1 person found this helpful
It's no secret that you should try to keep as much data as possible, all the way up to final output.
I would suggest to either use Dynamic Link to go between AE and PPro without having to render, or use the free and cross-platform Avid DNxHD QuickTime codec which has similar quality and file size options as ProRes.
If you render as 16 bpc (or 32 bpc) from AE as "Trillions of colors" to 1080p DNxHD 185 10-bit RGB, you'll have great video with minimal lossy compression.
Thank you very much for the quick reply and great info, Jonas.
Does anyone know of any articles or videos that fully explain codecs and exporting so I can fully understand what is happening from a technical point of view? I seem to only ever find part of the story.
1 person found this helpful
A few hints:
QT Animation+ supports alpha and is lossless, but only 8 bpc and is large (unless you have lots of areas with the exact similar color.) QT PNG is similar but can be smaller than Animation.
DNxHD can be > 8 bpc and support alpha and have several compression levels, but is not lossless. Same goes for ProRes, but ProRes is not cross-platform. ProRes can be read on Macs, but on Windows the user needs to download the codec to be able to view (but you can't write to ProRes from Windows.)
DNxHD needs a download on both platforms, but both can read and write.
H.264 and other MPEG-4 is for viewing only, avoid for any production. Same goes for MPEG-1 and MPEG-2.
QT PhotoJPEG doesn't support alpha and is 8 bpc, but can be viewed on everything that has QuickTime Player and VLC. Can be almost lossless with Quality >96%.
Again, I cant thank you enough.
1 person found this helpful
Regard to when to rely on Adobe Dynamic Link and when - on a digital intermediate, see this thread in PrPro Forum. Pay attention to Todd's comments.
Regard to codecs for digital intermediate:
- on Windows you can decode (read) ProRes once you have QuickTime installed, you can encode (write) to ProRes via FFmpeg;
- have a look at UT codec - in my testing the quality of UT 4444 (ULRA) and UT 444 (ULRG) and render time is identical to TIFF and TGA sequence, whereas file size is 3+ times less.
Thank you also.
I keep forgetting to mention that I am on a Mac.
1 person found this helpful
As is stated above there are Production codecs which can be lossless, nearly lossless, and at higher bit depth. Production Codecs include Uncompressed (huge files but fast do decode and will not playback in real time) Editing Codecs like Black Magic, Avid, ProRez, and others that are smaller, can still be lossless, may support alpha channels and higher bit depths, and Archival Codecs like Microcosm, JPEG2000, PNG, which support alpha channels, are lossless, and on some systems, will playback in real time.
Editing codecs are specifically designed to be played back in a NLE. These are good choices for a Premiere Pro / AE workflow. So are Archival Codecs. If your project involves a lot of sharing of files between users and apps, then pick a format common to everyone that is at the highest bit depth your storage system can accommodate.
There are also image sequence formats. These include PSD sequences, TGA, TIFF, PNG and others. Some support higher bit depth (32bit Float). Image Sequences work great, have advantages when it comes to correcting just a few frames here and there, and are used to a great extent in professional production pipelines. They are much more flexible than video but they have a distinct disadvantage when it comes to keeping track of the files and keeping things organized.
Then there are delivery codecs. These are designed for playback. They are not suitable for production because delivery codecs use GOP compression, are limited to low bit depth, and are not lossless. IOW, if you render to a delivery codec, then you process and render again to the same codec the artifacts start to build up. H.264 for example, after just 2 or 3 passes, even at very high quality settings, falls completely apart.
Consumer and ProSumer cameras use delivery codecs for image capture. Your Cannon 5D does. This is OK, and it's the only option you've got with some cameras, but the best workflow for these types of capture systems is to transcode the footage to a production codec before you do anything. Production codecs will render faster because it takes less time to decode the footage for processing. You should never render AVCHD or MPEG footage back to the same codec. Always use a production codec until you're project is ready to archive and deliver.
One note on ProRez on Windows. This is not a good idea for production. You cannot maintain proRez native 10bit color space. You should not render to ProRez using 3rd party workarounds because you're just building a proRez container for a windows compressed video. Sure, it will playback on a Mac, but the actual image will have lost important detail and color depth.
Let me give my humble opinion on Dynamic Link. This works great for a shot or two in a Premiere Sequence. This works great for working with Audio in Premiere and Audition. This works great for rendering for delivery or even production using the Adobe Media Encoder. Did you note my second sentence? Dynamic link between AE and Premiere Pro for a few shots in a sequence is a good idea and a time saver, but if you're cutting a film that's over a few minutes, each scene should be it's own sequence. If the scene has more than just a few shots that need to be processed in After Effects then using Dynamic Link can easily become a monster you don't want to deal with. The key to efficient post production of an effects laden production is planning. You've got to cut for story first, then produce your effects shots with sufficient handles (extra frames) to allow some fine tuning, then recut the final. Any other workflow on a long project is just wasting time.
Then after familiarising yourself with the basics do not hesitate to run your own tests so as to decide which codec serves your needs better in a particular circumstances (unless you just have to comply with e.g. a broadcaster requirements). I use this 'Difference' one.
