondesign.com
Loren, your comments on sophisticated uprezzings are interesting but not relevant to the question of this topic: the lossiness of uprezzing followed by an almost equal degree of downrezzing. Sophisticated uprezzing can produce a HD which looks, and is, slightly more detailed than the SD it came from. But this improvement can't be maintained when the HD is then downrezzed to SD. My point was that even unsophisticated uprezzing is hardly lossy -- the HD doesn't have visibly less detail than the SD it came from. The downrezzing is the lossier step.
I just now had a laboratory make me a 1440x1080 10-bit uncompressed transfer from a Beta-SP tape. I chose this to get everything I could out of the tape. My first need is to intercut it into a 720x576 ProRes sequence, so I'll reduce it for this. The lab man knew I'd do that and he'd tried the reduction himself and reported that it looked worse than the Beta-SP original. So he suggested that I make the transfer straight to 720x576 ProRes and skip the uprez-downrez business. This is parallel to Logan6's case. But the lab man had only compared the downrez with the tape image, not with a 720x576 made from the tape image. Had he, he would have discovered the loss from the transfer of Beta-SP to 720x576. Like Logan6's near match, this is the lossy step, and the uprez-downrez won't be better, but neither will it be worse.
Dennis Couzin
Berlin, Germany
When it comes to interpolating data, less is better than more. DVD is hardly a good deliverable. And I am looking at it that you are working with principal footage as HD. That in itself should answer questions about future proofing your work as well a uprez/downrez.
www.strypesinpost.com
Quote
strypes
When it comes to interpolating data, less is better than more.
When it comes to interpolating data, more is better than less.
Suppose you know the color at the four corners of a square and wish to estimate the color at the center of the square. It obviously helps to also know the color at the corners of the neighboring squares. More difficult calculation, but generally better estimation.
General discussion of rescaling/resampling, which isn't exactly interpolation, deserves a new strand. This strand is specifically about uprezzing with the intention of downrezzing.
Quote
strypes
DVD is hardly a good deliverable.
It's a lot better than VCD, which is still pretty popular in India, the most color-conscious culture I know. Well-scaled well-compressed DVD can be visually satisfying. Unprofessionally compressed DVD isn't. Blu-ray, with its higher resolution and bitrate, is harder to mess up.
Where I work (play really) the delivery format is generally not known until the video is finished and out in the world. It might go no farther than YouTube or it might make it into cinemas. So we must work in high resolution. Is this uncommon?
Dennis Couzin
Berlin, Germany
Well, what I'm seeing here, is that his source footage is HD. The additional stock footage and stills are smaller than SD. Down converting his HD footage through a Teranex or an Alchemist would give him optimum results for final output to DVD (and working in 10 bits Uncompressed or DPX if he can afford it). But that is not the point. The point is if the client comes back the next day and asks if he can deliver in HD, he won't be able to, even if he shot his principal footage in HD.
>Where I work (play really) the delivery format is generally not known until the video is
>finished and out in the world. It might go no farther than YouTube or it might make it into
>cinemas. So we must work in high resolution. Is this uncommon?
So I presume you work in DPX throughout? Heck no! I would work in HD (since Youtube accepts HD), and when a film out is required, output a pull list for a DPX scan, then grade and online the thing in Smoke.
www.strypesinpost.com
> Where I work (play really) the delivery format is generally not known until
> the video is finished and out in the world. It might go no farther than YouTube
> or it might make it into cinemas. So we must work in high resolution. Is
> this uncommon?
A project that doesn't know whether it's going to theatrical exhibition or just YouTube? Yeah. Very uncommon.
www.derekmok.com
[Loren, your comments on sophisticated uprezzings are interesting but not relevant to the question of this topic: the lossiness of uprezzing followed by an almost equal degree of downrezzing. [
That's only ONE of the topics. It's completely relevant, to the OP's workflow I was addressing, which involves two steps, the uprez AND a downrez. Especially the stock footage-- again, this is a hot topic! We documentarians use it all the time and it's all gotta be treated for HD these days, or left small in the frame usually both letterboxed and pillarboxed, an unsatisfactory- if acceptably sharp, solution.
Logan wrote-
[Can I simply enlarge the lower res footage in my HD timeline,]
And you correctly replied, translating, that it would look like crap.
But Dcouzin, you also wrote-
[When the codec itself adds no garbage, a considerable uprez is not visually lossy. (In fact, in your particular case of transforming 640x360 to 1920x1080 where 1 pixel becomes 3x3 pixels, if the uprezing is of the crudest "nearest pixel" kind, the 1920x1080 image looks identical to the 640x360 image.)
At what viewing size? Everything looks good as a postage stamp. I simply can't believe this, not without intervening enhancement-- a strategy usually involving borrowing data from neighboring FRAMES, not nearby pixels. You simply challenge pixel physics. What am I missing here? An enlargement you describe requires inventing pixels and that translates to visual softness.
I would appreciate an A-B example of 1-to-3 SD to HD with no intervention resulting in visually lossless scaled to match the target HD frame. Specify the codec you use. I'm always willing to learn.
I would rather Logan uprez with intervening enhancement for HD in hardware, relevant hardware or software as suggested, and keep his non-enhanced original material strictly for an SD delivery and avoid up and down steps, which we agree can step on the image due to GIGO.
I have certainly uprezzed with the techniques I've described to ProRes 720P, and then downrezzed for SD widescreen DVD delivery, with decent results. But I like the idea of keeping the SD original for a DVD delivery and keeping the enhanced uprez for BluRay or file based.
