REDAlert Settings

[Graeme Nattress]

DPX is lacking metadata at the moment while we figure it out. DPX is not the best documented format in the world.....

REDLog - yes, I'll come up with some docs. In the meantime, treat it as viper filmstream log (without funky green though) and you won't go ar wrong.

Graeme

 

[Brenton]

Thanks for the DPX T/C response Graeme

 

[Emery Wells]

But also maintain your love for the rich, saturated images that you may ultimately create from this raw material—and hope/beg/plead for tools to allow shooting with RED One under a non-destructive LUT that is included with the footage as metadata, so that you can preview your image as it may ultimately appear, record that nice flat raw image, and later have the choice of applying your shooting LUT or some other awesome color correction.

-Stu

Stu, this is close to how it is currently set-up. While you cant load LUTS per say, in the future one would be able to load a shot into Redcine, save that look as metadata, and then import it back into the camera. You currently have all the redcine controls inside the camera though they are not as easy to exploit as they would be when working in redcine.

 

[Brenton]

DPX extension not being added on export

DPX extension not being added on export

After successfully converting multiple shots to 2k log DPX, I am now encountering 'strange behaviour' from the export. Instead of naming the clips
[A001_C002_070904_001_00000.dpx] it's naming them as
[A001_C002_070904_001_000002007/09/05/09/12/15012] so it looks like it's adding the date and time in place of the DPX extension.

Here's my 'workflow'. Open the .R3D, make sure 'Red Log' is selected. Click 'Save Sequence 2k DPX', click on the .R3D original file and delete the .R3D from the name in the save dialogue and click save.

This has worked fine so far, the only difference I can see is that these shots were shot earlier in the day and potentially on a different CFcard.

The folder I was trying to work in had A001_C002_070904_001.r3d & A001_C002_070904_002.r3d. Trying a different shot in a folder that had only one .r3d still gave me the same extension problem.

I've just been kicked off the machine, but prior to that I was able to get one file *002 to export with the dpx extension by saving it to a new folder, but *001 wouldn't. Changing the name by removing excess 0's didn't make a difference, neither did giving it a name manually.

These are just camera tests and we have plenty to move on with but I did find this behaviour odd... just thought I'd pass it on

Cheers
B

[edit] Got back on the machine and now it won't output the files as dpx using the filename from the camera. Renaming it 'bob' however works as does saving it as tiff using the filename from camera.
Tiffs it is I spose as the DoP is coming in tomorrow and I need to get it into the Baselight.

 

[Joel Kaye]

Thanks Mike! MB Looks is a passion project for me, and it should be the ultimate complement to a RED One

Great to hear that. I'd love to see an article round tripping a RED offline edit via QT proxy through the new Looks for finishing.

LOOKS is going to be a monster application based on what I've seen so far.

 

[Stu Maschwitz]

...in the future one would be able to load a shot into Redcine, save that look as metadata, and then import it back into the camera.

That sounds ideal, although plain-old-3D-LUT support would be nice too. I don't know how much control Redcine is going to offer, but no reason to limit oneself to what one CC app can do.

Presumably the benefit of sticking with what Redcine can do is that you could store it in the raw stream as metadata that you could selectively apply—or not—anywhere throughout the workflow, even as an option within the Quicktime codec! That would be badass.

-Stu

 

[Teague Kennedy]

I urge new RED One users to learn to love underexposed, low-con, low-saturation images as they come off your camera, for they contain the broadest range of creative possibilities for you later.

-Stu

I thought that we were to expose to the right in order to maximize linear data, (would mean that we should have OVERexposed, low-con, low-sat etc.)

?

 

[Joel Kaye]

I thought that we were to expose to the right in order to maximize linear data, (would mean that we should have OVERexposed, low-con, low-sat etc.)

?

Overexpose film, Underexpose RED. Once those highlights are gone they are gone, but RED can keep pulling info out of the shadows. At least that's my plan until I hear differently.

 

[Teague Kennedy]

Hmm. I thought that in linear sensors, more information is allocated to the brighter luma range, so you want your image to be as bright as possible without clipping. Graeme?

