editing 4K at home...

You should be able to get a $15K computer, $120K Avid DS Nitris and a 1PB DiskXtender for about $200K. I would think that you'd have to convert your REDCODE file to an uncompressed file to edit it. Maybe Avid will come out with a real time 4K editing system that supports REDCODE.
 
I'm all about the offline. Besides the obvious fact that it's faster, requires far less storage, and is more efficient, I don't see a point in doing 4K online if I don't have a monitor that can properly display the footage.

REDCINE really is the feather in RED's cap, I do believe.
 
Speaking very roughly, when a codec is wavelet based isn't is it possible, to dynamically drop frequency bands to ease the processor load and deliver a lower resolution. How would such a precedure affect RAW de-bayering? I assume you, prior to encoding split the pixels into three seprate color channels and then compress each one seperately.
 
Haakon,

I disagree. Even without a 4k display device (which is not stopping interest a the camera) working with online material is the only practical solution of any effects work, indie guys are more used to online from a DV background, and the online really takes less space and time as you don't need to render out proxies. Plus proxy workflows are more prone to conform errors. Our 4k dissolve was performed without any wavelet multi-res tricks, so with these tricks, 4k can be manipulated very quickly (showing you 512, 1k, 2k or 4k as needed.)

Regarding : "REDCINE really is the feather in RED's cap, I do believe." Without that feather they would be bald. :) REDCINE is good idea, but also a necessary one. When you develop a propriatory format you either have to convert it to standard (e.g. REDCINE) or build direct support in as many apps (e.g. CineForm RAW.) Of course there are issues with either approach. REDCINE should be able to convert to CineForm RAW, and CineForm RAW can be loaded into tools like After Effects (PC or Mac(soon)) and exported to anything else as needed. Plus neither Red or CineForm is taking one approach exclusively, CineForm has conversion tools just as I'm sure Red plans direct integration. REDCINE is a good first step, but it is not the only way people will edit the content.
 
Anders Holck said:
Speaking very roughly, when a codec is wavelet based isn't is it possible, to dynamically drop frequency bands to ease the processor load and deliver a lower resolution. How would such a precedure affect RAW de-bayering? I assume you, prior to encoding split the pixels into three seprate color channels and then compress each one seperately.
Click to expand...

Yep. You get the idea. It is possible to manipulate 4k like it is uncompressed SD.
 
I'd do edits at a proxy res (probably DVCProHD 720P) and comps from the original Red footage in Shake. You can also color correct in Shake.
 
Schmovies production said:
David,
are you developing a software vertical for red or there will be an upgrade in you top end products?
Click to expand...

Not sure yet. Most likely the will a feature bump on CineForm Prospect 2K (which is Prospect HD with the filmmaker pack) to support 4K.
 
Everybody's been talking about the post production workflow. Can we talk about the production workflow? Say we had a feature that shot 1 hour's worth of footage a day for 30 days at 4k raw or 2k raw. What are the ways of capturing footage that RED has thought about, Jim or Graeme? What is necessary to store this data and what is necessary on set? These are questions producers are going to ask and I want to be able to tell them when I'm promoting the RED camera. I know you haven't nailed down the red drive yet, but what are your initial thoughts about what workflow is feasable, using red drives as switchable mags and then dumping to another larger raid? It seems to be about 1.5 terrabites per hour at 4k. What would a production be able to back up 30 hours to? It sounds to me like thousands of dollars in RAID drives. Could you shed some light on this? Thanks.
 
Red is currently promoting their Recode Raw workflow.
You use a small 160 GB harddrive magasine to capture Redcode Raw at 27 MB/s. That's around 98 mins of 4k 24 fps compressed raw media.
The magazine drive will have SATA, firewire 800, 400, usb-2 so you can offload to a raid when it's full.
30 hours is around 3 TB.
 
So, RED isn't promoting a way to capture 4k raw, 2k raw or even 4:4:4 1080p? It's only making capture devices for a redcode raw workflow? That doesn't sound right.
 
