We grouse about digital audio... What say about digital VIDEO?

[offramp]

Not a dumb question at all, ras.

 

Color gradients, for one. Depending on the parameters of the system, subtle variations in color get assigned the next higher or lower value, resulting in banding.

 

Another approach to compression is called GOP--group of pictures. It's a Sony technology; I learned a lot about it at a Sony tech school in New Jersey, several years ago. The concept is this: The computer takes each frame, looks ahead several frames (I want to say something like 18, can't remember) to determine if any significant changes occur. If there are none, then it sends only that original frame down the pipe, one time, and holds it there at the output until there is a frame that shows change above the threshold of the original frame. Cuts down on a lot of unnecessary data.

5 or so years ago, when I was the Chief Engineer at a post facility, we'd have lots of feeds monitored, and you could see evidence of this on CNN and the like, when they'd have a guest talking head on a show via satellite; if the person didn't move around a lot, and there were only changes in facial expression, you could see weird artifacts around the edge of the face, as if the speaker were really just some sort of robot or digital creation. If the speaker was in an environment with movement in the background--newsroom, say--then you could see artifacts as people walked through the background.

Another example would be: say you have a scene where two people are sitting on a rock, talking, and behind them is a huge, deep blue sky, maybe a couple of clouds. The computer sees all that blue, looks ahead, sees all that blue, and says "whatever. you get one frame". Now, the clouds are going to be the challenge and the giveaway. If the system is not advanced enough, the clouds will appear to sort of 'dart' along the sky in short little hops, 'cause the processor can't look ahead and redraw the advancing frames in smooth enough manner. If they're moving slow enough to begin with, possibly, but remember it's also busy redrawing the two people talking on the rock, so the portion of the frame with the most action/change is going to receive top priority.

[Oliver Chesler]

I just had to learn about Quicktime codecs. I perform with Ableton Live and I have the sequencer also playing a video which my Mac dual screens out onto an LCD projector. I realized that some codecs like H.264 use up to 45% of your CPU to decode on the fly. For web viewing this is fine but not when you have a few audio channels and a mic channel with plug-ins also running.

Check out my blog post to see my final codec decision and some other surprising things I discovered:
http://www.wiretotheear.com/2007/10/25/the-best-codec-for-video-in-ableton-live-on-a-mac/

[Rudolf von Hagenwil]

dumb question: What parameters are actually being "compressed' during video compression?

 

 

Overview of the MPEG-4 Standard:

http://www.chiariglione.org/mpeg/standards/mpeg-4/mpeg-4.htm

 

MPEG-4 compression is the standard of today in digital television. Before MPEG-2 was the official standard for digital TV compression. Using MPEG-4 compression provides better definition of quickly-moving. MPEG-4 is the global multimedia standard, delivering professional-quality audio and video streams over a wide range of bandwidths, from cell phone to broadband and beyond.

 

When you watch a H.264 Quicktime on a website, that also MPEG-4:

http://www.apple.com/quicktime/technologies/mpeg4/

 

This is a HDV (1920x1080i) 12-bit 2K video downscaled to 720p H.264:

http://timeline.ch/Showreel.html

[Rudolf von Hagenwil]

This is a screenshot, left Standard Definition (SD), right High Definition (HDTV)

[alphajerk]

But what if episode IV came out this year, in all it's digital-perfection glory? I don't know... I just really like the "feel" of those analog effects. Even though they look less realistic, there is (for me) a certain X-factor to episodes 4-6 -- As if it's almost more believable, because I don't automatically think, "Oh, that beam of light coming from that guy's gun is just a computer file". And I totally love Corpse Bride.

for me, especially if you watch them all in order [which my son has made me do] the old shcool analog approach comes across as cheesy and dated. you can see the black box masking around the tie fighters in space for instead against the starry backgroup, or that scene with obi wan shutting down the tractor beam. or the part where the emporer falls to his demise.

i dunno. i think compositing today really kills what was capable before... that isnt to say the STORIES are any better but it really is amazing how much can be done to enhance whatever story they have created.

of course there are always movies like magnolia which never seeings compositing really and heavily relies on its editing style, the work of the DP, the music and backing tracks, the intensity of the story.


anyway, film works the way the human eye does. better detail in black, less so in whites. digital works the opposite really. its resolution is black is nothing short of terrible. however, there are a LOT of things that most new digital guys miss, which is lighting a scene properly, knowing about DOF and how to use it to your advantage to lead the viewer through a scene. the lack of use, or improper use of dollies. its not just about slowing things down to 24fps. its not about adding grain. like everything, its really about skill and technique and using the medium to your benefit rather than its downsides run/ruin your work.

[alphajerk]

I just had to learn about Quicktime codecs. I perform with Ableton Live and I have the sequencer also playing a video which my Mac dual screens out onto an LCD projector. I realized that some codecs like H.264 use up to 45% of your CPU to decode on the fly. For web viewing this is fine but not when you have a few audio channels and a mic channel with plug-ins also running. Check out my blog post to see my final codec decision and some other surprising things I discovered: http://www.wiretotheear.com/2007/10/25/the-best-codec-for-video-in-ableton-live-on-a-mac/

 

 

uncompressed, while larger doesnt take the cpu cycles to decode the compression. and sometimes its better to just get a 3rd party card like balckmagix or aja to output your video and do the muscle.

[rasputin1963]

....and when I was growing up, my Catholic friends used to attend a catechism course after school which they simply called "CCD". I don't know what the initials stand for.....(?)

 

 

 

* Cleidocranial dysostosis, a genetic abnormality in humans * Central Core Disease, a rare neuromuscular disorder * Cortical collecting duct in the collecting duct system of the kidney

 

[Rudolf von Hagenwil]

.... some codecs like H.264 use up to 45% of your CPU to decode on the fly.

