Somewhat dumb question about Cineform at DVinfo.net

Jim Andrada · September 29th, 2007, 09:30 PM

I've been looking at the Cineform website and as I understand it (maybe wrongly!) I can capture over firewire and go to Cineform 4,2,2.

But I thought that the data stream from the camera was in 4,2,0.

So I'm confused about how one goes from 4,2,0 to 2,2,2. Is there sufficient information remaining in the 4,2,0 stream?

I also see that cameras will upres from 4,2,0 to 4,2,2 so I guess I'm equally confused about how this is possible and how much if any different it is from native 4,2,2.

Marc Salvatore · September 29th, 2007, 10:29 PM

Jim,

My understanding is that it upconverts/trancodes the footage to an AVI in the 4:2:2 colorspace. This does not add quality on capture but will give you better results when you start color correcting, compositing etc.. This is because you're now rendering your effects in a higher quality color space so they are less likely to break down like a DV or HDV codec would when pushed to it's limits. I should note that you also go from about 12 gb (HDV) to 38 gb (Cineform) per hour of footage.

You will want to use the "High" or "Film Scan" setting when capturing to Cineform for best results. I've heard on this forum that "Med" or lower will look slightly soft.

Regards,

Marc

Jim Andrada · September 29th, 2007, 11:00 PM

Marc,

Thanks for the comeback.

So if I understand correctly, there is no improvement over what was already captured, but there should be less "generational loss" down the road because any additional changes etc to the video will be done in a higher quality color space.

In other words, the 4,2,0 space is in some way a subset of the 4,2,2 space so the original video just sits there like it was in the beginning with no additional information being inferred or interpolated and addtional edits have more - shall we say "headroom" than they would have had if left in 4,2,0.

Hernan Vilchez · September 30th, 2007, 06:00 AM

The same concept would apply to the 10bit of Prospect vs the 8bit of Aspect?

David Newman · September 30th, 2007, 10:01 AM

4:2:0 to 4:2:2 (and 4:4:4) upconverting filters are required for NLE, compositor and 4:2:2 display technology. While there is no new chroma information, there is a visual quality improvement for doing filtered 4:2:0 to 4:2:2. see http://www.cineform.com/technology/H...ysis051011.htm We could perform this opteration on decode, but instead we do it before we encode into CineForm. Do the upconvert before compression, improves the chroma channels quality through compression.

The same goes for 8-bit up conversion to 10-bit. While there is no new information in the upconverted 10-bit, it goes through the compression with improved precision for downstream reconstruction, greatly reducing the banding often associated with compressed 8-bit.

David Taylor · September 30th, 2007, 11:07 AM

Pardon me for the very simply metaphor but it is somewhat descriptive....

When you shoot HDV which is highly compressed at 8 bits and 4:2:0 color space, you have little ability to "move" the image in post. Think of it like filling a bucket to the top with water, and then somebody telling you to carry the bucket 100 yards up a rock-strewn hill without spilling a drop.

When you use CineForm, before you "move the bucket up the hill", we give you a much bigger bucket - about 3X bigger in the case of HDV. So you dump your original full-to-the-top bucket ("HDV") into our 3X larger bucket ("CineForm Intermediate") then carry it up the hill. Because the bucket is so large you haven't lost any water.

The association here is that our files are larger, but they create much more headroom to allow you to "move" your images in post in a similar manner to a larger bucket making it easier to move water up the rock-strewn hill without spilling a drop.

Cary Lee · September 30th, 2007, 04:05 PM

I just upgraded from NEO HDV to NEO HD and I was wondering since I shoot HDV can I up my output to 1920x1080 vs the HDV 1440x1080 and work in a HD timeline in my NLE (vegas 8) and output it out as HD 1920 on a blu-ray burner?

David Newman · September 30th, 2007, 04:35 PM

Cary,

Yes to can upconvert to 1920 when capturing 1440 HDV content.

Brandon Freeman · October 3rd, 2007, 12:36 PM

Question --

What do you recommend, regarding the colorspace issue and color correcting? Currently, I capture .m2t files, color correct and then render out with CineForm. Am I to understand, based on this thread, that it is better to FIRST render out a CineForm .avi BEFORE any color correction?

Thanks.

