1st 1,2,3...

[Andrew Wilding]

That looks like a tasty histogram. Since overcoming my fear of underexposure (thanks build 20), my images have been a whoooole lot nice. Much more "filmic", if I may.

 

[Ketch Rossi]

I agree with you Michael, I have had more then once difficulties making DP's understand that RED is not as difficult as some make it to be, it just as in my opinion for anything new, needs to be learned and mastered, and your points are just great in that, they show the way.

IMHO one of the most important things that can be done is to truly Push the Informative teachings of Proper Exposure with the R3D files, and many more will come to use the RED cameras with confidence, and delivering to Post what will be Greatly Exposed Images that in turn can be Color Graded and transformed to the Director's taste, with fantastic results.

 

[David Mullen ASC]

Yes, I believe in the importance of viewing R3D frames later, just to check how I'm doing.

You hit on an important issue, which is to strip the metadata settings off and get back to as close to the original RAW image as possible before beginning color-correction.

I've had this problem before, and I think one of the reasons why the efx shots in "Manure" (as you probably noticed at Lightiron, Michael) were darker than the surrounding non-efx shots is that they must have been processed with some of the metadata settings intact.

Sometimes when I've questioned a post house about this, they say "we want to protect your vision by starting with your metadata settings" but that only makes sense for quick dailies for offline editing, when you want an automated way of getting timed dailies (not that I've found that system to be bug-free either, getting the same metadata to look the same for Quicktime dailies as they did on the set monitor). But for final color-correction starting back with the R3D files, I think it's important to ignore metadata and work from as much information as was recorded originally.

Nothing as as difficult as some people make it out to be -- but on the other hand, nothing is as simple as some other people make it out to be...

After 24P HD hit the scene, I got a lot of the "HD is too complicated to shoot, it slows you down" from film-centric DP's and my response was "HD is as complicated or as simple as you want to make it." At its heart, the F900 was a glorified ENG camera and could be shot as simply as any news camera, handheld, run-and-gun, or it could be surrounded by tents and cables and engineers and reference monitors, etc. Some people have a tendency to let fear turn the whole affair of shooting digital into a circus, as if they were launching the Space Shuttle.

 

[Mark Parry]

1. Incident meter for initial aperture setting. I like to meter into the keylight to understand the highest incident value and then set the aperture to place the exposure where I think it belongs. In scenes of high contrast, the spot meter comes out to judge the highlights and shadows.
2. Check RGB Histogram - adjust aperture slightly if necessary to best capture the scene.
3. Monitor stoplights in RAW to protect highlights.

 

[Ian Laurie]

1. adjust light for 320
2. use false color to find acceptable blowouts and not acceptable blowouts.
3. use histogram to tweak final shot for best "neg"

 

[Bob Gilles]

Contra DP

Contra DP

I knew that I would catch allot of flack for my controversial method of setting exposure, but I have to tell you that with the Realtime grading tools in Scratch, I have not blown one shot yet.

My team has to nab more than 40 shots a day in a heavy VFX environment with mucho compositing and using a scafold to isolate tones, we are on the WITGID program and it works for us.

 

[Joel Kaye]

RED sensors expose with more information in the toe that they do the highlights. By this logic, exposing based on the stop-lights or zebras at the high-end is likely not the best practice.

The stop lights are confusing because they are triggered by specular highlights so I don't think you can use those easily. Zebras are confusing because scenes present all manner of overexposure that are unknowns.

I think I'm right to say the only way to nail it is if you have a known white value in the scene and you have your cameras tools calibrated to it (histogram, false color or zebras all would work). Once you're there you can look at a RGB histogram and make an informed choice about dropping or adding exposure. The result of this will very likely be a histogram that looks just like the one you've posted.

I've shot so many scenes under so many types of lighting now that I can safely assert it's not as simple as "make a pretty histogram". Out in the wild when you're moving fast histograms are sometimes confusing even to the trained eye. Now imagine the untrained eye that's 16 hours into an 18 hour day shooting a documentary.

