Shot Redspace, going to film out, NEED HELP UNDERSTANDING WHEN TO LIN INTO LOG

[Viggenboy]

Up here in sunny, (translate:Rainy) Portland Oregon, we are conforming our first Red shot feature on Scratch. We're comfortable using Red and Scratch in the HD world, but this is our first film out and I have a huge hole in my understanding of the workflow:

When is the correct stage in the process to make the lin to log conversion, and what is the right way to do it?

The film was shot in Redspace. So when our colorist is working on it, in the process menu we have the color space set to Redspace, but what is the correct display LUT for him to be working in? Respace, or do we want RED Log because our intention is a film out? PDLog 985? He has been working in Red Log.

If we're on the correct path so far, (Colorspace at Redspace and Display LUT on RedLog), then when we render out dpxs to send to the lab, should we render those out with the same Red Log output LUT to create the dpxs using the log curve, or keep them clean? We are having a call with our lab this afternoon, but I don't know what their Red/Scratch experience is and would love to get some insight from those of you who have already walked the walk.

I know this is a newb question, but we're toddlers taking our first DI steps and we're trying to walk to the couch but gravity is a harsh mistress, and oh crap that hardwood floor could really hurt. Please help!!

Thanks much,

-Tony

[John Tissavary]

I don't know about "THE" right way to do it, but after some decent results using my own customized set of the PD958 log curves, I'm experimenting with Red Raw Log (in the actual Rednode in Scratch 4.2, so no LUT required). I then use a print emulation 3d lut for Vision Premiere (the intermediate stock most of my clients print to) as the display lut (LUT module) to grade in. I've not had a chance to print yet, but the monitored results look very promising.

It actually doesn't make a bit of difference what lookup the DoP had the monitor set to when exposing, other than to have a reference to that work so you don't deviate too far from his/her intentions unless that's mandated. Frankly, I reset all the metadata prior to grading, especially for film.

Redlog does not, in my experience, transfer that well to film.

PD685 works, sort of, but you're now reducing the information in your digital value by nearly half to fit into PD685, which is a massive compromise right out of the gate.

If you really want to see on film what you're seeing while grading you're going to have to develop your own lookups that take into account your monitoring profile, print stock, and lab that will do the print. There are a number of software packages out there to do this, I can recommend Cinespace (http://cinetal.com/products/cinespace.asp)from personal experience.

Since it's not possible to describe film response to light with a single formula, grading for film without print emulation LUTs is like grading with really lousy sunglasses on.

If no one at your facility has experience grading for film then I strongly suggest you hire someone who does. Grading for a successful filmout is not a trivial process, and feeling your way through this without experience will almost certainly cost quite a bit of time and money.


regards,

John

[Viggenboy]

John, thanks much for the response! Very helpful stuff.

We've talked with the lab and are taking the exact path that you recommended here: They have cinespace and are making us a LUT for Scratch that will adapt the full range 4:4:4 view on our Cinemage to their specific printer and stock.

We're definitely on a steep learning curve. Can't wait to see the results of this test.

Thanks again,

-Tony

[Michael.B]

The best way would be to color in linear space and have the filmout house do the log conversion. The log conversions provided by the RED tools are not really log. A good filmout house will have the mathematical inverse lut that should convert everything you've looked at in the linear world into a log equivalent.

[Hans von Sonntag]

The best way would be to color in linear space and have the filmout house do the log conversion. The log conversions provided by the RED tools are not really log. A good filmout house will have the mathematical inverse lut that should convert everything you've looked at in the linear world into a log equivalent.

I've done quite a few filmouts in the last year of prrojects I shot with the RedOne. Because I'm not an engineer and my knowledge concerning colour is ambitious but not scientifically deeply founded I leave the interpretation of the REC 709 RGB space to the lab and its talented staff.

They are able to adopt with their LUTs the picture you graded and reviewed on a calibrated REC 709 screen to the needs of film.

I do it this way:

1. Calibrated REC709 monitor. In the lab they will look at the identical pictures you have seen an approved.

2. I have an slighty seasoned Sony VPL100 projector. The projector gives you a clou how "dark" you can go and how contrasty a szene would look like in the cinema. Bare in mind that working off a projection one tends to crush the blacks too much. The reason: on the big screen black is not black but more or less grey, depending how much available light hits the screen. Here the scopes are your best friend.

3. Film has a different shoulder in the whites. You have to keep in mind that white on the screen is not necessarily white - only if it reaches 1023 or above in a typical 10 Bit environment. In this regards film tends to make whites light greyish. Different LUTs in the Lab can help here.

