F-Log2C is a colorists dream come true, will now have extreme grading options. Have tested the new firmware and can confirm that it resolves the auto focus issues.
Also the colors aren’t mudding together. Smooth color transition like B Raw. Fuji just brought more value to the Xh2s and this lets me know companies can do more with the cameras they have already released. They could have just saved this log for a newer camera but didn’t. I’ve been testing it against my bmcc full frame and I’m really blown away. They did their thing
Would be also interesting to see how greens and blues look like now, hopefully more accurate as well, but reds and oranges alone are already a great improvement, thanks for sharing
As a photographer, this reminded me when I was shooting with Velvia and all the reds were just pink, thanks for showing this, cause I had no idea about this whole FLOG2C update
Super helpful! Those bright reds were the worst. Curious if this works with colorspace transform or its only LUT right now. Excited to see the arri comparison too!
I always set my white balance -1 R for the reason of the magenta Fujifilm Cameras have and it definitely helps. I'm hoping F-Log2 C comes to the X-S20. Open Gate + F-Log2 C = AWESOME
With proper IDTs, I don't think the difference should be this huge, or even noticeable at all once transformed. It should only make a difference for extreme values (ie lasers, LEDs). I don't think Resolve has been updated to support the F-Log 2 C colorspace yet, though it would be possible to create your own IDTs using the values in their PDF. While I understand Fuji has released a "709" LUT for Flog2C, that is their subjective method of doing that image formation, which has likely changed in recent years due to their increasing push into cinema workflows, and thus the differences in results are likely largely due to that. Putting image formation philosophy aside, I believe FLog2C mostly comes down to making their 3x3 matrix math easier in the camera hardware, and less to do with any major changes to "look" (if there even is one) in that stage of the pipeline. Thanks for making a video on this topic though. :)
Yes I do understand what you're saying. In this case I didn't use the LUTs released by Fujifilm, it was done in Resolve. Either way I appreciate that this color gives you a much better starting point to work with than flog2.
Can confirm, while the differences are more subtle in IDTs, I've used a proper IDT based on Fujis datasheet in ACES for FLOG2C and on some shots it is night and day. All the colors are more accurate, the ACES IDT just seems to close the difference gap a little, but the FLOG 2 is still leagues behind. Shooting tail lights, neon, that's obvious, but same with flowers, and extremely complex mixed lighting or high noon / brutal overhead lights. I tested on all white balances including customs, shot it out against all modes, and the best surprise, not only is it the best of the in camera modes with the least chroma smoothing under the hood, basically no dead pixels even when changing modes back and forth all day on a low battery, and with the same rolling shutter as F LOG 2, but it's also to my eye and the scopes/waveforms superior to uncompressed CinemaDNG raw, using Assimilate Play Pro to extract 12 bit - it's not as good and doesn't have the same level of color detail in the skies. It's by far the best mode to shoot in and possibly the single thing that makes it like shooting on a Venice or a classic Varicam 35/Alexa. Hats off to Fuji, this is the coolest fucking shit.
@@area51pictures this is great to hear. I recently did a volunteer multicam shoot of a concert where every camera was a different brand (Several Sonys, two Fujis, and a few Canons)… and believe it or not they mostly matched bang on, except the red hues were weirdly bright on the Fuji. It was admittedly low CRI concert lighting, but still. I ended up solving it with a custom 3x3 matrix, using a MacBeth chart and my own python script that is akin to “mmColorTarget” in Nuke. It solved it perfectly, but it was certainly annoying and strange it was so far off. So I’m guessing this update fixes that. In fact I think that’s most of what Fuji Flog 2 C is about. It’s probably a working space closer to whatever sensor native is for Fuji’s recent cameras. I think BT.2020 is not the best as a working space out of camera, as it probably required some tricky “best fit” trades offs to reshape things, which is all much better done in post with a proper tonemapper these days when possible, especially when trying to match with other camera observers.
At 1:51 I can see the saturation change of your red bed sheet in the background. Even with the second example outside, look at the reds of the leaves. Is it looking more like Eterna maybe?
Those were great examples. I’m looking forward to seeing more of the footage you take. I always find myself having to tone down the red, so this new profile looks extremely promising
Hi Tom, In your comments you state that you are using in Davinci Resolve CST Arri Logc base. Which one? Arri LogC3 or Arri LogC4? My guess from my trials LogC3. Is this correct? Don
Hi Tom, which Arri Wide Gamut did you use, 3 or 4? I’ve been experimenting with flog2c using a CST with the following parameters: AWG4/ArriLogC3 -> Rec.709 Gamma 2.2. Is this similar to what you did here? My results look really good with better color representation and better color density with some specific colors like red, green, and orange (I haven’t tested footage with other prominent colors). Cheers
Do we know if F-Log 2C fixes the red channel downsampling issue Fuji has always had? Been with them for years and was never a huge fan of their approach with it
@@OzanTanOh sorry! Typically I expose about 1 stop over but depends on the scene. Always make sure your skintones are exposed well if you’re filming people. I’ve found the xh2s recovers highlights very well, so I always prefer over exposing a tad Its native iso is 1250
@@TlarkArts I guess I need to use exposure compensation to expose 1 stop over, but I can't change the exposure compensation when shooting in manual mode. How can I expose 1 stop over, I would be very happy if you could help me.