David Newman · October 3rd, 2007, 01:08 PM

It is a tricky subject. In an ideal world an 8-bit sources could be decoded to a deeper format (16-bit integer or 32-bit float), color corrected and sent to a deeper format like CineForm to preserve the quality (and it work fine that way in AE.) However, when render from non-CineForm files with Premiere we can't assume it can handle 32-bit float, so we only request 8-bit RGB. If you convert first, this problem goes away, as we can decode to 32-bit float, allowing for much nicer color correction.

Ray Bell · October 3rd, 2007, 07:40 PM

David when you say " convert first " are you talking about converting in
AE first and then bringing the converted footage into Premier??

David Newman · October 3rd, 2007, 07:51 PM

Quote:

Originally Posted by Ray Bell

David when you say " convert first " are you talking about converting in
AE first and then bringing the converted footage into Premier??

What are talking about, I've lost track? My last answer was regarding converting M2T data to CineForm before your edit/CC.

Jim Andrada · October 3rd, 2007, 07:56 PM

The way I read the answer was that in AE you'd get great results by taking native source in and color correcting before outputting to Cineform, but this wouldn't always be the case for other applications, like Premiere, so in general it would be best to go to Cineform right away.

Of course, this was just my interpretation, so I'll wait for the definitive answer.

David Newman · October 3rd, 2007, 08:09 PM

Quote:

Originally Posted by Jim Andrada

The way I read the answer was that in AE you'd get great results by taking native source in and color correcting before outputting to Cineform, but this wouldn't always be the case for other applications, like Premiere, so in general it would be best to go to Cineform right away.

Of course, this was just my interpretation, so I'll wait for the definitive answer.

Yes. Try working native M2T in AE and you find the other reason to convert to CineForm first -- speed.

September 29th, 2007, 09:30 PM	#1
Jim Andrada Inner Circle Join Date: Feb 2007 Location: Tucson AZ Posts: 2,211	Somewhat dumb question about Cineform I've been looking at the Cineform website and as I understand it (maybe wrongly!) I can capture over firewire and go to Cineform 4,2,2. But I thought that the data stream from the camera was in 4,2,0. So I'm confused about how one goes from 4,2,0 to 2,2,2. Is there sufficient information remaining in the 4,2,0 stream? I also see that cameras will upres from 4,2,0 to 4,2,2 so I guess I'm equally confused about how this is possible and how much if any different it is from native 4,2,2.

September 30th, 2007, 10:01 AM	#5
David Newman CTO, CineForm Inc. Join Date: Jul 2003 Location: Cardiff-by-the-Sea, California Posts: 8,095	4:2:0 to 4:2:2 (and 4:4:4) upconverting filters are required for NLE, compositor and 4:2:2 display technology. While there is no new chroma information, there is a visual quality improvement for doing filtered 4:2:0 to 4:2:2. see http://www.cineform.com/technology/H...ysis051011.htm We could perform this opteration on decode, but instead we do it before we encode into CineForm. Do the upconvert before compression, improves the chroma channels quality through compression. The same goes for 8-bit up conversion to 10-bit. While there is no new information in the upconverted 10-bit, it goes through the compression with improved precision for downstream reconstruction, greatly reducing the banding often associated with compressed 8-bit. __________________ David Newman -- web: www.gopro.com blog: cineform.blogspot.com -- twitter: twitter.com/David_Newman

September 30th, 2007, 04:05 PM	#7
Cary Lee Regular Crew Join Date: Mar 2004 Location: Hollywood USA Posts: 128	I just upgraded from NEO HDV to NEO HD and I was wondering since I shoot HDV can I up my output to 1920x1080 vs the HDV 1440x1080 and work in a HD timeline in my NLE (vegas 8) and output it out as HD 1920 on a blu-ray burner? __________________ Canon XHA1, SGpro,Flip,FF, RR Mattebox, Nebtek V-R70p-HDA with Canon, Nikkor Primes 24mm f2.8, 28mm f2.8, 35mm f2.0, 50mm f1.4, 85mm f1.4, 105mm f1.8, 135mm f2.0, and 300mm f4.0.