This following article pretty much nails it in my mind. If references Thomas Knoll.
http://www.luminous-landscape.com/tutorials/expose-right.shtml

I've had a similar conversation with a color math wiz named Steven Wright when I took a class from him. He said something to the effect of "think of yourself as a data acquisition engineer when you're shooting."
http://www.swdfx.com/about/resume.php

 

[mikeburton]

Yes, I believe in the importance of viewing R3D frames later, just to check how I'm doing.

You hit on an important issue, which is to strip the metadata settings off and get back to as close to the original RAW image as possible before beginning color-correction.

I've had this problem before, and I think one of the reasons why the efx shots in "Manure" (as you probably noticed at Lightiron, Michael) were darker than the surrounding non-efx shots is that they must have been processed with some of the metadata settings intact.

Sometimes when I've questioned a post house about this, they say "we want to protect your vision by starting with your metadata settings" but that only makes sense for quick dailies for offline editing, when you want an automated way of getting timed dailies (not that I've found that system to be bug-free either, getting the same metadata to look the same for Quicktime dailies as they did on the set monitor). But for final color-correction starting back with the R3D files, I think it's important to ignore metadata and work from as much information as was recorded originally.

This is where it can get a bit tricky, David. This is what we did on "Cafe" and worked out with "Dive" or VFX house in Philly. First, in Scratch we stripped all the metadata and designed a basic look for the VFX scenes (ie white and black point adjustments only) We gave Dive 10 bit Log DPX, processed through Scratch with our basic adjustments applied to balance out the image for them to time there VFX shots with. We also did a test beforehand with them to make sure we would get back a Log DPX that we could import into Scratch that would match our existing one light LUT and then applied our viewing LUT (Kodak 2383) on top of that for matching to film output.
That said, you can absolutely tell as a colorist the difference between a processed/baked 2K DPX file from the R3D's and an R3D file as far as the amount of latitude you have to work with in the image.
What was nice on "Cafe" was in my case be on the job from beginning to end as DIT/Colorist to make sure that everything is starting from the same reference point and all the different houses know what the other needs. I know its not often this happens this way but it ended up saving a lot of time and everyone in the post process was aware of how we were going to finish the film which made it a breeze. Having LightIron behind you the whole way through is another way to maintain consistency throughout the process.
I know most productions will never have the DIT and colorist be the same guy/gal but I guess I'm referring as well to a good post facility or a post production supervisor. In our case, it worked out well to maintain consistency from Set to Editorial, VFX, Conform and Final CC.

 

[Florian Stadler]

1. Light to meter and set rough T stop accordingly
2. Look at Monitor in Redspace, toggle Zebra & RAW, tweak stop slightly accordingly, commit to stop for scene
3. Tweak lighting according to taste/monitor

 

[Steve Freebairn]

For best results in optimally exposed R3Ds, evaluate the results of your preferred metering from the sensor's point of view based on a raw image data path.

This is key for people to understand. All the meters except for the RAW meter (the thin vertical bars) are based on the viewing space that you are in, so it's critical to set a user button to "View Raw" so you can toggle between REDspace and RAW. Some people get confused because in the lower left of the lcd and evf it says "RAW" that is the recording mode, not the viewing mode, in Build 21 it says "vmRAW" when you are in raw view in the upper middle right of the LCD/EVF. That was a good update by RED.

One more thing that I think could be really helpful to getting people to properly expose is to add REDlog to the viewing modes available. The reason why this would help is because then you would be able to see the exact same histogram in post that you can see on set and "exposing the body in the middle" would be perfect advice and you'd also be able to see your shadow detail without blowing out your highlight detail.

The next step would be to add false color modes to the RED post tools so that when Post is communicating with Set, they would have the same tools, with the same viewing modes.


For me the 1 2 3 changes based on what I care about more,

For Highlights (like shooting in the snow and preserving all detail and DR necessary)

1. View Raw and open up until you see the histogram start clipping, then close down a third stop.
2. Watch the shadows and fill with light to avoid "minimum exposure"

see "Snowmen" http://www.snowmenmovie.com/ the trailer looks like it's a mix between my dailies and final color, but rest assured, there is maximum DR in those images.