How did you convert it? Is there already a conversion lut? It seems like a lot of this could depend on how you convert and can be achieved with Flog2 as well.
All footage was converted with a CST flog2 to rec709 using davinci resolve and the arri logc as the base. This has always been my first step for converting flog2 to rec709 for all of my xh2s files. Everything you see in the video only has a conversion lut to get it to rec709 with no other adjustments.
Also - I was informed Fujifilm did release flog2C > 709 LUTs as well. I just tested with the official Fujifilm > 709 LUTs and results are pretty much the same. For my next test with the ARRI I will show that in further detail. Good question though, I should have clarified this in the video.
@@t_c Ok, because if you use the regular FLOG2 conversion lut which is designed for a smaller color space, the results should be totally off, if I understand correctly. I personally am quite new to color grading but have an understanding of color spaces from photography. For my tests with F-Log2 I have used Fujis LUTs but prefer color spaces transformation in Resolve so I am curious what kind of results are correct and/or possible.
@@jayzn1931 According to my test with the official flog2c LUT, there is a difference but it's not much. I also prefer the CST in resolve, that is what I used in this video.
@ yes obviously, but flog is only read out from the sensor at 12bit, so I’m assuming fLog2c is still being read out at 14bits like flog2 standard. Only seems reasonable to assume.
Of course, however this is a preliminary test as I mentioned. This was a very quick and dirty test that I was doing ahead of my shoot with the camera tomorrow to see if it was a viable solution. Thanks for watching!
F-Log2C is a colorists dream come true, will now have extreme grading options. Have tested the new firmware and can confirm that it resolves the auto focus issues.
So far loving the new update!
Fuji marketing enters the chat
Love the video dude
@ thank you!
Also the colors aren’t mudding together. Smooth color transition like B Raw. Fuji just brought more value to the Xh2s and this lets me know companies can do more with the cameras they have already released. They could have just saved this log for a newer camera but didn’t. I’ve been testing it against my bmcc full frame and I’m really blown away. They did their thing
Agreed!
Are you saying its better than the footage from the bmcc? Although to be precise B Raw is not a log profile but the codec...
WOW! Really happy with this update. Honestly such a gamechanger and in time for a shoot next week. Thanks for the great video:)
Looking forward to Arri - Fuji comparison. Great vid!
Thank you!
Would be also interesting to see how greens and blues look like now, hopefully more accurate as well, but reds and oranges alone are already a great improvement, thanks for sharing
Good to note, I will test in the next one!
As a photographer, this reminded me when I was shooting with Velvia and all the reds were just pink, thanks for showing this, cause I had no idea about this whole FLOG2C update
Glad to help!
this update has made me so happy
Super helpful! Those bright reds were the worst. Curious if this works with colorspace transform or its only LUT right now. Excited to see the arri comparison too!
Fujifilm released an Flog2C LUT, I tested briefly, looks pretty good. I will have to test that further in the next video. Thanks!
Nice breakdown, Thomas! Interesting to see the examples 👍
@@wearetrackclub thank you!
Great one. subbing to see tests against Arri. I wonder if you can idt to aces or even dwg already.
Thank you, I will keep you posted with that.
I always set my white balance -1 R for the reason of the magenta Fujifilm Cameras have and it definitely helps. I'm hoping F-Log2 C comes to the X-S20. Open Gate + F-Log2 C = AWESOME
With proper IDTs, I don't think the difference should be this huge, or even noticeable at all once transformed. It should only make a difference for extreme values (ie lasers, LEDs). I don't think Resolve has been updated to support the F-Log 2 C colorspace yet, though it would be possible to create your own IDTs using the values in their PDF. While I understand Fuji has released a "709" LUT for Flog2C, that is their subjective method of doing that image formation, which has likely changed in recent years due to their increasing push into cinema workflows, and thus the differences in results are likely largely due to that. Putting image formation philosophy aside, I believe FLog2C mostly comes down to making their 3x3 matrix math easier in the camera hardware, and less to do with any major changes to "look" (if there even is one) in that stage of the pipeline. Thanks for making a video on this topic though. :)
Yes I do understand what you're saying. In this case I didn't use the LUTs released by Fujifilm, it was done in Resolve.
Either way I appreciate that this color gives you a much better starting point to work with than flog2.