 

 

45% is too high. You have either an older processor, slow hard drives, or both, or you play 4K video.

 

I just checked that on a old G5, the CPU usage is:

 

01% CPU usage for standard NTSC playing via graphiccard.

28% CPU usage playing full 2K HD 1920x1080 video.

[Jon Gnash]

uncompressed, while larger doesnt take the cpu cycles to decode the compression. and sometimes its better to just get a 3rd party card like balckmagix or aja to output your video and do the muscle.

Afaik none of the Blackmagic DeckLink cards have a hardware accelerated IDCT engine.

 

It

[Oliver Chesler]

Hi Angelo,

I did my tests on a Dual 2.5 G5 with 2GB ram. In general the 740x480 H.264 file was ranging between 39% - peaks around 45%.

The best codec I found was believe it or not was PHOTO-JPEG. It has a medium/small file size but the CPU never went over 21%.

If I played uncompressed files DV NTSC the activity monitor would should around 20%.

I'm definitely not a video expert but as I mentioned in my post several other people on the Creative Cow and Ableton forums recommended PHOTO-JPEG.

But keep in mind this is a recommendation for what I am trying to do... a live show with the video playing in Ableton, etc...

My new Macbook Pro 2.2 with a 7200RPM drive arrives on or about November 5... that's going to make my life a lot easier.

[Jon Gnash]

Another approach to compression is called GOP--group of pictures. It's a Sony technology;

 

I thought that GOPs were just part of the MP2 & MP4 standard.

[Jon Gnash]

The best codec I found was believe it or not was PHOTO-JPEG. It has a medium/small file size but the CPU never went over 21%. If I played uncompressed files DV NTSC the activity monitor would should around 20%. I'm definitely not a video expert but as I mentioned in my post several other people on the Creative Cow and Ableton forums recommended PHOTO-JPEG.

 

That

[Jon Gnash]

I did my tests on a Dual 2.5 G5 with 2GB ram. In general the 740x480 H.264 file was ranging between 39% - peaks around 45%.

The peaks of 45% are peaks. Don

[alphajerk]

Afaik none of the Blackmagic DeckLink cards have a hardware accelerated IDCT engine. It

[Rudolf von Hagenwil]

Hi Angelo, I did my tests on a Dual 2.5 G5 with 2GB ram. In general the 740x480 H.264 file was ranging between 39% - peaks around 45%. The best codec I found was believe it or not was PHOTO-JPEG. It has a medium/small file size but the CPU never went over 21%. If I played uncompressed files DV NTSC the activity monitor would should around 20%. My new Macbook Pro 2.2 with a 7200RPM drive arrives on or about November 5... that's going to make my life a lot easier.

 

 

You probably have an older ATA controller and hard drives. But actually your SATA or even the older ATA hard drive should be fast enough, but not fast enough to run the digital audio production software simultaneously. If you want to lower the CPU usage, then you need an iSCSI Ultra with least 2 U320 SCSI 10 000 rpm drives, or a SAN or NAS storage solution. Then the CPU usage for a 740x486 H.264 drops to 2-3 % CPU usage.

 

 

GOP is a technic used in MPEG video compression. The standard is 15 frames per GOP (group-of-pictures), but can be changed to any other frame numbers per group with an encoder, or the encoder sets the start frame automatically when demultiplexing the MPEG video. The main drawback when cutting MPEG video is, the cut can only be at the first frame of each group, respectively if you cut in between, the cut jumps automatically to the nearest first frame, at least that's how it is with in video applications

 

.

[Hard Truth]

Alphajerk wrote "there are a LOT of things that most new digital guys miss, which is lighting a scene properly, knowing about DOF and how to use it to your advantage to lead the viewer through a scene. the lack of use, or improper use of dollies. its not just about slowing things down to 24fps. its not about adding grain. like everything, its really about skill and technique and using the medium to your benefit rather than its downsides run/ruin your work."

This is very true. Experienced, skilled videographers, such as the shooters of primetime TV dramas, know how to create a quality look with those techniques acheiving results that most viewers would not notice is video rather than film.

Inexperienced people will use a "film look" plug-in to add grain and shoot at 24fps in a usually unsuccessful attempt to acheive a film look.

It is analogous to an inexperienced audio engineer insisting on recording to analog tape to get a "classic" recorded sound when they don't know how to properly use EQ, compression etc.

[coyote-1]

THAT is what I was talking about.

While there is a sort of singularity at work here -- there are a finite number of grains in a photograph (high resolution film has more than low)...

Unlike audio, where analog audio captues one continuous 'image', photography of any sort breaks an image into discrete tiny elements. Photography of any sort is more analogous to digital than to analog because of that IMO. I did not mean to imply that it IS digital though.

[alphajerk]

well photography is a whole other bag of worms. there are literally loads of enlargers that are now worthless because of digital printing. there is absolutely no reason anymore to bother with them. the advantages to scanning in a film negative, or shooting strict digital to begin with and then working on the image within a computer and then outputting to a hi-res inkjet is far superior it rendered the old way basically obsolete and the equipment worthless.

[coyote-1]

The pro business (magazines etc) has been digital for ages now as enhancement is de riguer in the fashion world. FAR easier to do in the digital domain.

well photography is a whole other bag of worms. there are literally loads of enlargers that are now worthless because of digital printing. there is absolutely no reason anymore to bother with them. the advantages to scanning in a film negative, or shooting strict digital to begin with and then working on the image within a computer and then outputting to a hi-res inkjet is far superior it rendered the old way basically obsolete and the equipment worthless.

 

[alphajerk]

its been that way for at least 10 years now... i remember doing stuff almost entirely digital at Discovery Channel for their print stuff longer than that ago.