4. Talk to the colorist in the lab. They usually know a lot and want to satisfy your needs. Listen to them. When I did my firts home-grow DI the colorist asked me why I did not pushed contrast. He was right. If I would have had a decent projection the whole film would habe been more contrasty.

This approach is NOT the way Pros with big budgets are doing it. But leaving the LUT/log/lin-madness to the ones who deal with this all the time makes a lot of sense.

Hans

Sidenote: RedSpace is not a colour space such as DCI is. RedSpace is a LUT that Red developed to help the users to create instantly vibrant and nice looking pictures for REC 709 colour space. Rec709 in the RedOne settings again is an other LUT for, and named after the REC 709 colour space. To grade off RedSpace is surely not ideal. I would either grade off the RAW, RedLog or CineForm Log90.

[John Tissavary]

The best way would be to color in linear space and have the filmout house do the log conversion. The log conversions provided by the RED tools are not really log. A good filmout house will have the mathematical inverse lut that should convert everything you've looked at in the linear world into a log equivalent.

That may be a convenient & reliable way, but it is certainly not the 'best' way.

For instance, if you grade in full gamut you'll likely be creating colors that are not film legal, and if you work in Rec709 you'll be working with less colors than are film legal. If you rely on a Rec709 to print conversion you'll also probably lose a bit of detail in the highlights & shadows compared to what is possible had you been working with a specific emulation LUT right from the get-go.

My philosophy is you can only get part of the way by using a lin-to-log LUT after the grade, much better to do the grade itself in the print colorspace.

I'm assuming that since this is the Scratch section we're talking about Scratch (vs. Red supplied software), so an experienced colorist utilizing print emulation display luts will likely get a better result than a canned lin-to-log conversion.

regards,

John T.

[M Most]

That may be a convenient & reliable way, but it is certainly not the 'best' way.....
My philosophy is you can only get part of the way by using a lin-to-log LUT after the grade, much better to do the grade itself in the print colorspace.


John makes some very good points here. There are good cheats and bad cheats. Color correcting in video space, on a video display, for a film finish is a bad cheat. Using a properly designed print preview LUT, in a properly calibrated projection environment, combined with a log converted original image (many ways to do that) is a good cheat. Of course, if a film recording is not part of the picture, the "best" way is to be color correcting in an environment that represents the intended delivery, be that HD video or digital projection.

[Michael.B]

That may be a convenient & reliable way, but it is certainly not the 'best' way.

For instance, if you grade in full gamut you'll likely be creating colors that are not film legal, and if you work in Rec709 you'll be working with less colors than are film legal. If you rely on a Rec709 to print conversion you'll also probably lose a bit of detail in the highlights & shadows compared to what is possible had you been working with a specific emulation LUT right from the get-go.

My philosophy is you can only get part of the way by using a lin-to-log LUT after the grade, much better to do the grade itself in the print colorspace.

I'm assuming that since this is the Scratch section we're talking about Scratch (vs. Red supplied software), so an experienced colorist utilizing print emulation display luts will likely get a better result than a canned lin-to-log conversion.

regards,

John T.

Hey John. You bring up valid points. I should clarify that working in REC709 works well when you consider your DCP or HD finish as the master. Many consider film to be secondary, as the life of a picture resides on linear media forever, and the film is only in the theaters for a few weeks. Many never get the chance to do a filmout, or have limited film release. This method also avoids having to do a video trim pass, as you stated there are many video colors unavailable in film. Some, however, consider film to be the top concern. In this case it would be best to do a log conversion up front and color in log space with the proper film luts to ensure an exact match.

[Lucas Wilson]

Lots of good info from lots of smart people with experience.

Speaking strictly from the SCRATCH/RED perspective, the cleanest and mathematically purest way to do grading on R3D files is to keep them in linear light. Apply proper viewing LUTs to get into a normative space for calibration and output. But do all grading based on a linear light R3D with color metadata removed.

Mike M., Mike B., and John T. all have more practical experience with film outs than I do, so I'll leave that to them. : )

But for the purest results out of SCRATCH, use linear light and grade from there.

Lucas
-----
ASSIMILATE, inc.
LA, CA, USA

[Michael.B]

Lots of good info from lots of smart people with experience.

Speaking strictly from the SCRATCH/RED perspective, the cleanest and mathematically purest way to do grading on R3D files is to keep them in linear light. Apply proper viewing LUTs to get into a normative space for calibration and output. But do all grading based on a linear light R3D with color metadata removed.

