FS7: FS7 matching AMIRA

Donny 123 said:
But isnt Slog3.cine ..purposely almost exactly the same as Arri Log C.. which is based on Cineon .. made for telecine by Kodak.. no need for a LUT..?
Click to expand...

Donny,

S-Log3 is a response curve or transfer function, ie the contrast response of the recording. S-Gamut3.cine is a colour gamut; it defines what 'red', 'green', 'blue' and 'white' represent in the recording relative to a base colourspace known as XYZ. S-Gamut3.cine is determined by four pairs of values known as the primaries and white point which are then converted into a 3x3 matrix. On it's own it has no fundamental connection to S-Log3 (that's why Sony can offer more than one option with S-Log3).

Every single log curve I know of at some point has been referred to as 'based on Cineon'. This is because, in the sense that Cineon mapped linear levels to a log-based response curve, they all are. Equally, all cars could arguably claim a connection to a Ford Model-T as they have wheels to drive on and one to steer with. The Cineon connection is convenient for marketing. Some log curves are closer to Cineon than others, and S-Log3, LogC, Canon C-Log2 and Panasonic V-Log have reasonably close responses, but closer to each other than to Cineon, which covers a somewhat narrower dynamic range than the newer ones are designed for.

As response curves, S-Log3, LogC, C-Log2 and V-Log are very close relatives; the largest difference is that V-Log places black at 10-bit 128, where the others just mentioned (and Cineon) place it at 10-bit 95.

That doesn't mean that a Cineon LUT won't work, just that it definitely won't look as it would on a Cineon clip of the same scene.

On the colour side, all of the manufacturers take different approaches to their wide-gamut choices. Alexa Wide Gamut has different primaries to S-Gamut3, to S-Gamut3.cine, to V-Gamut, to Canon Cinema Gamut and certainly to Rec709. Again, you can use a 3D LUT on a colourspace it isn't designed for, it just won't look the way it was intended. If it looks nice, then great, but for it to work 'properly' you need to slip in a conversion stage. That doesn't have to be difficult.

I've just noticed one other thing a few posts back:
But how does Canon C - Log resemble Arri..? Canon C log is like HG7/8..
Click to expand...

Canon C-Log is fundamentally rather different to a hypergamma. It is a true log curve, albeit one tuned to mitigate the huge limitations of 8-bit. On the colour side, however, CP Lock on the C300 mkI is rather a hodgepodge, tweaking Rec709 in a rather nonlinear way. This means that its principal corrections can be made with a 1D LUT (Canon supply a bunch) or just using basic correction tools. Again, this is basically because 8-bit is not great for wide-gamut recording, but it does make simple grading easier, at the expense of more advanced stuff (the Canon ACES IDT for CP Lock C-Log on the mkI C300 is a horror show compared to pretty much any other log recording camera).

The C300mkII opens up 10-bit recording, and with its wider dynamic range capability Canon have shifted towards the approach of S-Log3 and LogC with C-Log2. If you have a go with LUTCalc, all these log curves are calculated using manufacturers' parameters; there is a chart of IRE vs stop on the bottom right in which you can compare any of them - or hypergammas - in terms of response.

Ben
 
Last edited:
MichaelFischer.at said:
You just have to go into LutCalc and select S-Log3/S-Gamut3.cine to Arri LogC/Alexa Wide Gamut and off you go ;)

You can do this with SL2/SG, SL2/709 or Canon C-Log/709 as well!
Click to expand...
So where is a fault in my thinking? Why we even use bertLUT if we could just use two stage LUT's - one to get you from Slog3 to logC and then use a rec709 LUT that Arri provides. Shouldn't that give as almost exactly same results as an slog3 with bertLUT?
 
You don't need to go to logc. You can just apply the ARRI LogC -> REC709 directly to slog3. (This gives you around the same look as LC709 (non-A) applied with 0.02 lift and 1.02 gain settings in Resolve by the way). From there you can do a small bump in the lower tone of the blue channel and you'll get this slightly warmer look. Add a contrast tweak from between 1.2 to 1.4 and you should be good to go with a pretty great looking image.

