What's the best intermediate codec to use?

September 22, 2005 12:17 PM

I have some video on a DVD that I want to edit in Final Cut Pro. What's the best intermediate codec to use? I don't have space for uncompressed video, and the rest of my footage is DV-quality, so I can afford a bit of quality loss.

I've been using Apple Pixlet at 50%, but is there a better choice? There's one listed in QuickTime's export dialog called "Apple Intermediate Codec", but I haven't been able to find much information on it.

Also, what's the best codec that supports alpha channels? (i.e., if I'm keying footage in After Effects but compositing with the background in FCP. I've used Animation for this, but it produces huge filesizes as it's not optimized for video, and is difficult for my system to play back in real time.)

Is there a simple guide somewhere to what codecs are best for what purposes? (i.e., "don't edit in this codec", "this codec is too lossy for intermediate use".) I've done some googling, but for many codecs reviews of compression quality and possible uses is hard to find.

posted by tweebiscuit to technology (6 comments total)

To clarify, I'm not going to be outputting to DV, or even editing the footage into a DV sequence -- just that the end result need look no better than standard DV footage.
posted by tweebiscuit at 12:31 PM on September 22, 2005

You might consider huffyuv. That page only has binaries for win32, but since it's open source there should be an OS X version somwhere, but I don't know anything about OS X.

Huffyuv is a lossless video codec, much like FLAC is to audio. It will not be nearly as large as uncompressed raw video, and it should be quite fast to encode/decode. However, it will still be much larger than a "normal" video codec.

Another advantage is that it stores the data in the YUV colorspace (as the name implies) which is good because many codecs naturally operate in this colorspace. That means that if you take source material -> huffyuv -> final codec, you don't have to do any colorspace conversions if you keep everything in YUV. Converting between RGB and YUV is not an exact process and does introduce some loss in quality.
posted by Rhomboid at 1:59 PM on September 22, 2005

Hmm, that's interesting. However, not really perfect for my purposes, as it looks like it's Windows only, and not really industry-standard. I'm a high-level amateur (working to break into professional work), so I'm curious about what the accepted wisdom is. Also, judging by the datestamps on the rest of his pages, that codec is 5 years old! Thanks for the suggestion, though, the YUV info in particular was enlightening.
posted by tweebiscuit at 2:07 PM on September 22, 2005

The Quick answer: Animation codec. You may not like it, but it supports an Alpha channel and keeps 100% of the quality. It does what you need.

Workflow is: Animation with synthetic AE elements, composited on DV elements.



Long side- one by one.



Video that starts on a DVD

This is MPEG-2. It's a very lossy format to begin with. Can you get the original footage in the original format? You shouldn't be too worried about losing any further quality if your original is going to be transcoded MPEG-2. If this is what you're going to use...MPEG-2 that you're going to transcode...you might as well transcode it to DV. It'll be fast + cheap.

So, the standard DV+Animation workflow should work great- especially if your source is MPEG-2 via DVD. This is what 99% of people use (not the mpeg-2, the DVD+animation)

What are you finally going to use this as? DVD? Digibeta? Knowing this format helps pick a hardware format. It's ugly to transcode from one Raw format to another - it's best to always work in that finished format.

Now, FCP is primarily optimized to act as a DV program. It's capable of doing much, much more...but 99% of it's usage is DV.

The AIC (Apple intermediary codec) was developed to handle HDV in a non-b/p frame codec (all I frames). FCP 5 can handle the MPEG-2 of HDV on the fly...but it takes quite a bit of horsepower. AIC is how iMovie and FCE HD handle HDV.

FCP has always meant to have agreement of capture format + sequence format. Make sure stick to the Apple prebuilt sequence formats if you want RT capability (at all!).

The Pixelet Codec was developed by Pixar for animation to preserve quality and reduce file size.

Using AE?
Use Animation @ 100% - that's smaller than uncompressed. Runs about 1 gig/min.
DV is 5:1 Compressed and runs about 5 min/gig (but does not support an alpha channel)

It sounds like when you say "if I'm keying footage in AE" are you actually talking about creating an alpha channel to your footage....Keying generally refers to the creation of a black and white channel that gets inserted into the RGB file...so it becomes RGB + Alpha.