Can confirm, while the differences are more subtle in IDTs, I've used a proper IDT based on Fujis datasheet in ACES for FLOG2C and on some shots it is night and day. All the colors are more accurate, the ACES IDT just seems to close the difference gap a little, but the FLOG 2 is still leagues behind. Shooting tail lights, neon, that's obvious, but same with flowers, and extremely complex mixed lighting or high noon / brutal overhead lights. I tested on all white balances including customs, shot it out against all modes, and the best surprise, not only is it the best of the in camera modes with the least chroma smoothing under the hood, basically no dead pixels even when changing modes back and forth all day on a low battery, and with the same rolling shutter as F LOG 2, but it's also to my eye and the scopes/waveforms superior to uncompressed CinemaDNG raw, using Assimilate Play Pro to extract 12 bit - it's not as good and doesn't have the same level of color detail in the skies. It's by far the best mode to shoot in and possibly the single thing that makes it like shooting on a Venice or a classic Varicam 35/Alexa. Hats off to Fuji, this is the coolest fucking shit.
@@area51pictures this is great to hear. I recently did a volunteer multicam shoot of a concert where every camera was a different brand (Several Sonys, two Fujis, and a few Canons)… and believe it or not they mostly matched bang on, except the red hues were weirdly bright on the Fuji. It was admittedly low CRI concert lighting, but still.
I ended up solving it with a custom 3x3 matrix, using a MacBeth chart and my own python script that is akin to “mmColorTarget” in Nuke. It solved it perfectly, but it was certainly annoying and strange it was so far off. So I’m guessing this update fixes that.
In fact I think that’s most of what Fuji Flog 2 C is about. It’s probably a working space closer to whatever sensor native is for Fuji’s recent cameras. I think BT.2020 is not the best as a working space out of camera, as it probably required some tricky “best fit” trades offs to reshape things, which is all much better done in post with a proper tonemapper these days when possible, especially when trying to match with other camera observers.
@@t_cwhich IDT did you use in Resolve?
DEFINE_ACES_PARAM(IS_PARAMETRIC_ACES_TRANSFORM: 0)
/**
* F-log2C to ACES AP0 conversion.
* Goes from F-Log2C and F-Gamut C gamut
* to scene linear, ACES AP0 gamut.
*
* This can be used as an IDT in Resolve.
*/
__DEVICE__ inline float flog2c_to_linear(float t)
{
const float a = 5.555556f;
const float b = 0.064829f;
const float c = 0.245281f;
const float d = 0.384316f;
const float e = 8.799461f;
const float f = 0.092864f;
const float cut1 = 0.000889f; // Should be equal to ((cut2 - f) / e)
const float cut2 = 0.100686685370811f; // should be equal to (e * cut1 + f)
if (t >= cut2) {
return _exp10f((t - d) / c) / a - b / a;
} else {
return ((t - f) / e);
}
}
// Used values calculated from color.support/colorspacecalculator.html
// source color space F-Gamut C and Output ACES AP0, with CAT02 chromatic adaptation.
__CONSTANT__ float matrix[9] = {
0.840895 , 0.027528 , 0.131577,
0.000920 , 1.007390 ,-0.008310,
-0.000560 ,-0.000776 , 1.001336
};
__DEVICE__ float3 transform(int p_Width, int p_Height, int p_X, int p_Y, float p_R, float p_G, float p_B)
{
float r = flog2c_to_linear(p_R);
float g = flog2c_to_linear(p_G);
float b = flog2c_to_linear(p_B);
float r2 = ( matrix[0] * r) + ( matrix[1] * g) + ( matrix[2] * b);
float g2 = ( matrix[3] * r) + ( matrix[4] * g) + ( matrix[5] * b);
float b2 = ( matrix[6] * r) + ( matrix[7] * g) + ( matrix[8] * b);
float3 res = to_float3(r2, g2, b2);
return res;
}
At 1:51 I can see the saturation change of your red bed sheet in the background. Even with the second example outside, look at the reds of the leaves. Is it looking more like Eterna maybe?
The way your shirt went from purple to blue!
Haha yes i noticed that after I filmed the video. Crazy!
Those were great examples.
I’m looking forward to seeing more of the footage you take.
I always find myself having to tone down the red, so this new profile looks extremely promising
Thank you, yes it seems like they controlled everything much better with this.
Good vid! To me, FLog2C looks like if you were to drag down the Red channel saturation and luminance sliders in Lightroom a bit.
Yes exactly, that's how my correction LUT works for flog2. However flog2c seems to be a lot more accurate in terms of capturing the proper hue.
Hi Tom, In your comments you state that you are using in Davinci Resolve CST Arri Logc base. Which one? Arri LogC3 or Arri LogC4? My guess from my trials LogC3. Is this correct? Don
I want to know this as well!
Yes LogC3!
@@Mikeaguado LogC3!