(By the way Bert's LUT is not just the ARRI LogC -> REC709 LUT. It's a bit more than that.)
 
starcentral said:
You don't need to go to logc. You can just apply the ARRI LogC -> REC709 directly to slog3. (This gives you around the same look as LC709 (non-A) applied with 0.02 lift and 1.02 gain settings in Resolve by the way). From there you can do a small bump in the lower tone of the blue channel and you'll get this slightly warmer look. Add a contrast tweak from between 1.2 to 1.4 and you should be good to go with a pretty great looking image.
Click to expand...

An alternative you could try is to take a LUT designed for Arri LogC with Arri Wide Gamut colourspace (33x33x33, larger than that is not read by Arris, so may well be a linear interpolation), read it into LUTCalc (box on the bottom left, under 'LUTAnalyst'), set 'input gamma' in that box to LogC (Sup 3.x & 4.x), input gamut to 'Alexa Wide Gamut' and 'Camera' in the top left box to 'Arri Alexa / Amira' and click 'Analyse'.

You should then have the Arri LUTs gamma and gamut available as output options, effectively alohype's two stage suggestion rolled into one behind the scenes, using the actual Sony and Arri primaries and white point.

If you specifically load 'X-2-ALEXA', in the top left box where you have set the camera to 'Arri Alexa / Amira', temporarily set 'Recorded ISO' to 1600 before clicking 'Analyse'. The curve will be an almost perfect map to LC709A from 18% gray up, but with shadows scaling a bit lower (to 2% for black).

This is still an approximation, but it's simpler and should be closer to accurate than hand judgement,

Ben
 
One last thing,

if an Arri Rec709 look on an Arri LogC camera shooting Alexa Wide Gamut is like an artwork you'd like to have a print of for your wall but the limited run is all sold out, Dennis' suggestion is like a hand-drawn sketch and my LUTAnalyst approach like taking a photograph. Ideally, the best solution would be to get hold of the original plates and print your own 'perfect' repo.

As it happens, in the case of the original Alexa 709, Arri is actually nice enough to take us half way there. In a document titled 'ALEXA Log C Curve - Usage In VFX', they not only state that the Arri Rec709 process involves applying a 3x3 matrix to LogC/Wide Gamut footage which has first had a tone map applied to it (as opposed to purely linearised data), they actually give the matrix values.

I say they take us half way there - unfortunately they don't publish the tone map! What it isn't is just the final Rec709 curve. I've been working on an algorithm to figure out tone curves like this where the matrix is known, with the ultimate aim of being able to approximate the matrix values 'blind'. If I can crack that, I would have much, much smaller and potentially more flexible and accurate representations of some of the nonlinear colourspaces in LUTCalc which are currently stored as LUTs. It also makes these kind of colourspaces reversible.

Anyway, as Alexa Classic 709 is effectively LogC / Arri Wide Gamut -> Tone Map / Wide Gamut -> Tone Map / Alexa709 -> Alexa709 / Alexa709 where the only unknown is the 'Tone Map' bit, I'm 95% sure I've managed to pin that down. This is the 'Classic' Alexa 709, as titled 'X-2-ALEXA' on an Amira. I've not really intended to include it in LUTCalc so much as use it as a path to figuring out the more general case of tonemapped colourspaces.

That more general situation of Log / input gamut -> linear / input gamut -> linear / stage 1 matrix -> Tone map / stage 1 gamut -> Tone map / output gamut -> output gamma / output gamut - particularly where there is also a linear space matrix stage - is currently proving trickier!

Fun and games ;-)

Ben
 
cameramanben said:
Donny,

S-Log3 is a response curve or transfer function, ie the contrast response of the recording. S-Gamut3.cine is a colour gamut; it defines what 'red', 'green', 'blue' and 'white' represent in the recording relative to a base colourspace known as XYZ. S-Gamut3.cine is determined by four pairs of values known as the primaries and white point which are then converted into a 3x3 matrix. On it's own it has no fundamental connection to S-Log3 (that's why Sony can offer more than one option with S-Log3).