Although DNxHD is widely accepted, I notice sometimes such things while performing the test:
Whereas the test for the same frame with UT 4444 (which is available for both Mac and PC) looks something like this (exactly the same way it looks like with TGA or TIFF sequence):
Thanks everyone again. This kind of info is invaluable. I really appreciate it.
Based on what Rick said, I have another question(imagine that) and hopefully I ask it correctly.
In a video on the adobe site, the instructor mentioned that the Mercury Playback Engine in Premiere Pro brings footage in as/at 4:4:4 32-bit float and that there is no need to transcode. While I understand that I may need to transcode for AE, is it still better to transcode in general? Even directly into Premiere Pro? And does AE do the same thing? The footage looks ok when i bring it in to AE in the form it came it from the 5D Mark III.
Also, regarding transcoding. Do you transcode files for editing or effects, and then transcode them a different way and re-import them into the NLE when you are ready to output a final product?
BTW, the AVID codec is working well. Havent had a chance to try the others.
Well, first of all about encoding to ProRes on Windows. The only reason I mentioned that option was I had misread Jonas's comment and it somehow printed in my memory as 'one can (easily) write to ProRes on Windows'. However, since I have never tested it by myself (and not in the mood to do that in the nearest future), I would argue neither in favour of nor against. The fact that an application is an Open Source one doesn't mean it is less smart than a brand.
Second, If you had a look at the thread in PrPro Forum, link to which I submitted in my previous comment, you would see that we had exchanged a few words about Jason Levine 'evangelical' presentations as well.
And if you were addressing me of whether it is better to transcode in general, I would say 'no', because it highly depends on your workflow. Transcoding may yield significant time savings (e.g. if you need to render more or less complex composition several times - either RAM preview or media file), whereas I don't see any advantages in terms of quality (Rick may have an opposite opinion).
To sum up, I would just quote Rick, 'The key to efficient post production of an effects laden production is planning'.
You can bring your camera original into PPro and it will do a great job of producing video. If the project is small and simple this will work just fine. If the project is large, you have double system sound or are doing extensive sound editing, and you've got a bunch of hands in the pie then I'd suggest using Magic Bullet Grinder (Mac only) to transcode, add timecode, and notes, and start a footage log.
FYI, I prefer the Black Magic codecs to the AVID because I've seen fewer bugs as versions change.
As far as transcoding, that's taking the original to another codec. I only do that once. Every render from there on goes to a production codec. Once you're using a production codec keep using the same one unless there's a specific reason to change. I don't just transcode footage to put it in another format. I never re-encode footage unless I'm adding something to it.
I was going to comment on the planning part. I agree in many ways. And perhaps over time even more.
But this goes into another area that I would like to ask you guys about, just some other time, because I have to finish this.
The whole idea of what I will just call the "studio model" that comes with 15+ people that are all experts and cost alot of dough vs. the "new production model" that utilizes way fewer people, often just one person, and achieves that same results for far more profit at the same level of quality.
Not to mention the process of planning. I agree that planning is important, but as a graphic designer also, it doesnt leave me room to discover anything or at least as much.
Just a little background, I was a film prod major at a small college, but took the electives for graphic design and that is what pays the bills...i didnt have a trust fund to be able to afford USC or anything like that...lol...but i am kind of glad i couldnt after all...
I understand the old model and why it evolved, but the ever lowering cost of high quality, intuitive tools cant keep a good idea or great storytelling down like it used to. Its all at your fingertips and economically accessible. Or at least in reach. With Adobe Creative Suite Master Collection, an HDSLR, some prudently purchased equipment, and apps, it is now possible for someone like me to not only come up with the content/idea, but execute/produce it at every level from the motion graphics to the marketing in all channels, composing and mixing the score/tracks, etc...for the cost of a small regular car...about 30K.
Now THAT, is utterly transformatory...
Well, enough. Thanks again to everyone for helping me wade through the technical stuff. I really appreciate it.
I've used DNxHD on a daily basis between AE and Avid for years and
have yet to encounter any bugs (other than the other side using a
really old version of the DNxHD QuickTime codec, which was easily
fixed by having them upgrade.)
Jonas, the codec has worked perfectly so far...
What kind of computer would take 80+ hours to render? That seems really long. And why wouldnt you break the project up into smaller more manageable chunks?
I am using a 2X3 GHz Dual-Core Xeon with 19GB of ram, System 10.7.5. Or my powerbook maxed out with ram...8GB if i remember correctly. Either one I didnt think was too beastly...LOL...neither seem to take that long to render.
Now, my sequences are probably very simple, but I just cant wrap my head around 3+ days of rendering...LOL...i dont think i have that much patience in me to wait for that and find out it didnt look so good or there was a tiny error or something...LOL...
Man, that would be frustrating...LOL...
The spec and the composition description I was referencing, including CPU usage level while rendering, is presented in this thread (you could find this link inside of my comment within the 'slow to render' thread as well).
Since I had time to dabble with and was interested in gaining an understanding, I passed trough up to the end. Now this is a perfect example in which circumstances and why it is unwise to rely on Dynamic Link, because the difference is dramatic.