Animation is absolutely optimized for video, it is a lossless compression. But it was never meant to run at RT. For what it's worth, uncompressed video was usually meant to be accelerated with hardware. Apple, for the 'professional' formats are using either DVCPro 50 or Sony's iMX format.

The workflow would be Animation + DVCPRO 50.


Now, as to the 'codec' talk...
Apple has to keep inside of QT, a number of legacy codecs that you have no need for.

Legacy codecs: Cinepak, Compnent Video, Graphics, VIDEO
Codecs shared with a 'still' format: BMP, JPEG2000, PNG, TGA, TIFF, Planar RGB

Distribution formats (meant for web, etc.): DivX, 3ivx, MPEG-4, Sorenson, Sorenson 3

Legacy hardware formats: Motion JPEGA, Motion JPEGB

Specific uses (such as video-conferencing): apple VC H.263, H.261, 262, 264

Independent format meant for compositing: Animation

Pixar generated optimized for animation: Pixelet

Apple uses (with FCP) P-Jpeg as an offline codec for DV

Current Hardware formats:
Avid (1:1, DV, Meridian Compressed, Meridian Uncompressed)
DV/DVCPRO (Pal+ntsc, HD 1080+720, DVCPro 50)
MPEG IMX (30,40,50) (sony IMX decks/cameras)
HDV formats: AIC, HDV 1080+720
Targa Cine YUV+YUV 16 (additionally cinewave)
*Uncompressed 10 bit + 8 bit

The hardware formats are meant to be utilized with specific decks, and almost never have the ability to imbed an alpha channel

The advantage of Animation..it's uncompressed no matter which of these formats you use and it supports an alpha channel. Need to see it playback in RT? Render.
posted by filmgeek at 2:58 PM on September 22, 2005

Thanks filmgeek, a lot of that was very helpful. I should have clarified MORE by saying that I'm not concerned with realtime output meaning viewing on an external NTSC monitor -- just being able to play the video without rendering. FCP isn't great at this with animation, but Pixlet seems to work out.

Final output is to DVD -- the FCP sequence will be converted directly to DVD format. And this project in particular doesn't need an alpha channel (and yes, by keying I meant defining an alpha channel), I was just curious to know what codecs with alpha channels were good for intermediate use.

However, though you advise converting my MPEG-2s directly to DV, I tried this first -- and after comparing the results to the same footage encoded as Apple Pixlet at 50%, the Pixlet was half the size and looked much sharper. Are you sure DV is the best solution here?

Still, though, you did answer one of my main questions when you said that the standard workflow was DV+Animation, which is good to know. :) Thanks!
posted by tweebiscuit at 4:32 PM on September 22, 2005

What's the footage of? Is it animation? Coming from an MPEG-2 file?

Traditional animation or Computer animation? Do you know what the FPS was originally? Traditional animation is 24fps (really 12 fps x 2), where often, much computer animation is 30fps.

Rather than transcoding the MPEG-2 to DV (which gives you both RT output and the ability to end up in NTSC space..) you coudl pick a different Video format. For example, DVCPRO 50 has 2x the data rate...(and therefore woudl be 2x larger). But it will play in RT.

And that it is an NTSC compatible format is hugely important...if you want the video to stay legal (off in the distribution arm, when you make it a DVD again.) You can't really skip thsi step (monitoring the footage in NTSC...or at least in scopes.)

Pixlet is a wavelet codec...as you decrease the percentage it gets more compressed. Animation would be much smaller if you made it 50% too...but in both cases you're throwing out info.

BTW, FCP5 has the abilty to turn off the 'red' render bars...and go 'unlimited RT' even with animation codec...but still may need to be rendered unless you're on the very fastest macs.

We can continue this here, or direct from my mefi email...the details would hugely change how and why) I'd approach the subject.

I don't think necessarily DV is the best solution...but I know that Pixlet was planned for playback of Animation...not compositing or editing.
posted by filmgeek at 5:20 PM on September 22, 2005