September 30th, 2007, 04:35 PM	#8
David Newman CTO, CineForm Inc. Join Date: Jul 2003 Location: Cardiff-by-the-Sea, California Posts: 8,095	Cary, Yes to can upconvert to 1920 when capturing 1440 HDV content. __________________ David Newman -- web: www.gopro.com blog: cineform.blogspot.com -- twitter: twitter.com/David_Newman

October 3rd, 2007, 01:08 PM	#10
David Newman CTO, CineForm Inc. Join Date: Jul 2003 Location: Cardiff-by-the-Sea, California Posts: 8,095	It is a tricky subject. In an ideal world an 8-bit sources could be decoded to a deeper format (16-bit integer or 32-bit float), color corrected and sent to a deeper format like CineForm to preserve the quality (and it work fine that way in AE.) However, when render from non-CineForm files with Premiere we can't assume it can handle 32-bit float, so we only request 8-bit RGB. If you convert first, this problem goes away, as we can decode to 32-bit float, allowing for much nicer color correction. __________________ David Newman -- web: www.gopro.com blog: cineform.blogspot.com -- twitter: twitter.com/David_Newman

September 29th, 2007, 10:29 PM	#2
Marc Salvatore Major Player Join Date: Jun 2007 Location: Medford, OR Posts: 351	Jim, My understanding is that it upconverts/trancodes the footage to an AVI in the 4:2:2 colorspace. This does not add quality on capture but will give you better results when you start color correcting, compositing etc.. This is because you're now rendering your effects in a higher quality color space so they are less likely to break down like a DV or HDV codec would when pushed to it's limits. I should note that you also go from about 12 gb (HDV) to 38 gb (Cineform) per hour of footage. You will want to use the "High" or "Film Scan" setting when capturing to Cineform for best results. I've heard on this forum that "Med" or lower will look slightly soft. Regards, Marc

September 29th, 2007, 11:00 PM	#3
Jim Andrada Inner Circle Join Date: Feb 2007 Location: Tucson AZ Posts: 2,211	Marc, Thanks for the comeback. So if I understand correctly, there is no improvement over what was already captured, but there should be less "generational loss" down the road because any additional changes etc to the video will be done in a higher quality color space. In other words, the 4,2,0 space is in some way a subset of the 4,2,2 space so the original video just sits there like it was in the beginning with no additional information being inferred or interpolated and addtional edits have more - shall we say "headroom" than they would have had if left in 4,2,0.

September 30th, 2007, 06:00 AM	#4
Hernan Vilchez Regular Crew Join Date: Feb 2007 Location: Munich, Germany Posts: 135	The same concept would apply to the 10bit of Prospect vs the 8bit of Aspect?

September 30th, 2007, 11:07 AM	#6
David Taylor Major Player Join Date: Sep 2003 Location: Solana Beach, CA Posts: 853	Pardon me for the very simply metaphor but it is somewhat descriptive.... When you shoot HDV which is highly compressed at 8 bits and 4:2:0 color space, you have little ability to "move" the image in post. Think of it like filling a bucket to the top with water, and then somebody telling you to carry the bucket 100 yards up a rock-strewn hill without spilling a drop. When you use CineForm, before you "move the bucket up the hill", we give you a much bigger bucket - about 3X bigger in the case of HDV. So you dump your original full-to-the-top bucket ("HDV") into our 3X larger bucket ("CineForm Intermediate") then carry it up the hill. Because the bucket is so large you haven't lost any water. The association here is that our files are larger, but they create much more headroom to allow you to "move" your images in post in a similar manner to a larger bucket making it easier to move water up the rock-strewn hill without spilling a drop.

October 3rd, 2007, 12:36 PM	#9
Brandon Freeman Major Player Join Date: Feb 2007 Location: Boise, ID Posts: 226	Question -- What do you recommend, regarding the colorspace issue and color correcting? Currently, I capture .m2t files, color correct and then render out with CineForm. Am I to understand, based on this thread, that it is better to FIRST render out a CineForm .avi BEFORE any color correction? Thanks.

October 3rd, 2007, 07:40 PM	#11
Ray Bell Regular Crew Join Date: Nov 2004 Posts: 1,414	David when you say " convert first " are you talking about converting in AE first and then bringing the converted footage into Premier??

October 3rd, 2007, 07:56 PM	#13
Jim Andrada Inner Circle Join Date: Feb 2007 Location: Tucson AZ Posts: 2,211	The way I read the answer was that in AE you'd get great results by taking native source in and color correcting before outputting to Cineform, but this wouldn't always be the case for other applications, like Premiere, so in general it would be best to go to Cineform right away. Of course, this was just my interpretation, so I'll wait for the definitive answer.