Fascinating Thread on Reduser

[Keith Walters]

WHY!!!!-are-people-still-using-color-baked-DPX-s-F-amp-*-*-K!

 

I don't think I've ever seen a thread on Reduser with so much solid (and frank!) "signal", and so little noise! :rolleyes:

 

It's sort of comical that the original poster asks "Why?! ..." and a whole lot of industry experienced and technically literate members proceed to tell him why.

 

Probably nothing that hasn't been said before, here or elsewhere, but nicely condensed. And not a single one-line sycophant post (at least not yet :D )

 

Wonder if it'll stay up. You better get in quick....

[Phil Rhodes]

I find myself with a mixed reaction to that.

 

First, I've long held (and expressed) the view that the higher-end a post house considers itself to be, the more likely it is to do catastrophically stupid things like transcoding red to DPX, baking in a look, and then trying to grade the result. I'm stretching my brain trying to think of a reason as to why this might have happened, but since there isn't really even a time penalty for doing it right it can only really be a lack of competence - or at least, I'd love to hear the excuse. Post houses seem to become "high end" by doing work on big, visible projects, which seems to get them stuck in a rut of doing things in exactly one way. There is a terrible tendency to assume, or at least robustly try to insist, that their way is the only way, and they are often simply wrong, or at least trying to justify their investment in equipment that's been obsoleted by massive advances in information technology. Specialists at the top of a wide variety of disciplines often get terrible tunnel vision in this way, and television postproduction is no exception. People who work on smaller, independent stuff where they can't dictate the workflow, where they may be approached with various types of media and asked to perform a wide variety of tasks, often have a much broader base of experience and are often much more capable.

 

In short, posting a big film or TV series is not really evidence of problem-solving capability, it's evidence of copying exactly the process used by dozens of other places, repeatedly, in circumstances where huge amounts of money are available.

 

The contrary issue is that which has always been the controversy with Red (and now, sadly, other camera systems): they built half a camera, and let post do the rest of the work. It is not reasonable to expect every post house to equip specially to handle the output of every camera, and while it may be reasonable to expect that most places would cater to something as popular as Red, that does set a dangerous precedent. Red's output is not only hard work to use, they have also been wilfully difficult by encrypting the original material and ensuring that it is impossible for third parties to work with directly. If we're going to keep building Bayer cameras that output the mosaic, compressed or otherwise, standardisation is imperative.

 

Expecting everyone to read format-o-the-week (be that DNxHD, Prores, Red, whatever) is extremely daft. I don't think DPX sequence has much to recommend it; I think it's often used because it's the closest thing to a standard we have.

 

Whether that's a good thing or not is a discussion for another time.

 

P

 

PS - What a "Red Rocket" does is clearly parallel processing - has anyone ripped the heatsink off one to see what it really is? Presumably some sort of big FPGA, but it's exactly the sort of thing you'd expect put into CUDA, really.

[Keith Walters]

 

The contrary issue is that which has always been the controversy with Red (and now, sadly, other camera systems): they built half a camera, and let post do the rest of the work. It is not reasonable to expect every post house to equip specially to handle the output of every camera, and while it may be reasonable to expect that most places would cater to something as popular as Red, that does set a dangerous precedent. Red's output is not only hard work to use, they have also been wilfully difficult by encrypting the original material and ensuring that it is impossible for third parties to work with directly. If we're going to keep building Bayer cameras that output the mosaic, compressed or otherwise, standardisation is imperative.

 

Expecting everyone to read format-o-the-week (be that DNxHD, Prores, Red, whatever) is extremely daft. I don't think DPX sequence has much to recommend it; I think it's often used because it's the closest thing to a standard we have.

They could put all these arguments to rest by supplying a optional module which simply clips onto the "brain" and offers all the industry standard formats that people keep converting the RAW files into anyway. But I suppose that would be tantamount to admitting they were w-r-o-n-g....

 

All this reminds me of the big "M2" push that Matsushita wasted so much money on the 1980s. They seemed totally convinced that it was only a matter of "education" to get the major players to virtually bulldoze their entire multi-million dollar post-production infrastructures onto the loading dock and replace everything with component video (from Matsushita of course, not that hoik-splat! Sony rubbish! Actually, I think Graeme Nattress is about the only person in the RED camp who even knows what M2 is now :rolleyes:)

 

And certain self-styled "leading lights" really need to pull their f*cking heads in; you are not doing the Red cause one bit of good.