Mike M., Mike B., and John T. all have more practical experience with film outs than I do, so I'll leave that to them. : )

But for the purest results out of SCRATCH, use linear light and grade from there.

Lucas
-----
ASSIMILATE, inc.
LA, CA, USA


Hey Lucas, good to see you back and feelin better.

[John Tissavary]

Micheal, you're right - if HD or other linear medium is the primary delivery then Rec709 or some reasonably similar viewing space is the way to go.

Lucas - it's definitely not a trivial job to squeeze every single drop of goodness out of a Red filmout, and such a breeze to make stunning images with Scratch in linear.

But even though there are probably some complications when grading Red in log, my experience shows that applying the log conversion up front, then grading in a print emulation viewing environment ensures that the colorist has the most control to land the detail where it will 'line up' with the way film responds to light.

I am experimenting with the new raw log space in the Red node settings for a filmout in March - seems really promising. Can you elaborate some on exactly how this works?

cheers

John

[jake blackstone]

Autodesk did some testing with RED and had published a decent Lustre white paper on this subject.

[Steve Shaw]

Just as a point of note - You need a Film Calibration LUT for the process you are describing, and these have nothing to do with the source image at all...

Most (all?) film recorders assume the data they are going to be fed is 'Cineon/dpx Log'... If you present a tv gamma image (linear...), it is converted to Log via internal LUTs before the film-out process is performed.

So, the best way to approach a project that is going to be filmed-out, and you want to see an image that is accurate to the final film you need a 'film print' LUT, and then grade through that.

In this way you will 'push' the Red images into 'Cineon Log' space, ideal for the film-out.

Because of this, RedLog is the best way to go, with all other metadata turned off.

Does this help?

Have a look at the info on my website (www.lightillusion.com) to read more about this. The DI and Scene-to-Screen tech papers may be a good starting point.

Steve

[Steve Shaw]

Oh, and I should have added that you can then use a seperate LUT for any tv deliverables, matching them to the film look, but taking into acount the different viewing environment/monitoring...

Steve

[Antoine Baumann]

so why not have a lut that do linear light to log to print, so you actually see a "printed log" image while grading on the linear light?


, and these have nothing to do with the source image at all...

can you elaborate?

Thanks,
antoine.

[Viggenboy]

I just circled back on this thread and wanted to thank everyone for the input! In the end, we've taken a path pretty close to what Steve Shaw has suggested: Worked closely with the Lab who created an emulsion LUT for us using Cinespace. They factored in us working with Scratch going 444RGB into a Cinetal Cinemage, wide gamut. Its incredible how much the emulsion LUT alters the image.

In Scratch, (Before we knew about working with REDRAW LOG), we set the Red Color Space to Redspace, then used the RedLog Display LUT.

Then on the output of the DPX files, we baked in the Redlog LUT. I was always suspicious of this, not understanding what rendering that on top of rendering out Log DPXs would do. Was it potentially doubling the log curve? But we sent many DPX render tests to the Lab, and the fore-mentioned gave the best results.

We just saw our first film out test projected the other day, and the proof is in the pudding. Its impressive how much our image on the Cinetal is mirroring what the projected image ended up looking like.

So to those in the know, display and emulsion LUTs aside, if you take Red footage, (Linear), and render it out to Log DPXs, is that process applying a log gamma curve to the files in the render process?

We started this project on a certain path, but I'm looking forward to testing the same footage using RedRaw Log as the red colorspace. Partnering with a lab thats willing to test with us has been clutch.

Thanks again all.

[John Jonathan]

I just circled back on this thread and wanted to thank everyone for the input! In the end, we've taken a path pretty close to what Steve Shaw has suggested: Worked closely with the Lab who created an emulsion LUT for us using Cinespace. They factored in us working with Scratch going 444RGB into a Cinetal Cinemage, wide gamut. Its incredible how much the emulsion LUT alters the image.

In Scratch, (Before we knew about working with REDRAW LOG), we set the Red Color Space to Redspace, then used the RedLog Display LUT.

Then on the output of the DPX files, we baked in the Redlog LUT. I was always suspicious of this, not understanding what rendering that on top of rendering out Log DPXs would do. Was it potentially doubling the log curve? But we sent many DPX render tests to the Lab, and the fore-mentioned gave the best results.

We just saw our first film out test projected the other day, and the proof is in the pudding. Its impressive how much our image on the Cinetal is mirroring what the projected image ended up looking like.

