Cinema camera or Video camera?

[Jannard]

There are a few ways to make your RED ONE look like it has 7 stops of dynamic range... two primary ways.

1. Over-expose. Just clip the highlights while you record and there won't be any need to screw up anything in grading. This is the easiest way to turn a Digital Cinema Camera into an ordinary looking video camera (although you will still have more resolution).

2. Grade it improperly. Use ISO or exposure to set the mid-grey (or make most of the image look right) in RED Alert!, REDCINE, Scratch or Final Cut... without any regard for the histogram. Let the highlights go because the "middle of the image looks right". Don't use a curve to pull up the middle because that would hold the highlight you wish to blow out. Clipping is an art. If you can get really good at it, you may also learn what not to do "just in case you want the full info for a specialty shot".

As has been pointed out, the RED ONE does not have as much DNR as film. While most people think it is plenty to work with... it certainly does not if you throw away info.

Here's some good news. We are here to help. Build 16 makes it MUCH more difficult to miss exposure. And REDspace does a great job of helping to see your image without tempting you to make grading errors.

There are some of you that are saying "I don't do that. I know exactly what to do and how to get the most DNR with my RED ONE". OK, then. I'm not talking to you.

Here is an example of how to get an image to clip with one grading button.

I know... this grab is not that sharp. That isn't the point here...

Jim

 

[conrad gaunt]

Is there ever any chance of getting the full 12bit sensor data out of redcine in a DPX, I know 10bit log doesn`t lose much useful data, but I agree with you about not throwing it away?

 

[Graeme Nattress]

That's a great example Jim.

Graeme

 

[Graeme Nattress]

10bit Log or Gamma encode from a 12bit source does indeed loose you some code value precision. However, it does so from the brightest stops where you have an over-abundance of code values, and hence, if done correctly, does not diminish the image quality in any meaningful way.

Graeme

 

[10s]

Do you mean I have to actually use the camera properly? that's crazy! :blink:

I've thought for a long time there is a need for a special DVD on how to properly use the camera. It should go out with every unit.

 

[mrmuccy]

Jim,

Can you tell me about the wavelette compression you're doing & where it is introduced in the image chain? I'm wondering if this is connected to DNR? Although I imagine its after the AD conversion.

 

[I Bloom]

Jim,

Did you just bust out some build 16 preview frame grabs!

IBloom

 

[Jannard]

We are constantly amazed at how much footage we see being "roughed up" during grading. More than we see over-exposed. I think it is safe to say that much of the "RED footage looks like video" and "I see so many clipped highlights" are a result of grading. RED RAW footage is not like film, in an inverse way. With film, the tendency is to under expose a negative to keep the shadows from blocking. RED RAW footage is exactly the opposite... which I'm sure has contributed to some difficulties.

Build 16's primary goal is to help our customers make easier decisions that produce better results. It is a major change in how to shoot and grade RED footage.

Jim

 

[Jannard]

Jim,

Did you just bust out some build 16 preview frame grabs!

IBloom

Nope... old Build 12 or 13 footage.

Jim

 

[Graeme Nattress]

With RAW recording to R3D files, the process is simple - we take the RAW sensor data, and pipe it into the wavelet compression. It does not effect dynamic range at all, being 12bit as the sensor is 12bit.

Graeme

 

[Jim Exton]

A little instructional tape or clip put up on Red.com, like you did with the myth buster one, would be really helpful.

Glad to hear build 16 helps, but it sounds like there is still further instruction needed.

Hopefully you all or someone else will get to that pretty quickly once build 16 is out.

 

[Mark Crabtree]

I'm buying a preset lens. Preset iris. Preset focus. Fool proof!
It's what we've all been hoping and praying for, true point and shoot technology. No need to look at histograms or learn how to use a complicated digital readout light meter. Your choice of Pinhole or Bell & Howell optics. PM me for group buy info.

 

[I Bloom]

Jim,

Can you tell me about the wavelette compression you're doing & where it is introduced in the image chain? I'm wondering if this is connected to DNR? Although I imagine its after the AD conversion.