SF Geek said:
Everybody's been talking about the post production workflow. Can we talk about the production workflow? Say we had a feature that shot 1 hour's worth of footage a day for 30 days at 4k raw or 2k raw. What are the ways of capturing footage that RED has thought about, Jim or Graeme? What is necessary to store this data and what is necessary on set? These are questions producers are going to ask and I want to be able to tell them when I'm promoting the RED camera. I know you haven't nailed down the red drive yet, but what are your initial thoughts about what workflow is feasable, using red drives as switchable mags and then dumping to another larger raid? It seems to be about 1.5 terrabites per hour at 4k. What would a production be able to back up 30 hours to? It sounds to me like thousands of dollars in RAID drives. Could you shed some light on this? Thanks.
Click to expand...

I have to say that I am a tad curious as to why you would want to shoot RAW when you can shoot Redcode RAW. Is it for quality reasons? Because as far as what Greame has told us, Redcode Raw and RAW are pretty much exactly the same unless you intend to do pretty nasty things to the footage.

Off the top of my head I cannot think of any type of usage that you could do in RAW which you cannot do with redcode RAW, bar scientific, medical and astronomy applications.
 
RED has stated many image sizes that you'll be able to capture to in the RGB REDCODE codec and 4k uncompressed RAW. But the most flexible format at this point is the compressed RAW format which is 4k REDCODE RAW.
There will also be a highspeed data port which will supply uncompressed RAW data, but this demands a very highspeed RAID, and if the REDCODE RAW codec will be as good as we all hope, there is no need to go uncompresssed in most cases.

The beauty of REDCODE RAW is that you use the full sensor size, get full post tweakability (it's "raw" sensor data) at 10 bit log and maintains a manageable datasize of 27 MB/s at 24 fps.
 
Last edited:
OK, I'm not as up on codecs as I'm sure a lot of people are so could you explain some things to me. Is there color compression happening? How is it that wavelet compression is regarded as lossless and that other compression isn't. All I know is that compression is bad and if I want to push and tweak my looks to the limit like I would with film, then I would want RAW footage. I'm sure there are books that could explain all of this in lengthy detail, but could you give me short story on the whole thing.
 
Current specs:
REDCODE RAW is 4k 10 bit log RAW data from the sensor with VBR wavelet compression. There is no normal color subsampling as this would ruin the bayer encoding, but if you look at the bayer pattern of the sensor itself it works a bit like color subsampling, being that in the 4k frame, G has 50% of the pixels, Red and blue 25% each.
When you convert to RGB by de bayering, G, R and B each as 100%, which equals 3 times as many pixels.
The format is lossy, but as the efficiency is very high it's almost visually lossless.
The current aim is for this format to work at up to 30 fps.
 
Last edited:
Anders is right!

Wavelets are quite nice for compression. They don't add nasty noise like DCT does, and indeed, the only artifact you'll see, and only if you really pull the footage so far it's destroyed, is some mild noise reduction / softening.

The data is still raw, so it's still able to be pushed and pulled around to a great extent. That is not altered by the compression. The prototype REDCINE now handles REDCODE RAW files, and you'd never know you're working with one untill you look at the file and see it's 1.1MB rather than 24MB (working with 16 bit data at the moment). And that's in the RAW converter, pushing , pulling and tweaking.
 
At 27MBytes/s for 4k, no codec is mathemathically lossless. However my company has been spreading the message for years that wavelets can be visually lossless. Visually lossless is a concept that takes the human visual system into the equation. If you can reduce the data size in way that the eye doesn't notice, you have acheived higher compression without nasty artifacts. One reason wavelets are attributed with this characteristic more so than other compressors which are DCT based, is wavelet doesn't have block boundaries. DCT compression divides the image in 8x8 regions, when artifacts occur they have edges (squared off by the 8x8 grid), and the human eye is a really good edge detector. This oversimplified reason why a DCT image will look worse than an wavelet image at the same bit-rate. There are many more reasons that make wavelets a better real-world compressor.
 
David Newman said:
Vladimir,

Based on Red announcements, CineForm RAW has a similar bit-rate to REDCODE RAW (both 10-bit wavelet), so if you use an average of 27MBytes/s for 24p, RAID-0 with two 500GB drives will hold 10 hours of 4k RAW.
Click to expand...

Thank you David.
 
Thanks for the info. So if I was shooting redcode raw would I only be able to shoot 24fps? Anders said something about the goal of this format reaching 30fps. Does that mean I could shoot 2k at a higher framerate, or am I alltogether wrong?
 
You must log in or register to reply here.
Back
Top