I also have a hard time untangling your lab example-- a straight transfer of PAL BetaSP to PAL ProRes seems the right idea for strictly an SD delivery.
Best, as always,
Loren S. Miller
www.neotrondesign.com
Home of KeyGuide Central
Quote
Loren Miller
At what viewing size? Everything looks good as a postage stamp.
I wrote: Quote
...transforming 640x360 to 1920x1080 where 1 pixel becomes 3x3 pixels, if the uprezing is of the crudest "nearest pixel" kind, the 1920x1080 image looks identical to the 640x360 image.
"Looks identical" can only mean when viewed at the same visual angle.
"Visually lossless" means there is no loss in detail. There's no question of that HD image looking as good as a specimen of HD as the original image looked as a specimen of SD. It's the HD image having as much detail as the SD image that's relevant for the uprez-downrez question under discussion.
Quote
Loren Miller
I also have a hard time untangling your lab example-- a straight transfer of PAL BetaSP to PAL ProRes seems the right idea for strictly an SD delivery.
Sure it would be if PAL Beta SP were 720x576. But Beta SP is analog (at least in the scan direction) so transforming it first to 1440 pixels across and then down to 720 pixels across is hardly different from a single transformation to 720 pixels across. (I don't know what's going on vertically.)
Dennis Couzin
Berlin, Germany
What are you doing here?
Get back to work.
All the best,
Tom
Tom-
I'm doing these instructional discussions during coffee breaks. This one captures my attention.
DCouzin-
["Looks identical" can only mean when viewed at the same visual angle. ]
And what angle is that? Sideways? So far I'm not learning anything from this. I see nothing in your reply which tells me I can upscale an SD image to HD without special treatment. If it's possible, I'll be delighted.
- Loren
Today's FCP keytip:
Set a motion effect keyframe instantly with Control-K!
Your Final Cut Studio KeyGuide? Power Pack.
Now available at KeyGuide Central.
www.neotrondesign.com
Loren Miller: Google "visual angle" to learn what the expression means.
"Special treatment", or what I called "sophisticated uprezzing", is highly desirable when the goal is to make SD footage look more nearly like HD footage. Techniques that look at neighboring frames can be very powerful, because if the image moves a tad this gives fractional pixel information.
But the topic here is different. The uprezzing is expected to be followed by downrezzing, and the final output will be SD -- that's Logan6's case -- so there's little gained by sophisticated uprezzing. That's all I've claimed.
For example, suppose you uprez an SD video to ultra-HD, and suppose the uprez algorithm is so good that the ultra-HD practically captures the original scene from which the SD was shot. Now when you downrez that ultra-HD to SD all you can hope for is what SD can get from the original scene, like the original SD.
If the uprez algorithm is the worst possible -- nearest pixel -- this will sometimes give the exact same result in the final downrez as the most sophisticated uprez algorithm, but generally a somewhat worse result. Any half-decent uprez algorithm will give practically the same result in the final downrez as the most sophisticated one.
Dennis Couzin
Berlin, Germany
Visual angle defined as 0 degrees of arc, great. I suppose that can be assumed?
OMG, UltraHD... I'll have to add another page to my ScanGuide when that goes commercial in 2015.
[But the topic here is different. The uprezzing is expected to be followed by downrezzing, and the final output will be SD -- that's Logan6's case -- so there's little gained by sophisticated uprezzing. That's all I've claimed.]
Fair enough, thank you, Dcouzin. I'm focusing on uprezzing. I'm interested in seeing a visual example of this somewhere and with what codec:
[where 1 pixel becomes 3x3 pixels, if the uprezing is of the crudest "nearest pixel" kind, the 1920x1080 image looks identical to the 640x360 image.)]
And to pick your brains, another question: if one MUST downrez a 1080i timeline to SD, and if you had a wide choice of codecs available, say using Episode Pro, what would be your ideal choice of 1) codec, 2) container format be to minimize artifacts?
- Loren
Today's FCP keytip:
Set a motion effect keyframe instantly with Control-K!
Your Final Cut Studio KeyGuide? Power Pack.
Now available at KeyGuide Central.
www.neotrondesign.com
Quote
Loren Miller
Visual angle defined as 0 degrees of arc, great. I suppose that can be assumed?
Who said 0 degrees? All I said was "same visual angle". The 1920x1080 uprez made by the "nearest pixel" method from the 640x360 original has exactly the same visual detail as the original, so the two images look identical when they occupy equal angles of view. For example, if you display them each at 100% on two similar monitors you'd have to move the monitor with the 1920x1080 image three times as far away. This is so simple that it is uninteresting in itself. It was a part of a broader discussion of uprezzing followed by downrezzing.
It is interesting and important that a 1920x1080 uprez made by more sophisticated methods from a 640x360 original can have more visual detail than the original, but this belongs in another strand.
Quote
Loren Miller
if one MUST downrez a 1080i timeline to SD...?
The first thing I'd do is to deinterlace it, even if the SD will be interlaced. The only deinterlacers at my disposal are Compressor's, Video Purifier's, and Shake's. The first two are amateur junk and the third I haven't tried but suspect it is now incorporated in the first. We need a strand on state-of-the-art deinterlacing. Second step is downscaling. Here I think still image algorithms are sufficient. Shake defaults to a simple sinc filter. Third step, if necessary, is to interlace, which Compressor does properly. When doing multiple steps like these I work in either uncompressed or ProRes HQ.
Dennis Couzin
Berlin, Germany
|
|
All the news now and in your digital future
