What is your favorite workflow for keeping all DR available on MX?

[Gunleik Groven]

Back on subject...

A quick way of checking "what is in there", is to set ISO at 250 and gamma to linearlight. Then you'll see what is in the highlights.

Then you could set the gamma to redlogFilm and drive the ISO towards 1000, then you'll have a good idea of what's in the lowlights.

Then start developing.

Obviously that is laborous and after a while not neccesary at all, but it is a good way to train your eyes as to what you want to manage to take care of in your developed image...

 

[Ido Karilla]

This corresponds well with what I see.
RedlogFilm - when developing charts - seems to be ok at 800, but when developing real-life images skin and highlights seems to be too compressed "up there" to be really gradeable at anything above 320 for daylight images.

I think RedLogFilm is set to comply with Cineon standards, if you want all your data from the sensor you will have to stick to the mysterious RedLog.

 

[Gunleik Groven]

Agreed.

I just see that I still have to push the ISO back to get everything.
Not on charts, but on practical shots...

I also agree that redlog probably is technically better, but a lot more workintensive to use. I will probably use it again, though when I see it the best solution. Redlog does not protect the highlights the way redlogFilm does, though...

 

[paulherrin]

...can you say again why doing a tiff export is not pertinent?

i haven't read what exact method sanjin uses for pseudo-dr, or hdr, or however you choose to call it, but the fact is that the data is there and you can get at it if you choose to do so...

...so i'm just wondering what the big discussion is about?
(and i'm still confused as to why monitoring comes into play, besides the fact that it would be nice to have an easy way to monitor and record quickly and independently)

 

[Gunleik Groven]

Not pertinent... -

Sorry for my english, but not sure what that means...

If it means "not necessary", I think it isn't.

I don't disagree with him that that is one way of getting the full latitude. Much easier though (if you want to go 16-bit tiff) just to export one tiff pr frame at 320/250 linear-light. You'll get the same info out.

Exporting three out at redgamma, doesn't really give you any extra info.

Not sure if that answers your question, though...

 

[Paul Leeming]

So what I see, is that I can recover all the skintones and all but the left elbow from his dress...

I still like the dramatic development you did, though -

But if I was to prepare it for a grading session, I'd have attempted to get as much of both the highlights and lowlights recovered as I possibly could...

The first attached image is the only part of the image where I didn't manage to recover texture (1/1 crop from the left elbow)...
Not much really...

the rest is pretty much intact... (2nd image...)

@ 800/redgamma, though... (3rd image)

(btw, hope it is ok to use your image as an example. I have my own, but the most relevant are NDA... Will remove if you want me to!)

No worries Gunleik, it's all good and what you recovered is exactly right. I allowed a tiny bit of overexposure as he flashed through the sunbeams coming through the trees, as it is natural to our eyes as well and gave the best balance in the dark areas without blowing out the important textures of the jacket and reflections off his head. It's only for a few frames at a time and was intentionally let go for that reason. Nice thing is that with the DRX slider you can basically balance it to his jacket white perfectly so in motion it's a non-issue.

This is a great discussion! I love technical stuff like this and dissecting it with others who are as passionate about image quality as I am

That's what makes this forum great - passionate professionals who are just interested in the imagery and making the best of it. There's a core group here who have given me so much combined knowledge over the four or so years I've been here (and DVXUser before that) that it's my pleasure to give back in my own way and continue learning from such a treasure trove of helpful information.

Cheers from Japan,

Paul

P.S. How great is it that RedCine-X now exports R3D single frames?! Makes these sorts of discussions SO much simpler when you can give the RAW to everyone so no colour decisions cloud the process, since we can all extract the best info from each shot. Thanks Red Team!

 

[Anson Fogel]

Gunleik, (and Paul)

Thank you, thank you. Wonderful discussion, and addresses specifically some of my core questions shooting and post-ing MX of late. I've stumbled on some of these issues and solutions myself, but never spent enough time really playing with it to settle into a flow that works. Developing daylight exteriors at 320 RLF has been useful when grading the output online elsewhere, non R3D - it sound like that may be what you refer to Gunleik when you say online/preparing/developing for later online.

I am still using that workflow mostly for commercial projects for speed and simplicity, but am moving towards resolve, where that is no longer the case - I suspect that the same issue/questions arise to really maximize that latitude. And I still (again, for daylight) land at shooting 800 (with annoying weaknesses, as you note, in the monitor path) then developing at 320 RLF, for shots where I need to.