Every single log curve I know of at some point has been referred to as 'based on Cineon'. This is because, in the sense that Cineon mapped linear levels to a log-based response curve, they all are. Equally, all cars could arguably claim a connection to a Ford Model-T as they have wheels to drive on and one to steer with. The Cineon connection is convenient for marketing. Some log curves are closer to Cineon than others, and S-Log3, LogC, Canon C-Log2 and Panasonic V-Log have reasonably close responses, but closer to each other than to Cineon, which covers a somewhat narrower dynamic range than the newer ones are designed for.

As response curves, S-Log3, LogC, C-Log2 and V-Log are very close relatives; the largest difference is that V-Log places black at 10-bit 128, where the others just mentioned (and Cineon) place it at 10-bit 95.

That doesn't mean that a Cineon LUT won't work, just that it definitely won't look as it would on a Cineon clip of the same scene.

On the colour side, all of the manufacturers take different approaches to their wide-gamut choices. Alexa Wide Gamut has different primaries to S-Gamut3, to S-Gamut3.cine, to V-Gamut, to Canon Cinema Gamut and certainly to Rec709. Again, you can use a 3D LUT on a colourspace it isn't designed for, it just won't look the way it was intended. If it looks nice, then great, but for it to work 'properly' you need to slip in a conversion stage. That doesn't have to be difficult.

I've just noticed one other thing a few posts back:


Canon C-Log is fundamentally rather different to a hypergamma. It is a true log curve, albeit one tuned to mitigate the huge limitations of 8-bit. On the colour side, however, CP Lock on the C300 mkI is rather a hodgepodge, tweaking Rec709 in a rather nonlinear way. This means that its principal corrections can be made with a 1D LUT (Canon supply a bunch) or just using basic correction tools. Again, this is basically because 8-bit is not great for wide-gamut recording, but it does make simple grading easier, at the expense of more advanced stuff (the Canon ACES IDT for CP Lock C-Log on the mkI C300 is a horror show compared to pretty much any other log recording camera).

The C300mkII opens up 10-bit recording, and with its wider dynamic range capability Canon have shifted towards the approach of S-Log3 and LogC with C-Log2. If you have a go with LUTCalc, all these log curves are calculated using manufacturers' parameters; there is a chart of IRE vs stop on the bottom right in which you can compare any of them - or hypergammas - in terms of response.

Ben
Click to expand...

Thanks for the info Ben.. the Canon C log likeness to HG is not technical I guess .. just that it doesn't have the flat,washed out look of "true" log footage.. you could just air it as it is.. that you could never do with Slog..
 
Donny 123 said:
Thanks for the info Ben.. the Canon C log likeness to HG is not technical I guess .. just that it doesn't have the flat,washed out look of "true" log footage.. you could just air it as it is.. that you could never do with Slog..
Click to expand...

As Ben said, this has more to do with the color space then with the transfer function. C-Log uses the 709 color space and a more contrasty log curve. S-Log and LogC both also use a wide color space, that's why it looks so dull.
 
Yes but at the end of the day.. it looks very much like HG than Slog.. thats all Im saying.. what ever the process to get there..
 
When using Bert's "True-Balanced" on a two camera F5 shoot recently, we kept the bright skin highlights under 63 IRE.
 
David Kudell said:
What are you setting zebras on bright white using this LUT?
Click to expand...

Yes! That is the right question to ask. Setting exposure by skin tone is a bad habit that would only work if everyone was a clone. I happen to shoot whites, blacks, asians, and latinos on a regular basis. A number such as 63% is meaningless with such varied shades of skin. On the other hand, bright white is a reliable and consistent target that eliminates the guess work.
 
Doug Jensen said:
Yes! That is the right question to ask. Setting exposure by skin tone is a bad habit that would only work if everyone was a clone. I happen to shoot whites, blacks, asians, and latinos on a regular basis. A number such as 63% is meaningless with such varied shades of skin. On the other hand, bright white is a reliable and consistent target that eliminates the guess work.
Click to expand...

Doug, it's because I took your excellent FS7 course that I knew to ask the question! :)

I think I'm going to have to figure it out myself. I should theoretically be able to if I know that Slog3 whites should be 61. I just need to set camera at Slog3, point at a white card that's zebra-ing at 61, then apply the LUT in camera, point at same white card, and raise the zebra levels on the camera until they show up on the white card. Does that sound right?
 
You must log in or register to reply here.
Back
Top