Setting the record straight about ProRes workflows

guomo

New member
First off, I realize the title of this post sounds a bit arrogant. It is not my intention to be so, I am trying to be helpful and hopefully save a lot of folks time and effort and disk space. I have deep experience as a former QuickTime developer for Apple and come at this from both sides: a film maker and a programmer. I got a little frustrated reading posts where well-meaning folks are essentially suggesting the most bizarre and wasteful workflows or flat out give wrong information. In the end their workflow may work, but it is likely because of luck and their willingness to waste time with bigger data.

So there are countless threads wherein people are asking about what codec to transcode their AVCHD footage (And even other formats) to. Even if they should at all. It is a deep topic and I will limit my points to those whose footage originates in AVCHD; though much of the facts and concepts are poratble to other formats as well.

Truth is, anything more than Pro-Res Proxy for any AVCHD footage is a waste of time and space. For most projects you can transcode, edit, and finish your entire film with just Pro-Res Proxy. There is little reason to use LT, HQ, or 4444.

The short answer why is two fold: AVCHD tops out at 25mbs so any codec that is greater is useless and that your destination output is unlikely to be one that can handle the higher color space - possibly bitrate too.

Because the AVCHD spec limits maximum bitrate to 25mbs for any size video, no amount of extra bandwidth in the transcoded format will get you better video. To think otherwise is analogous to thinking if you could make $10 bill physically bigger it would buy you more stuff. You simply can't get more data from something that isn't there.

However, I can think of a few reasons that most might want to use the other ProRes codecs besides ProRes proxy (not an exhaustive list but the top 3 reasons):

1) You shoot some other HD format besides AVCHD that exceeds the 45mbs data of the proxy codec like RED, Viper, Phantom, uncompressed HD etc.

2) You have a portion of your sequence that uses green screen, lots of filters, heavy alpha channels, etc. and you don't want to wait around for the render every time you touch something in your timeline. In this case you would transcode the SFX-heavy portions from AVCHD to LT, HQ, or 4444 depending on how many effects and alpha channels and round trips you expect. Do all your wizardy in a separate sequence set to the better codec, then export using QT conversion as Proxy and drop the clip into your Proxy version timeline with the rest of the footage. You can always tweak the hi-res sequence and re-render if you need to change things.

3) Your final destination output is for a 10-bit digital cinema projector, film, or another format that is uncompressed HD; some film festivals will ask for this if you get accepted but your submission is invariably on DVD with a top bitrate of about 10Mbs.

But what about the better color grading idea that everyone is talking about with the ProRes 422 or 4444 versions? Shouldn't I use that to get the best quality grade? Maybe. It is partially true in that the app you will get finer-grained adjustments because of the wider colorspace, but your ability to adjust that finely is directly limited to your ability to see it. Is your grading monitor a $3600 sony LMD-2450MD or equivalent that has true 10-bit signal processing? If so then go for it. Or perhaps you outsource the grading to a pro colorist and she has top of line gear. Then 422 or 4444 is your best bet...maybe, read on.

However, if you are like the vast majority of us and are using stock montiors in the $1000 or less price range, then you will hardly be able to see this difference. The biggest factor though is your final output format. If it is anything on the web or a disc of some sort then all your color grading efforts - if you were even able to see the difference on your crappy monitor - will, in an instant of export, be lost in the downsampling and quantization to the lesser-grade format. And worse, you have no choice in the matter.

That last part is critical. Think about the difference between 1024 values per color channel (10-bit) color and 256 color values per color channel (8-bit). When down converting it means that possibly 3 out of every 4 colors gets pinned/quantized to some value that it wasn't originally. The point is you may not like the effect. If you keep it in 8-bit because you are going to 8-bit in the end, then well, you will always know what you are getting. I realize this is an over simplification and that ther are various sampling means within the world of video, but the principal is true: data is being thrown out.

So in summary, set your FCP Log and Transfer preference to ProRes proxy and ingest and edit away. You save gobs of HD space and won't lose a bit of quality (scenarios describe above excepted).
 
lol- great first post but couldn't disagree more. ProRes Proxy falls apart much faster than any of the other flavors and yes you'll see it real quick the second you start applying filters, transitions, etc i.e. anything that adds a generation worth of render. And that compounded over the life of post will definitely result in lower quality. Get a bigger hard drive and use ProRes- doesn't have to be HQ but at least ProRes regular not Proxy. Drives are dirt cheap these days i.e. 1TB internal = $75. If you follow this gents advice you'll be pennywise and pound foolish.

Noah
 
And that compounded over the life of post will definitely result in lower quality.

Compounded over the life of post?? What do people do in post, re-render after re-render? You just create the master files and everyone works from those. If you happen to use programs that can deal with the native files great! If not and you use FCP convert them to prores (not proxy though, that is a BAD idea).
 
Faulty premise! ProRes is an intraframe codec, while AVCHD is interframe. This means that AVCHD can do a lot more with 25mbps than ProRes can; and H.264 is much more efficient than even other long-GOP codecs. So, one needs a much higher bitrate in ProRes than in AVCHD to get comparable quality.
 