@@t_c Thank you!
Hi Tom, which Arri Wide Gamut did you use, 3 or 4? I’ve been experimenting with flog2c using a CST with the following parameters: AWG4/ArriLogC3 -> Rec.709 Gamma 2.2. Is this similar to what you did here? My results look really good with better color representation and better color density with some specific colors like red, green, and orange (I haven’t tested footage with other prominent colors). Cheers
Super interesting!!
Hey thanks for stopping by! 🤠
Do we know if F-Log 2C fixes the red channel downsampling issue Fuji has always had? Been with them for years and was never a huge fan of their approach with it
Thanks for making this I subscribed and I’m looking forward to seeing the Arri comparison test
Thanks! Hopefully I can get some testing done soon.
Hi. Subscribed and waiting to see future updates and comparison especially Davinci workflow for this new log format
Sure thing, I can include that in the next test. Thanks!
Have you tried converting using a color space transform since this color space doesn't exist yet in Davinci Resolve
@@Mrcheesydancer not yet! Next test I will try that out
Great test. I noticed a lot of the colors seem to be desaturated which is great.
Yes, a little more wiggle room for coloring in post which is nice.
Never realized that Flog2 had such a bad problem with the reds... Is that the same with Flog?
yes, F-log always had this problem because of BT.603 matrix . But, HLG is accurate
They finally fixed their pink skin tones, wow!
they fixed it in HLG years ago
I think Lumix V-log is suffering similarly with the red tones
First of all thank you for the video. What is the recommended iso to get less noise when shooting with flog2c?
It's basically idential to flog2 in terms of how you should expose. Native ISO remains the same.
@@t_c Then i'm asking the same question for flog2, thanks.
@@OzanTanOh sorry! Typically I expose about 1 stop over but depends on the scene. Always make sure your skintones are exposed well if you’re filming people. I’ve found the xh2s recovers highlights very well, so I always prefer over exposing a tad
Its native iso is 1250
@@TlarkArts I guess I need to use exposure compensation to expose 1 stop over, but I can't change the exposure compensation when shooting in manual mode. How can I expose 1 stop over, I would be very happy if you could help me.
Will it be appropriate to use with ‘standard’ flog2 davinci colour management?
Not sure yet, may be too soon to tell. I saw Fujifilm released an flog2c LUT though.
How did you convert it? Is there already a conversion lut? It seems like a lot of this could depend on how you convert and can be achieved with Flog2 as well.
All footage was converted with a CST flog2 to rec709 using davinci resolve and the arri logc as the base. This has always been my first step for converting flog2 to rec709 for all of my xh2s files.
Everything you see in the video only has a conversion lut to get it to rec709 with no other adjustments.
Also - I was informed Fujifilm did release flog2C > 709 LUTs as well. I just tested with the official Fujifilm > 709 LUTs and results are pretty much the same. For my next test with the ARRI I will show that in further detail. Good question though, I should have clarified this in the video.
@@t_c Ok, because if you use the regular FLOG2 conversion lut which is designed for a smaller color space, the results should be totally off, if I understand correctly. I personally am quite new to color grading but have an understanding of color spaces from photography. For my tests with F-Log2 I have used Fujis LUTs but prefer color spaces transformation in Resolve so I am curious what kind of results are correct and/or possible.
@@jayzn1931 According to my test with the official flog2c LUT, there is a difference but it's not much. I also prefer the CST in resolve, that is what I used in this video.
Is rec709 better than rec2020?
Is fLog2c still being read out at 14bit?
@@DanielTroyer yes! Think of it as a picture profile.
@ yes obviously, but flog is only read out from the sensor at 12bit, so I’m assuming fLog2c is still being read out at 14bits like flog2 standard. Only seems reasonable to assume.
That's safe to assume, otherwise we would see a difference in DR, and I didn't notice any.
I noticed the M Coupe you were holding and hoped you had one. It didn’t take long to see you do. Haha.
I hope you don’t sell it.
Don’t worry I’ll yell at him if he sells it
Yes! I still have it, needs work! Thanks for watching.
oh wow that lamp was PINK pink thats wild
I never knew how bad we had it before! Haha but that is an extreme use-case. In real life the lamp is pretty vibrant lol
It looks like colors have more density in 2C
that's how fuji is describing it
considerably more, even more than the uncompressed DNG raw extracted using Assimilate from the atomos shogun connect
Sorry but a real test should be done with Da Vinci Resolve and color wheels to identify how the new profil give possibilities with colors.
Of course, however this is a preliminary test as I mentioned. This was a very quick and dirty test that I was doing ahead of my shoot with the camera tomorrow to see if it was a viable solution. Thanks for watching!
lamp test tells it all homie
My god I had no idea how bad Fuji is
hey dad. 🧪
Son?
@@t_c missed u dad