Gunleik, you asked the core question - what is YOUR answer?

 

[jake blackstone]

I would like to contribute my 2c for this discussion.
As this thread is called "favorite workflow for keeping all DR available on MX" it is important to remember, that there are two parts to the process, be it film or electronic- an acquisition and post processing. Mistakes in both can adversely impact the DR of the final image. Obviously, with electronic acquisition technology, overexposure is a much more series offense, than with film. Conversely, bad film developing is much more egregious offense, than with electronically acquired image.
What does this all mean to me? Well, being a colorist, I am intimately aware, that no amount of post magic and experience can save an improperly exposed Red material. None! Now, bad post can definitely contribute to that same sad goal of bad finish, but, luckily, with Red, bad post is perfectly reversible. So, in this never ending RED exposure/post dance, I would give much more weight to the proper exposure, as it is not reversible

 

[Gunleik Groven]

Jake,
Problem with answering that question, is that the answer is relative... That's why I have tried to keep the conversation strictly theoretical and technical...

In real life I'll "look at the scene and decide what suits it the best..." when shooting, -

In post I'll "Look at the clips and make decisisions from what I see when preparing a grade..."

And when grading... You get my dig? -

There are really no "good for all" solutions. There are some technical pointers, though - and RLF has made them much more clear. I can now give "stock answers" where I know the result will not be outright bad. The baseline of RED post has been risen significantly through this. So let's go baseline and technical...

Main rule of thumb:
Don't confuse the processes:

Exposing
Preparing for grade
Grade

If you want your images to have maximum flexibility.
The considerations in each step differs slightly and you will get the best result if you focus on the part of the process you are in. This is true even with a RAW-online in mind...

On set.
All camera RAW metering is relative to 320/daylight, but at 320 there's no headroom in the highlights.

On the MX you can expose a lot lower without adding noise, it is quite touchy in the top, though, giving yourself some headroom in the top is good.

Thus, exposing at 800 or above is good advice.

Remember that you that your headroom is less at tungsten than it is at daylight settings. The key indicator of breaking the highlights at tungsten is not the RAW meters (because it doesn't actually break in RAW...), but "metallic" highlights in skintones and hair. They indicate that the RED channel is being clamped, and you have two choices: Lowering the exposure or go to daylight ballancing, accepting that the image will look "yellow" on set and first part of post.

Hey, what's the issue when the RAW image isn't broken?
The issue is that if you want to WB your image with the metadata-tools, you will break the image anyway...

In other words, it is good advice to set your ISO higher at tungsten then at daylight, IF YOU ARE IN DOUBT to get a bit extra highlight protection. (OTOH if you're not in doubt, do whatever suits you... - )

As Paul said, and I have implied, it is good technical advice to "expose to the right". That doesn't dig to well with a lot of DP's though... Many are used to much MORE latitude than the MX offers (this is not disregarding the MX, but a fact of film...)And they already expose well into the "right" because they're used to 2 stops headroom in the top.
I think it more often is a confusing term (because DP's easilly expect more headroom than what is really available...), and I have conciously stopped using it, but of course: The more you "fill up" the sensor, the more fun you can do in post. Just don't break it if you don't intend to.

Focus on composing a pleasing image, don't worry too much. A bad framing, makeup or wrong camera-move is a lot more distracting than the occational over- or underexposure... Be a photographer when you are on set. don't worry too much and trust your eyes and skills. Look for images and storytelling primarely!!! -
(I do agree with you Jake, but still I see people worrying too much and getting distracted by the technical stuff. Really, the RED is just a camera, trust your photographic and storytelling skills, after all that's why you are shooting... Be an artist!!! -

If you are the DIT, discuss and examplify exposure with the photographer IN ADVANCE, not on set (if you don't have to...) Many DP's will break the highlights, because they want to. Nothing wrong with that, as long as it is a concious decission! It leads to some extra work in the next step, though.

As it is not bad to conciously underexpose. It can be beautifully done in the right hands... Nothing is illegal but NOT knowing what you're doing -


Preparing for online
I really think this step should be separated from the grade, even in an all RAW workflow. Because THIS is the purely technical step which transports the info from the "artistical and aestetical" shooting process to the "artistical and aestetichal" coloring/VFX and finnishing process.

Stock answer:
Go RLF at 320 for daylight and 160-250 for tungsten, if it is a requirement to fully tungsten ballance the files prior to grade.