It depends on what your editing and the extensiveness of the edit.
If I have simple edits, family stuff I throw h264 files on a timeline, cut, ripple, export, done.
If I'm on a job where we are continually altering in/out points, cross fade edits, dip to color edits,
lower thirds you name it there is no way h264 would fly.
Remember in 24p, h264 is only two frames a second. The other 22 frames are interpreted sequentially from those two.
eg to see frame 4 you get frame 1, calculate frame 2, from that the processor calculates frame 3, from that the processor calculates frame 4.
a mental picture it would kinda look like this
i..........i...........i..........i...........
(where the i represents the i frame and the . represents the next eleven calculated frames)

if you edit and your second clip starts on frame 4 then it may look like this
i...........i...........i...(edit point).......i...........i...........
so you see the frames after the edit point are now dependent on an i frame that is not even in the timeline!
And that's just on a simple cut. Dissolves are the worst and most prone to errors.

h264 is a great delivery codec and was never meant to be an editing codec.

http://www.dvxuser.com/V6/showthread.php?p=2071129#post2071129
 
As a fellow programmer, I will tell you two reasons why your logic is faulty.

1) If you process your footage through multiple passes of filtering/effects, working at 8-bits as your working bit-depth and as your target bit depth is going to propagate quantization errors at each pass. There is a reason people work in 24-bit audio when they are going to mater it to 16-bit. It allows you to reduce quantization effects for when you get to the final edit. You ALWAYS want to work in a higher bit depth unless you are making NO changes at all to the footage.

2) Data rate is NOT an indication of compression quality, period. If you read as many threads as you have, you'd see all the discussion even with people comparing the AVCHD compared to something like XDCAM an how mb to mb, the AVCHD is better at the same rate. Further, Panasonic actually released a document where they tested all the ProRes formats and recommended ProResLT as being sufficient.

And to the person who asked about re-rendering, at least the editor I use, yes. I've watched him work and he'll go through and match all the footage (is that called a one light?) and export it, then he'll start cutting it all together and export it, then he'll take that and pass some of it to me for VFX in AE, which then comes in and out of AE. Then he'll go through a final pass with color correction. So in this case, it at least gets written to disk maybe 5 times at a minimum (including the initial conversion to ProResLT from whatever source).
 
Not to mention round-tripping shots into other programs for color correction/effects compositing, etc which is precisely when you want a codec that stands up to repeated rendering/generations. Like I said- this workflow makes sense maybe for cuts-only super simplistic editing when you're at the South Pole research station and there's no Fry's/Amazon.com ready to ship you a new, cheap external drive. But in any other instance will result in noticeably reduced image quality. Hard drive space is so dirt cheap these days- even on laptops- this shaving MB's at all costs mentality is totally unnecessary.

Noah
 
ProRes Proxy was created originally for use with Final Cut Server, but has lots of other uses. There is so little information per pixel, it is for pre-viewing low res, and yes, falls apart quickly when you start stacking up filters, round tripping, compositing, etc. Not to mention it totally suck for color grading (ask any professional, experienced colorist, they won't touch Proxy). And data rate does have a direct bearing on the quality of the footage. Higher data rates can carry more data per pixel. So you absolutely want a data rate in ProRes slightly higher than your original Long-GOP, highly compressed original. So you want a ProRes flavor that support at least, if not slightly higher than your original format. Plus that data rate is influenced by frame size and frame rate. It's not nearly as cut and dry as the OP claims. I'd never use Proxy for anything but preview QT's that need small file size or offline edits.

http://documentation.apple.com/en/finalcutpro/professionalformatsandworkflows/#chapter=10&section=0

ProRes White Paper
http://www.apple.com/finalcutstudio/docs/Apple_ProRes_White_Paper_July_2009.pdf
 
Boo! Now im confused again... I will stick with ProRes HD

I've played a bit with the other tools, but for fast workflows and simple work (i.e. Weddings, live events, youtube/vimeo stuff). I think FCP's default for my GH2 - Prores HD - is the safest to deal with.
When the project is done, I'll probably trim what I dont use and trash it, keep the AVCHD RAW footage, buy more Hard Drives :P
I'm of the belief that bigger is better. I rather edit with FCP/AE with confidence... even if it will end up on a DVD or some cloud out there...

When it comes to more "professional work" (Short films, festivals, etc), there are even more tedious tools/formates I can use to create even larger files...which ill probably do and use at the end of the day.
There are tools one can use to squeeze every drop out of your video files. Here are a few:

5DtoRGB (Hands down the best quality)
Clipwrap
MPEG Streamclip

I'll adjust my given workflow depending on the project and time constraints.

Im only providing my 2 cents because all this pixel peeping is really all over the place, so for the average joe... this is my take.
And heck, Apple will be native AVCHD in only a few months. Adobe is already native AVCHD, so all this talk may be pointless....

Dai-Lon
http://www.dailonweiss.com/
http://www.kristinaweiss.com/Magyar/
 
I am more than happy to transcode my Nikon footage to ProRes HQ. Yes it may give a larger file size / data that is unnecessary, but I have found it holds up well to multiple filters, and I never see any noticeable difference in render times when something does need to be rendered.

I am sure the HQ is overkill, but I tend to stick with what works. Since I most times cut transferred 16mm into digital files (and get the film telecined or scanned at a bit depth that will allow for maximum color correction) I also stick with the same codec, as all my transfers get dropped onto my hard drive as Uncompressed and ProRes HQ
 
Back
Top