grading red footage

[Johnny Friday]

REDLogFilm is pure log, and can be described in the form y=A log (Bx + C)
Linear is of the form y=Ax+C, although for it to be linear to light we want C=0
Video gamma is of the form y=Ax^C, although more complex versions are used where a small linear section is placed near x=0.

Graeme

Yes, i see now......very simple indeed. Thanks for the explanation. I had a z where there was an x and that is what i was missing

 

[M Most]

I guess the next question would be with ACES you should be applying LUT transformations before the grade, correct? So you would go REDLogFilm > ACES > corrections > Linear > grade as needed/wanted or am I thinking about this the wrong way where ACES would theoretically replace the Linear transformation?

The ACES input transform is fed by the linear light representation. So you would set the Red SDK setting to linear light, and either Redcolor2 or Camera RGB depending on your particular flavor of ACES IDT. The linear light representation is a no-loss situation, it is the direct result of a 16 bit debayer process. In most cases, you would not necessarily do a pre-transform to linear light, you would most likely do a "live" passthrough from the Red SDK's linear light output to an ACES processing pipeline. That's the way it's implemented in Resolve 8.1, and that's the way it's handled in Baselight and Lustre through the use of 3D LUT based transforms. The differences come more on the output side, where some systems (Baselight in particular) allow you to transform back to your display format and append additional grading in that environment, while others are designed to do all color manipulations purely in ACES space, with the transform to display space happening as the final step. I like the flexibility of the Baselight approach, but both can work well.

The reason I suggested putting an initial grade prior to the LUT is that not every image is perfectly exposed. 3D LUTs in general are built on the assumption that the image has proper exposure. If that is not the case, the LUT will likely cause some clipping at the high end (for overexposure) or the low end (for underexposure). A grade prior to the LUT allows you to fix those exposure issues before performing the log to video transform. It also allows you to change values based on the original image rather than a transformed version of it. All of this is particularly useful when you're not dealing with Red images directly from the R3d file, but from a transcoded version of that. If you're working live from the R3d files, you have exposure control still available via the SDK, but if you're working from DPX files, you don't. Even in a properly done transcode, an overexposed image is going to cause clipping issues through a proper log to video LUT.

 

[Graeme Nattress]

Using cameraRGB is fraught with danger and should be avoided. If you do want to use ACES, and I'd recommend you don't as it's buggy (not the RED implementation of the colour space, but the Academy's rendering process) get REDCine-X to export a EXR with the ACES space.

Graeme

 

[paulherrin]

thanks for the info guys. it's really not that complicated to understand that the world is not simply made up of log and lin... there are endless possible variations. but i will say that doesn't make a real huge difference to 'civilians' like myself, as long as I'm aware of it and how it should be properly handled.

 

[Gabriele Turchi]

My apology ,
i menat Video Gamma when i said linear .

Redlog using a Print LUT : it make sense to me using a Print to if going to print
Redlog using a Log to video Lut : doesn't this produce basically red gamma 2 ??
red log using a custom curve : this is valuable , but i have a question : bending (design a curve) does not shift white and black level , so when is that supposed to be set stretching the image ? before or after the curve ? how? using lift gamma and gain ? i am hearing of people using red log film and going straight on the 3 way color corrector (lift gamma gain ) but . isn't is wrong using lift gamma and gain a log encode image (before making it suitable for video gamma space?)

tahnks
g

 

[M Most]

Redlog using a Print LUT : it make sense to me using a Print to if going to print
Redlog using a Log to video Lut : doesn't this produce basically red gamma 2 ??

No, it does not. Redgamma and Redgamma2 are interpretations that have never really been either fully documented or explained, other than to say "we think these look nice and so do some other people, so try them for quick results." RedlogFilm is coded to a standard (based on the Cineon log curve) and is therefore easier to interpret, understand, and manipulate. I've found that it also gives the best interpretation of the captured grey scale by far of any of the gamma curves Red has supplied when used in a proper pipeline. But that doesn't make it appropriate for viewing directly, either during production or in post. It requires a processing pipeline and is therefore not as useful for set viewing purposes as the other curves.

i am hearing of people using red log film and going straight on the 3 way color corrector (lift gamma gain ) but . isn't is wrong using lift gamma and gain a log encode image (before making it suitable for video gamma space?)