You could also go 16-bit linear-light TIFFs at daylight/320, but for most of us that is not a practical option. It is the only truly lossless way to go, though.
I must admit I rarely miss this option, even though it allows for true batchprocessing...

Practical answer:
Use your eyes and know what to look for and use whatever means that gives the best result...

Thus, I really feel strongly that this is NOT a batch-process. If it was, we could as well have been shooting a baked format...

So, what do I set out to do?
Protect the highlights
Protect the lowlights
Give some headroom for the colorist in both ends (it doesn't help much to get "all" the info out of the shot if it is either too compressed in the top or if it's no headroom to grade "up there", the same goes for the bottom)

The channels break unevenly in the top under different lighting conditions. This can give really ugly casts in the highlights just where it breaks. DRX is a fantastic tool for fixing this, but it is not available with Rocket transcodes (RAW workflow or not) , thus I have to go in and manually adjust the signal in the top to avoid uglyness.

The RLF is possibly technically capable of holding the highlights at any ISO, but I find the top to be too compressed to be easily gradeable, and have found myself to step back to 250 even under daylight, to get some headroom and "curves" in the top of the channels for the grading-session. (This would be true for RedLog as well...)

But it is a bit low in the lowlights, so I usually end up using a curve in the bottoml, lifting the lowlights into a range where the colorist can easilly see that he/she actually doesn't have to clamp it all down. It is important for the signal that this happens when still in RAW. When you're in RGB/YUV (which also applies when grading RAW: post the "RED development stage") the info is gone...

And this is a favourite subject of mine... So far only Clipfinder and RCX have a curve to use on the raw file prior to development. Resolve, Color and AFAIK even Scratch doesn't have this, and that leads me to frequently suggesting DPXs/10-bit AJAlog as online format even for RAW-capable applications. And after some A/B tests, I am usually listened to...

When Tungsten is exposed "to the limit", I very often step even further back on the ISO to avoid clamping the red channel and also not fully whiteballancing it, as I find the result more pleasing when the colorist finnishes that process with some headroom in all channels, than when he/she gets clamped channels into the grade.

Online/Grading:

Back to artistic stuff... -

When making dailies, I actually do much the same, just using RedGamma.

Step back to reclaim the highlights, compensating with a curve in the bottom to get the mids/lows back to where the DP shot them.

Welll....

That's some of what I think on the subject...

Cheers!

G

 

[Graeme Nattress]

I think RedLogFilm is set to comply with Cineon standards, if you want all your data from the sensor you will have to stick to the mysterious RedLog.

REDLog was designed primarily as "data transfer" to get a really good use of the bits going from the 16bit internal processing format to 10bit.

REDLogFilm is also perfectly usable for the above use, but it is also designed to be very gradable and compatible VFX workflows by being matched to the EXR output.

Graeme

 

[Rudi Herbert]

So Graeme,

If we do all of our grading in RAW on CS5, where we only transcode after all manipulation is done, would you still recommend RedLogFilm or RedLog? I seem to get better colors out of RedGamma in this case...

 

[Ido Karilla]

REDLog was designed primarily as "data transfer" to get a really good use of the bits going from the 16bit internal processing format to 10bit.

REDLogFilm is also perfectly usable for the above use, but it is also designed to be very gradable and compatible VFX workflows by being matched to the EXR output.

Graeme

Hi Greame,

First - Thank you,

For me REDLogFilm is a blessing.
I had so much trouble working with REDlog for film outs.

Never the less there are those shots (badly exposed) that REDlog helps me recover a signal in the highlights.

Any chance that you are working on a REDColorFilm that will preserve more color depth in "skin tones" through the digital to analog roundtrip?


Thanks again.

 

[jake blackstone]

Jake,
Problem with answering that question, is that the answer is relative... That's why I have tried to keep the conversation strictly theoretical and technical...


G

Unfortunately, with Red, theoretical doesn't cut it. In theory for a given exposure of 18% gray card with same ASA you should always get the same signal out of the sensor or the exposure on film. That works pretty well for film and once it is tested, developed and with ASA selected, you'd stick with same setting for pretty much the whole shoot.
Not with Red. In case of Red, ASA number becomes a derivative of particular scene DR and to a lesser degree, of color temperature. Yes, you can and should come up with various rules of thumb for Red ASA selection, but, at the end, nothing can replace the experience of actually shooting Red under various lighting conditions. So, when talking about Red, choosing proper ASA becomes of paramount importance, which can and will affect the final DR more, than anything else in the process.