If I could attempt to answer.

AFAIK the compression is the last step before storage, after some digital signal processing and prior to debayering (which as we know happens in post on Red.)

Wavelet compression is not connected to DNR in the sense that you aren't loosing any bit depth, nor are you changing the values in the image signal.

However it does affect subjective DNR in the sense that random noise cannot be compressed. So your noise floor "looks different" when it has been compressed, the grain is not as tight. It's "clumpy". If you think this texture is unattractive you might be less apt to let your image fall into the noise floor. So your effective DNR is reduced.

Who knows how build 16 works, but here is a question: If noise can't be compressed, and any compression will necessarily change the character of the noise, than perhaps it makes sense to change the character of the noise first ourselves, prior to compression, so that it compresses in a more attractive way.

We all have our theories.

IBloom

 

[conrad gaunt]

10bit Log or Gamma encode from a 12bit source does indeed loose you some code value precision. However, it does so from the brightest stops where you have an over-abundance of code values, and hence, if done correctly, does not diminish the image quality in any meaningful way.

Graeme

Understood, but it would be nice to not have to do any kind of `first lite correction` before outputting from RC, I`m simply going to convert the log back into its non-log form so I can manipulate it with other material, so it seems like needless processing which won`t save any disk space in the end. Log may be useful for filmout people etc, but nobody I know.
I`d rather not rely on RC for anything other than decompressing R3Ds, thats not to say that I won`t want to use RC for other things at other times, when its working well. I just want access to the full sensor data that the camera promises. 12bits = max 12stops. 10bits = max 10stops. Like jim says, you can shift/crush/expand colours many ways, so information that seems irrelevant might not end up that way with correct processing. why not have a non-log option, or at least a 12bit log option so we can dig a little deeper.

edit: I should point out that only having 10bit log output won`t be why some people will struggle to grade and lose information as Jim is highlighting.

ps, Can`t wait to see build 16 footage

 

[Daniel Browning]

Jim, you forgot to share the second half of the lesson: how to crush the shadow detail and sharpen the image into oblivion.

 

[Matthew Verkler]

Jim, you forgot to share the second half of the lesson: how to crush the shadow detail and sharpen the image into oblivion.

Yes, over-sharpening is key to achieving that much sought after "video look". I have successfully turned boring Red shots into much superior "digital video" with the techniques Jim has outlined, coupled with copious amounts of sharpening, or, as I like to call it, "enhancement".

Thanks Jim, for helping everyone figure out how to match the footage we have become used to!

 

[Jannard]

Jim, you forgot to share the second half of the lesson: how to crush the shadow detail and sharpen the image into oblivion.

That's next week...

Actually, this clipped highlight thing is the single biggest issue facing RED ONE users. If it was all about the camera, that would be one thing. But so much footage is "reduced in quality" during grading that it deserves most of our attention. Again, it is the primary benefit of Build 16.

Jim

 

[10s]

This RED experience is a learning curve for everyone. On top of this, there are the skeptics, and then there's actual technological challenges making this even more complicated. Sometimes the cutting edge really hurts!

Maybe build 16 will solve many issues but it seems to me that the best way to get the word out and reset everyone's learning curve is a great video via DVD & WEB on how the whole system works. This will cut down on all the chaos, chatter and skeptics who are having a hey-day with poor operator execution, and blaming it on the hardware. Of course this takes time, money and it's another project/headache, but isn't it even a bigger headache to introduce new a photographic system and end up having so many get it wrong and spread rumors.

...I do like Mark Crabtree's idea on just settling this once and for all, & getting pin hole cameras! :wacko:

 

[Jason Ing]

Edit: I reread Jim's preview post and answered my own question. Never mind.

 

[Brandon Fraley]

With film, the tendency is to under expose a negative to keep the shadows from blocking. RED RAW footage is exactly the opposite... which I'm sure has contributed to some difficulties.

did you mean over expose? I'm not sure i understand.