 

[jbeale]

> ..you want your image to be as bright as possible without clipping...

Even if you "expose right" to the very last bit of DNR, if you avoid clipping, your scene will most likely look dark until you apply some curves to it. The "film look" (well, "looks") we are all familiar with has/have a very strong built-in curve due to the inherent nonlinear nature of film. Most video cameras have such curves built in as well, making the image brighter but throwing away some info and limiting creative control later.

 

[Joel Kaye]

so you want your image to be as bright as possible without clipping. Graeme?

Well, maybe we're agreeing here then. To me overexposing is by definition going to cause clipping. In many situations "bright as possible without clipping" may be the perfect exposure. Not every situation though.

 

[GlennChan]

Here's how the Red would(/might) work:

The camera's sensor has a native white point / white balance. If the light coming in is a different white point (e.g. tungsten), then one/two of the channels will clip before the others. A simple+effective white balance method is to use digital gain to balance out the channels by multiplying them (by some set of numbers that makes r=g=b for colors that should be neutral). Normally you'd let those one/two channels clip.

In Redcine/Red Alert, you can do highlight recovery by "lowering" the exposure. Redcine will guess the clipped channel information based on information on the other channels that haven't clipped. How well this works depends on the scene... you can adjust the amount of highlight recovery.

The other approach is to get WB correct by using optical filtering... in which case you can't pull this trick. But the channels clip at the same time, which is another way of getting the most dynamic range.

Beyond that, if all three channels clip, then there's simply no information so you have no detail there. On specular highlights this is ok... little hits of highlights look the same whether they're clipped or not (unless you are doing image processing that uses HDR).

2- On the low end, you have sensor noise. The image gradually gets noisier and noisier, until you can't discriminate noise from actual image detail.

If you need to bring the image brighter (and therefore bring out shadow detail), you will bring up noise too. If you underexpose too much, then you can have a noisy image if you fix it in post. However, the Red images shown so far as very noise-free and look great when you bring the exposure up in post. So there's some room to play around with.

3- It would be a good idea to test your camera to see what works best. Bracket a bunch of exposures, and color grade them like you would in an actual production.

4- The behaviour of Red is a little different than traditional video cameras.

In traditional video cameras, there is no highlight recovery.

The highlights are subjected to a knee circuit, which color corrects the highlights so that detail is gradually rolled off. A simple knee algorithm would apply a knee onto each RGB channel; this causes the highlights to de-saturate (and shift in hue). So there are various algorithms that deal with that problem... this is different between manufacturers I believe.

The knee will mess around with the colors in the highlights. You can't undo it in post AFAIK.

On the low/shadow end, it's the same deal. Except you tend to have more noise (noisier sensor, inferior/noisier compression).

Extra highlight information is stored as "superwhites"... values above legal white level. These tend to be clipped off unless you recover them in post (the default behaviour is clipping).

5- In most scenes, there is really dark and really bright detail.

There is likely some pockets of shadow that are dark. At the same time, you have some zinging specular highlights (off metallic surfaces and such). You'll usually lose detail on either end of the exposure scale... so you should decide what you want to keep. (You probably don't need detail in specular highlights.)

Also decide if you want some safety... in case the talent steps into some pocket of brighter light or whatnot.

 

[SalaTar]

That sounds ideal, although plain-old-3D-LUT support would be nice too. I don't know how much control Redcine is going to offer, but no reason to limit oneself to what one CC app can do.

Presumably the benefit of sticking with what Redcine can do is that you could store it in the raw stream as metadata that you could selectively apply—or not—anywhere throughout the workflow, even as an option within the Quicktime codec! That would be badass.

-Stu

Sounds like CineForm

 

[DigitalPraxis]

A simple guide for Digital Cinematography:

Treat the camera as if using reversal film stock.

Expose for the highlights, and light for the shadows.

Simple!

 

[Michael Hastings]

RE: Optical filter white balance.

RE: Optical filter white balance.