 

[Gunleik Groven]

Yup.

Thus I tried to line out the things I have had success doing... -

But then we quickly enter "taste" land...

 

[Hrvoje Simic]

Give some headroom for the colorist in both ends (it doesn't help much to get "all" the info out of the shot if it is either too compressed in the top or if it's no headroom to grade "up there", the same goes for the bottom)

Depends...on what type of scene is and where you want to go in post.
Lowering down from the clouds is less of an issue than raising up from the dirt.

 

[Mark Wilkinson]

Thanks for the dialogue - but I'm curious about the logic of exposing to the right.

For film we exposed to the right to get what we called a thick negative to work with in telecine and the shoulder of film made that a safe approach.

If the MX sensor lowered the noise floor but didn't significantly improve the handling of the shoulder, wouldn't one expose to the left? Which is to say protect highlights and let mid range settle into the low noise middle-left of the histogram?

Is this what you are after by setting ISO at 800...?

So what's the logic of doing both? Or are we saying the same thing...

And also, changing the ISO and adjusting the curve are different mathematically (if I understand how ISO works on the chip), so oughtn't one keep the ISO at 320 to minimize noise and light or filter to expose about 1 stop to the left. Then, once you have the image with the least noise in RGB, apply your grading curve.

The only way to protect your highlights is by cutting the light hitting the chip - you can't trick the sensor into being less sensitive by ripping the Raw footage differently.

It's all about managing the light that hits the sensor - and minimizing the noise generated - which is to say exposing at 320. The rest is a shell game.

Graeme - do I have the data path on the ISO - Curve issue correct?

 

[Hrvoje Simic]

Thanks for the dialogue - but I'm curious about the logic of exposing to the right.

For film we exposed to the right to get what we called a thick negative to work with in telecine and the shoulder of film made that a safe approach.

You also want a thick negative here because more data is recorded in highlights.

http://www.luminous-landscape.com/tutorials/expose-right.shtml

If the MX sensor lowered the noise floor but didn't significantly improve the handling of the shoulder, wouldn't one expose to the left? Which is to say protect highlights and let mid range settle into the low noise middle-left of the histogram?

Forget about the shoulder in RAW. You make the shoulder as you wish.

Yes, the goal is to protect highlights which are on the right on the histogram, hence "expose to the right".

 

[Mark Wilkinson]

You also want a thick negative here because more data is recorded in highlights.

http://www.luminous-landscape.com/tutorials/expose-right.shtml



Forget about the shoulder in RAW. You make the shoulder as you wish.

Yes, the goal is to protect highlights which are on the right on the histogram, hence "expose to the right".

Hm. There is a shoulder, but it is just linear, right?

I get ETTR, but now that the new generation chips have reduced the noise floor, isn't protecting the hilites paramount? It seems like protect means "don't clip" and thus avoid the right side at all costs.

 

[paul engstrom]

Hm. There is a shoulder, but it is just linear, right?

I get ETTR, but now that the new generation chips have reduced the noise floor, isn't protecting the hilites paramount?

My understanding of ETTR is that that digital sensors do not record information equally from black to white/left to right. If you took one single frame of the same image exposed to the left and compared it to the same image, but this time exposed to the right, the file size would be bigger for the second take. As a result, you would have more information for your grading software to work with and the potential for a better result.

Furthermore, the intention behind ETTR is not primarily to protect against noise or to be lax with highlights; its always a given that you will protect what level of highlights you want to protect. But, for the reason I outlined above--recording as much of the image in the most data-rich exposure area of the sensor--you will continually push against the limit of loosing your highlights, frequently overexposing in the process.

The over-exposed raw image, when brought down in post, will look better, with more subtle color details, than if it had been exposed 'correctly' (not to the right) at acquisition.

 

[Mark Wilkinson]

All makes sense Paul - we're going back and forth on the same issue that Stu dealt with a while back in this blog:

http://prolost.com/blog/2008/3/2/exposing-to-the-left-vs-exposing-to-the-right.html

Notably blogged before the MX chip, I believe.

My only salient point regarding this string was that changing the ISO is a shell game that introduces noise before the curve. Better to rate at 320, avoid clipping, expose your histogram as much to the right as you are comfortable with and apply your curve last.

If you are feeling sassy rip through Redlog for your grade.