I am setting up a stock footage business, with everything shot on DV, I want to colour correct in CF, but I am unsure how I get from DV to CF, is it best to create a 4.2.2. uncompressed version of the native DV, or is CF capable of working with DV native files. I ask this question to see what I will need in the way of hardware.
Color Finesse will work with any format of video that After Effects or FCP can read.
For your application, I'd recommend bringing your DV footage into a 16-bit After Effects project, color correcting with Color Finesse, then rendering out of AE into whatever format you intend to deliver your stock footage. Most of the CD/DVD-ROM based stock footage seems to use JPEG at 90-95% quality.
Despite starting with compressed, 8-bit, 4:1:1 DV footage, you want to do your correction in as large a color space as possible. You can then deliver in the format you choose. The quality of this will be superior to doing all your correction in an 8-bit environment, such as FCP. Even if you choose DV as your final format (which I wouldn't recommend), the results of having a deeper intermediate stage will be visible.
Thanks Bob, so a couple of things come to my mind, first of all I am not yet familiar with AE, what is meant by a 16bit project, would decompressing the native DV footage into SDI/4.2.2. (either via the DV playback deck or software) give the necessary colour space you talk about? By the way, I do understand the need to do any colour correction in a large colour space.
Then I am not certain about your comments on the delivery format, I have this understanding, from reading articles on Ken Stone's site, that all this decompression and recompression really messes up the the final image. What I was figuring was that suppplying the footage as a DV file, would be the least damaging visually. But please feel free to reducate me!!
Thanks again, If we can get the process understood at the theory stage then the practicalities become simple for us.
DV is compressed, 8-bit, 4:1:1 video. Since it's already compressed and color subsampled, there's no way to repair whatever "damage" that might have done to your image. The goal is to avoid any further damage.
When you load that DV footage into After Effects, it gets converted from 4:1:1 Y'CbCr data into 4:4:4 RGB data as the DV codec decompresses the data into After Effects' RGB buffers. So there is no need to do any intermediate step to expand the color space.
When you're doing the color correction within AE, you want a color space with as much resolution as possible, so that as you do minute adjustments to the data you minimze rounding error. That's why I suggested using a 16-bit project (16 bits for each of the red, green and blue channels). When working in an 8-bit project, if you use multiple filters, each filter needs to round to an 8-bit value, regardless of how accurate the filter is internally. By using a 16-bit project, you reduce that rounding error. As you apply multiple filters, this rounding can become quite visible if you're limited to 8 bits.
Color Finesse takes this a step further by using 32-bit floating point to represent each color channel, and by combining most of the color correction tools you'll need to use into one application so that there isn't the need to combine multiple filters with rounding between each filter.
So now that you have the beautifully corrected footage in a big 16-bit space, the trick is to deliver it to someone else. If you deliver in DV format, then you'll need to go through another DV compression cycle. And the person using your footage will also need to go through a DV decompression cycle, unless they happen to be using your footage unchanged and are using a DV system.
By using a very lightly compressed delivery method (such as JPEG @ 90%), you eliminate that additional DV compression and color subsampling step. But it really depends on how you want to deliver footage, your business model, etc., any of which may outweigh the quality issue.
Thank you, thank you, seems you have 'solved' the problem, but FWIW what kind of data rate would I expect by jpeg'ing this 16bit file. To put it into perspective, each clip will be 30secs, are we looking at 100's Mbs.
TIA, Rob Dunford
I think we will be getting a copy of CF soon to start experimenting. I have years of experience on Rank Cintels/Pogle/RSQ sec colour corr. telecine, so I am looking forward to playing with an all software solution for the hardware solution.
I just did a really simple test and ended up with around 100 kB per frame (NTSC, 720x480) or around 3 MB/sec, which is a bit less than what DV is.
Depending on your footage, you might want to choose better quality compress than I did, but you really shouldn't see a huge increase in file size.
Ok just ONE more question, to handle this large file in CF, do I need to install a fast SCSI drive to avoid dropping frames, or is a fast dedicated IDE drive suitable. Bearing in mind, we will be handling up to a 1000 separate clips in the setting up of this library and every little speed increase is welcome. My proposed setup was going to look like this maybe, new DP Mac, 4 large IDE drives in the base of the case on a separate controller, boot drive in lower optical bay and finally maybe a fast SCSI external drive for CF/AE.
In your opinion, would it 'work' if the boot drive was the fast SCSI and doubled up for the CF/AE files, obviuosly I would use the biggest/ fastest SCSI drive avail.