 

I'd be tempted to add: "You know who you are..." but sadly, I fear they don't :lol:

 

Anyway, gotta go; It's about time for me to cry over what the RED might have been again....

[Paul Bartok]

I get it there's some problems with post houses especially Australia but his end theory is Alexa + Canon user are all dumb, off course i forgot there not RED cameras so they must be inferior, how silly of me to forget. RED USERS are all conformist elitists. look at me my camera it up scales to RAW. get over yourselves.

[Teddy Smith]

I shot a commercial on Red and had it graded in Davinci by a big post house in town with several very large film credits under it's name. The colorist didn't understand what I was asking when I told him to make the adjustments in "raw". I had to look it up on Google myself and show him how to change the color temp and other things in a raw menu inside of the Davinci itself. He said "we don't do it this way". I told him I didn't care how he normally does it, I wanted the major color adjustments done in raw then we can make tweaks his way.

[Phil Rhodes]

I'm not surprised, colorists can be absolutely terrible like that.

[Jon Rosenbloom]

Are we not splitting hairs? How wrong do you have to shoot it for RedRaw vs. DPX to be an issue? I just graded a Red-shot short at a "big name" post house here in NY. Indeed, they ingested all the raw footage, rebuilt the timeline as DPX files, and we graded from there. The colorist mentioned that he would have to make sure the engineers would not reflexively put a LUT on the footage before we even looked at it. But, that took just a word or two. We "baked in" the look when we were done. Would I have been happier had we not worked from the DPX conversion? I don't think so. I had spent the past few months worrying I had underexposed the project, but we had all the latitude we could ask for, and in the end, we graded it pretty much back to how we lit it - without a DIT, or even a 17" - on the day. It looks just fine. Go figure.

[David Mullen ASC]

In an ideal world, we wouldn't be forcing linear digital information into a film log workflow, but we don't live in an ideal world. In reality, 95% of the time, a conversion to Log DPX files should contain all the information in the original RAW image without much color-correction problems unless totally wrong parameters were chosen for the conversion (and if they were, you'd just re-convert it correctly.)

 

But eventually we should all be getting away from working in 10-bit Log DPX for digitally-acquired material, which is one of the reasons Red should support the ACES project. The Sony F65 short "El Dorado" at NAB was color-corrected using ACES in 16-bit linear and looked great, for a variety of reasons.

 

But as Mike Most has pointed out, one of the reasons for the use of a common 10-bit Log DPX system at post houses is that many projects come in with material shot on a variety of cameras, so it helps if all the footage is converted to a common format before color-correction begins.

[Bruce Greene]

I recently graded a movie shot with RED. I tested grading from the RAW file vs. grading from 2k.dpx.

 

I also tested REDlogfilm vs. REDgamma2 for the .dpx.

 

My observation was that it doesn't much matter which method one uses as long as one doesn't clip any data in the transcode to .dpx Unless the white balance is very far off in the RAW converter, it's very easy to adjust color temperature in the transcoded .dpx files.

 

The other observation is that there is no advantage to using REDlogfilm. In fact, the REDlogfilm transcode mapped the dynamic range of the camera into a narrow portion of the luminance range. Using REDgamma2 filled up most of that range with data, with no clipping if done at the correct ISO setting in the RAW converter.

 

In the end, the super highlights were saved and rendered into a log curve for filming out the negative, while they are clipped a bit for the digital delivery, as they always will be.

[F Bulgarelli]

David

 

Could you please explain "linear digital information into a film log workflow"

 

Thanks

[David Mullen ASC]

David

 

Could you please explain "linear digital information into a film log workflow"

 

Thanks

 

The sensor response to light is linear -- log is an artificial gamma curve applied after the fact, mainly because most D.I. systems are used to dealing with 10-bit log scans of film, plus anything recorded to film has to be log for making prints, and the other 10-bit video standard, Rec.709, has limited dynamic range compared to Log. But nowadays it is possible to work at higher bit depth in linear, like 12-bit, 14-bit, or 16-bit linear RGB so why deal with a log conversion, especially as recording to film becomes less necessary (and a conversion to 10-bit log could be done after the color-correction is done.)

[F Bulgarelli]

Thanks David