So to those in the know, display and emulsion LUTs aside, if you take Red footage, (Linear), and render it out to Log DPXs, is that process applying a log gamma curve to the files in the render process?

We started this project on a certain path, but I'm looking forward to testing the same footage using RedRaw Log as the red colorspace. Partnering with a lab thats willing to test with us has been clutch.

Thanks again all.


You choose your color space when you conform. .R3D files can export DPX as RedLog, RedSpace, Linear etc.. 10bit Log to 16bit Linear. Thats where you choose your Colorspace. When you burn in a LUT you are not changing the colorspace per say, you are changing how it is seen. But Depending on how your LUT is working that can be a grey area.

But keep your colospace the same along the way and don't burn in your LUT unless you know exactly what you are doing and that is decided by the film printers. You need Rules and Method or you are just muddying the waters it's gonna make it harder for yourself to ever understand DI.

Your LUT is just your glasses, how you see the Footage. Leave the footage colorspace alone, grade and conform in that colorspace. Or at least convert LOG to LIN, Grade and use the exact LIN to LOG method to reverse it. Try to Leave in 16bit Linear or 10bit Redlog, and use your LUT glasses to see the footage correctly.

Also buy a book or hire support to walk you through it. You might have had a happy accident by taking 16bit Linear and filming out a REDLOG LUT and having it look right. Be careful because you might be in for a Big Spanking if you rules and methods are not correct.

Also Steve is a Quantel guy, he knows what he is talking about, but I think he thinks the Scratch is kinda a toy, where the iQ Pablo is the real tool.

[Rainer Fritz]

We had to finish three features last months graded on nucoda, baselight and lustre, i have to say first, and had problems to view the footage with the LUT which the labs using normally on their scans, when we were using Redlog, PDlog685 or PDlog985 -2 Stops.

So Steve what would you suggest, if we are not working on scratch, how to get first a LUT for the viewing environment to grade in log, with which red export settings?

For now we decided on the three projects and grading systems different ways. One way is to export in redspace/redlog from RED Raw to 10bit RGB DPX and grade linear full range, doing the lin/log conversion for filmout on the outgoing side of the grading system or on the ARRI laser direct. Same with redspace/PDlog685. This seems to give a good working base on grading, but a more videoish base look of the material....

So, but who can provide a LUT, where it is possible to match the viewing environment to the red footage, to the filmstock? A truelight LUT? So without a special LUT no way to grade in log ?

thx
rainer

[Evangelos Achillopoulos]

We had to finish three features last months graded on nucoda, baselight and lustre, i have to say first, and had problems to view the footage with the LUT which the labs using normally on their scans, when we were using Redlog, PDlog685 or PDlog985 -2 Stops.

So Steve what would you suggest, if we are not working on scratch, how to get first a LUT for the viewing environment to grade in log, with which red export settings?

For now we decided on the three projects and grading systems different ways. One way is to export in redspace/redlog from RED Raw to 10bit RGB DPX and grade linear full range, doing the lin/log conversion for filmout on the outgoing side of the grading system or on the ARRI laser direct. Same with redspace/PDlog685. This seems to give a good working base on grading, but a more videoish base look of the material....

So, but who can provide a LUT, where it is possible to match the viewing environment to the red footage, to the filmstock? A truelight LUT? So without a special LUT no way to grade in log ?

thx
rainer

We do that in a daily basis... read the link in my signature... the cost of it is the 1/10 of what Viggenboy payed for... all these questions are problems solved for us... we are a LAB that you can partner with and have the best results... all the comparisons that they have made for us (clients) with other Labs that are following the REC709 style were in our favor because a closed loop LUT Log workflow always gives the best velvet blacks with the best color rendition. The REC709 style is for kids play, the real DI is always Log.

I want to add something more in the Log or linear debate... Linear is clipping blacks and whites like video (even on film) and on the other hand Log has the regular film behavior (compressing blacks and whites)... So just by an artistic point of view Log is much better for grading, RED footage takes something from the magic of film response. And this is the reason of the videoish look... Linear workflow...

Beware that we don't partner with everyone there are rules that we follow to have the best post houses in terms of tech expertise and market share. There are regions like New York and Israel that we don't accept new partners in order to protect the existing ones.

[Harcharan Singh]

Hi,
Just digging into this old thread. Just wanted to know how things are being done with Red Color Science and Scratch 5.1 so that we can plan for an upgrade?How different it is from version 4.4?
Thks
Harcharan