For Preserving Shadow Details

1. View Raw watching for shadow detail on the histogram, it's ok to clip speculars and even some highlights (if you absolutely have too) as long as you know that you'll never ever ever get them back. and that you're not planning on bringing them down in post.
2. Use Daylight Balanced lights or Gel your lights, or use an 80 A B C or D filter, based on how much light you have. Even an 80D will help populate the blue channel helping you to avoid fixed pattern noise.

One example that comes to mind is the Lady Gaga footage that is in one of the reels we show at REDucation, there are other examples floating around here on REDuser, where there are 9 lights in the background or car headlights that obviously in a night shoot you don't and shouldn't protect.

It's important to do testing, but it becomes really really important if you're going to start clipping or crushing things on either end of the histogram.

 

[Mark Mannschreck]

view rec 709 (build 20+)

rate camera at 160-200 interior low to medium contrast

incident meter into key

check grey card in key with false color in RAW

in my experience, at that point you've gotta by eye rather than histogram... alot of time in narrative stuff, a fat histogram or ETTR isn't the correct exposure


rate camera at 320-500 exterior

" "

 

[Mark Parry]

When it all comes down to it, experience is the most useful tool to setting exposure, in any format.

 

[Sarah C.]

BEFORE ADDRESSING JIM'S QUESTION, IT IS CRITICAL TO UNDERSTAND 2 THINGS FIRST.

A. RED ONE is NOT a VIDEO CAMERA but a RAW IMAGE CAMERA! There is no onboard processing baking in a look, INSTEAD processing happens AFTER you expose. THINK FILM NEGATIVE.

My post may have triggered this response. My apologies. I made an addition to my post to address this. Dynamic range of the capture was foremost on my mind. Thanks.. ~s

http://reduser.net/forum/showpost.php?p=525479&postcount=9

 

[Jannard]

I have changed my methodology over time (and Builds)... and it now has exposed "what I really want". We are using the new M-X sensor (which has a lot less noise) to do a few new important changes to how exposure will work in the future. One of our goals is to match up metering to ISO rating exactly. The other is to create a new color space (REDcolor) and gamma (REDgamma) that give you a much more consistent view of the full range from A-Z (camera through post). Additionally, Graeme has come up with an ingenious way to deal with ISO rating. The whole system is called FLUT Color Science. We will demonstrate it at RED Day. You still will have legacy options if you want to continue as you have in the past. We really have learned a lot in the last two years and believe that the future should always be better/easier than the past.

Our goal... 1,2,3 that everyone can understand. This year, the comments suggest a multitude of different ways to get to the same place. Next year when I post the same question... I think there will be a lot fewer preferred ways of setting exposure. And a lot less missed exposure.

Jim

 

[MichaelHalsell]

RED Day = Good Day

 

[Ketch Rossi]

Look forward to it.

BTW, is anything Graeme ever does not " INGENIOUS??" :~)

 

[Stephen Strangways]

Simplified procedure for most situations:

1. Set ISO 320

2. Get the highest exposure (most light) as much as possible without blowing highlights I want to keep, by checking the histogram, zebras, and false color in raw. Ignore the traffic lights and raw meter, because they don't show me where it's overexposed

3. Go back to viewing REDspace, and if the midtones are too low, raise the ISO until the image and histogram look pleasing.

(4. Post-production) Process most footage with Redspace color space and gamma, unless the contrast ratio is too high, in which case process with linear light gamma and create a custom curve to suit the scene.

 

[Tim Sutherland]

BTW, is anything Graeme ever does not " INGENIOUS??" :~)

Not that I've ever seen.

 

[Graeme Nattress]

Not that I've ever seen.

That's because you never see my failures

What I've learned is that software image processing is one thing, but as soon as you tie it into a camera you've got an interactive feedback loop through a human and a display, and that's a whole different problem, and a trickier one too.

Graeme

 

[Roberto B]

1) battery.. full

2) tripod because it is a red1 after all

3) record button