Here's how the Red would(/might) work:

The camera's sensor has a native white point / white balance. If the light coming in is a different white point (e.g. tungsten), then one/two of the channels will clip before the others.

The other approach is to get WB correct by using optical filtering... in which case you can't pull this trick. But the channels clip at the same time, which is another way of getting the most dynamic range.

Beyond that, if all three channels clip, then there's simply no information so you have no detail there. On specular highlights this is ok... little hits of highlights look the same whether they're clipped or not (unless you are doing image processing that uses HDR).

I have broached this subject earlier, but would appreciate a revisit. In shooting underwater we obviously have a very heavy overall blue. I have suggested that setting the white balance does nothing due to raw, but that it still might be useful to use red U/W filters to avoid clipping the blue channel so much sooner than any other.

It seems to me this is what you are saying. I asked Graeme earlier and he commented that he had no experience underwater. I am really just asking if my theory is correct - not necessarily the specific situation.

 

[GlennChan]

Hey Aqua,

Graeme would know more than me about that topic. I'd echo his earlier suggestion: do a test. Then you'll definitely figure out the right approach. Try various strengths of optical filtration.

In theory:
A- If you are in a low light situation, then adding optical filtration will reduce the amount of light hitting the sensor (which is bad).
B- The best type of filtration would only reduce the amount of light hitting the 'blue' photosites/sensors. If you look at the spectral transmission charts for the filter, the best one might be a filter with a very steep slope/cutoff for blue/violet wavelengths.

2- If you post your r3d files, you might be able to get some pointers on Redcine/Red Alert settings.

2b- Graeme mentioned that Redcine would do white point adaptation... I believe that algorithm assumes that there's no optical filtration over the lens. So you may or may not get slightly better results by compensating for the optical filtration.
I believe the underlying concept is this:
The white point of what you're looking at will affect your color perception.
If you look at a consumer TV with a ~9300K white point, reds will appear de-saturated.
Whereas if the consumer TV had a D65 white point, reds will look fine. The difference in white point affects color perception. You can compensate the colors so that they look right... e.g. consumer TVs will mess with the color of red so that it looks correct.
Redcine will compensate for that effect when white balance (white balance changes the white point so that all your sources' whites match D65 or whatever your target white point is). If you use optical filtration, you don't get the color compensation.

*This is likely not a huge issue at all.

3- Would you use lighting underwater? That might be a variable to consider??

 

[filmwhip]

Graeme and others,

Is there anyway to do are reverse transform of a file into a .r3d file? Ie can i take a full range 0-1023 10bit log file and convert it into a full range (native camera format .r3d file? This would be really useful for pipeline testing.

 

[I Bloom]

Hello Graeme,

I'm wondering about two settings in REDAlert...DRX and Black Offset. Got a pretty good idea about black offset, but Googled DRX and got nothing.

Oh and is Pick WB an auto white feature?

And also, can you explain the matrix checkbox.

Many Thanks,
IBloom

 

[Graeme Nattress]

Filmwhip - no. Sorry. Especially as R3D is currently RAW only.

ibloom - PickWB allows you to place your cursor on something you think is white and get a white balance from it.

DRX is RED proprietary Dynamic Range eXtension via allowing you to use a region of exposure where there's differential clipping between red, green and blue without it going funky colours.

Matrix governs colour space and white point. If you turn it off, none of that matrix processing is done - to see the raw underlying image.

Graeme

 

[I Bloom]

Filmwhip - no. Sorry. Especially as R3D is currently RAW only.

ibloom - PickWB allows you to place your cursor on something you think is white and get a white balance from it.

DRX is RED proprietary Dynamic Range eXtension via allowing you to use a region of exposure where there's differential clipping between red, green and blue without it going funky colours.

Matrix governs colour space and white point. If you turn it off, none of that matrix processing is done - to see the raw underlying image.

Graeme

Thanks Graeme,

I see, I didn't realize you have to pick the WB point first and then hit the button.

What parameters are governed by Matrix just kelvin and tint or everything?

How and when can one use DRX?

And also can you briefly explain Black Offset?

Many thanks,
IBloom