I don't like to use terms like "right" and "wrong" with these kind of things. Having said that, if you look at my recent blog posts (mikemost.com) you'll see my reasoning for avoiding that kind of approach.

 

[Steve Sherrick]

Mike, just want to say that the time you take to share your extensive knowledge and experience is remarkable. The blog is a great resource. Thanks for doing it!

 

[Tom Lowe]

Best starting point for grade is REDLogFilm, REDColor2. If you're doing a more video style grade, use REDGamma2.

Graeme

I only use RedColor2 and RedGamma2.

 

[Steve Freebairn]

It's also important to remember when making the choice between REDLogFilm and REDGamma2 to ask yourself, whether you are the final step in the color process. If you're the last person in the chain to do color, then use either. If you're setting a look on set, then you want to use REDGamma2 so that later on, if the workflow calls for it you have the option of getting Log files.

 

[Nick Shaw]

Redlog using a Log to video Lut : doesn't this produce basically red gamma 2 ??

It can do if your LUT is a REDlogFilm to REDgamma2 LUT (I have built a 1D LUT which does exactly that). But that does not mean it is not a valid approach. On the contrary, grading REDlogFilm and viewing and rendering through a REDlogFilm to REDgamma 2 LUT gives a very different result, and far more grading flexibility, compared to grading REDgamma 2 directly.

 

[Graeme Nattress]

Correct Nick - placing the grading operations either before or after the viewing LUT produces different results. Some colourists I know prefer a "pre-lut-correction" (film-style grading) and others like a "post-lut-correction" (video style grading) and some use both pre-and-post-lut grading to achieve the look that they're going for. Each way is valid - pick the one that works for your workflow and desired look.

Graeme

 

[Nick Shaw]

Correct Nick - placing the grading operations either before or after the viewing LUT produces different results. Some colourists I know prefer a "pre-lut-correction" (film-style grading) and others like a "post-lut-correction" (video style grading) and some use both pre-and-post-lut grading to achieve the look that they're going for. Each way is valid - pick the one that works for your workflow and desired look.

Graeme

I would say that the main case for grading "pre-LUT" is that the LUT compresses the highlights, whereas REDlogFilm (or any other log coding) is a representation of the uncompressed linear light brightness of the photographed scene, including values considerably higher than the nominal scene white. Grading in log means having this full range available when grading and then using the LUT to compress the "over-brights" to within the range of your display device. If you grade something that has the highlights already compressed, then it can be harder to pull out additional highlight detail, or even just reduce the exposure, without turning your whites grey.

As Mike Most advocates (and I rarely disagree with his viewpoint) a "trim" grade after the LUT can be useful in some cases.

 

[Gabriele Turchi]

i second that Mike Most shared knowledge is very very very valuable ...

NICK: very very very very interesting approach ... i would love to know more ...

when i said that Redlogfilm + log to video lut produce red gamma 2 , i did not meant exactly that , but i meant that the benefit of log gets kind of lost without a custom made curve ... red gamma 2 is quite flat and WAY WAY far superior that red gamma for color correction ...so i am wondering (NICK ?) how and why grading redlogfilm for video delivery is superior

ps: i mean grading R3D , because otherwise i do know that rendering out DPX 10 bit before the DI, LOG is the best way to pack 12 bit into a 10 bit container ... but having access to the RAW (using 16bit decode in Resolve in my case ) ...?

i would love to know more from Nick on that way to do it ...

pss: i can totally see that a custom del-log curve on shot by sot base would be the very best , but i guess is a bit of more time consuming , and doe introduce more variables for hum a error i guess ...

thanks a lot

g

 

[Gabriele Turchi]

I would say that the main case for grading "pre-LUT" is that the LUT compresses the highlights, whereas REDlogFilm (or any other log coding) is a representation of the uncompressed linear light brightness of the photographed scene, including values considerably higher than the nominal scene white. Grading in log means having this full range available when grading and then using the LUT to compress the "over-brights"

but grading the 12 RAW (using 16 bit on the SDK in my case) : isn't the same amount of info we are talking about ? meaning that from red gamma2 ca me extracted the same info on the highlight that from Redlogfilm (equal or maybe even a bit lower ISO or FLUT ...)

i have the impression (and i might be wrong) that redo film for video target takes more to grade (let say for a nice normal proper exposed with image (so no crazy particular look ) than redgamma 2 , and i quite never noticed how red gamma 2 could destroy anything on the image (in the same way that redspace does , or red gamma (meaning clipping the highlight etc...)

g

 

[Nick Shaw]

when i said that Redlogfilm + log to video lut produce red gamma 2 , i did not meant exactly that , but i meant that the benefit of log gets kind of lost without a custom made curve ... red gamma 2 is quite flat and WAY WAY far superior that red gamma for color correction ...so i am wondering (NICK ?) how and why grading redlogfilm for video delivery is superior

See my reply above.

ps: i mean grading R3D , because otherwise i do know that rendering out DPX 10 bit before the DI, LOG is the best way to pack 12 bit into a 10 bit container ... but having access to the RAW (using 16bit decode in Resolve in my case ) ...?

Even grading on a system which can work directly with the RAW via the SDK, you still need to choose the decode parameters. If you decode as REDlogFilm and then add a LUT downstream of the grade, you still get more flexibility than decoding as REDgamma2 with no LUT.

By the way, while I agree that REDgamma2 is far more gradable than REDgamma, I personally prefer to use a REDlogFilm to REDgamma rather than REDgamma2 LUT. Since the grading is happening before the LUT, the more aggressive roll-off in the highlights and shadows is not a problem, and produces a result which many people like.

 

[Gabriele Turchi]

mmm.....very very interesting nick , i 'll give it a shot

ps:is your lut only applied as viewing Lut while grading , than baked IN at the end?

thanks

g

 

[Nick Shaw]

I have the impression (and i might be wrong) that redo film for video target takes more to grade (let say for a nice normal proper exposed with image (so no crazy particular look ) than redgamma 2 , and i quite never noticed how red gamma 2 could destroy anything on the image (in the same way that redspace does , or red gamma (meaning clipping the highlight etc...)

You are right that since REDgamma2 does not clip highlight or shadow detail, everything is still there for you to grade with. At the end of the day whatever gets you to the result you want is the right approach for you. No right or wrong. I just personally prefer the log + LUT approach, and have found that most colourists I talk to who are used to working with film scans find the same, because it responds to grading adjustments the same way a film scan does.

Note, I am not advocating using a film print emulation LUT for a video deliverable. As Mike Most has said a number of times before, doing this means adding unnecessary processing to counteract the effect of the film stock. My LUT is simply a 1D curve.

 

[Gabriele Turchi]

You are right that since REDgamma2 does not clip highlight or shadow detail, everything is still there for you to grade with. At the end of the day whatever gets you to the result you want is the right approach for you

well , sometimes we think that what we think is good is the best ...but there might be better ...and talking a bout RED (RAW), the approach and the starting point will produce different results ...

Note, I am not advocating using a film print emulation LUT for a video deliverable. As Mike Most has said a number of times before, doing this means adding unnecessary processing to counteract the effect of the film stock. My LUT is simply a 1D curve.

i am 100% with you on this one guys ... sometime it does works (especially exteriors towards magic hour light ...but mainly cause issue ...


g

 

[Graeme Nattress]

Even if you are referencing back to the R3D raw data, you still need to decide upon a linearity of the data and colour-space to grade in. You can't really do much in the way of grading on RAW data! So I'd still be picking REDLogFilm / REDColor2. What referencing back to the R3D does give you is easy ability to re-choose white balance, and a better guarantee of data precision.

Graeme

 

[Gabriele Turchi]

Using cameraRGB is fraught with danger and should be avoided. If you do want to use ACES, and I'd recommend you don't as it's buggy (not the RED implementation of the colour space, but the Academy's rendering process) get REDCine-X to export a EXR with the ACES space.

Graeme

Graeme, could please elaborate a bit on the usage of camera RGB ?
sometimes seems to me that does have nicer skin tone , but what i do notice is a less color separation on the elements of the image (let say compared to RC2) + HUE changes ...what is Camera RGB mean to be used for ? is the gamut larger than 709 ? so is not recommended for 709 color work ?

tahnks

g