Logarist Color Correction for DaVinci Resolve, Vegas Pro, Final Cut Pro X

[joe12south]

I suspect that it's not the same as a 14-bit RAW image, but the Logarist color space does seem a lot better with 8-bit and 10-bit footage than anything else I've worked with.

I'm guessing that higher precision math is part of why Logarist works better, and maybe why it's slower on some computers.

The precision of the math use to make adjustments is a function of the NLE. As long as the color space is capable of holding all of the desired colors, Resolve is working with MORE than enough precision for any reasonable transformation.

 

[balazer]

- The NLE is working at high precision (at least 16bit) and a "wide" internal gamut. How do these LUT's make doing transforms "better"?
- If what you're doing is transforming to a different color space, then isn't knowing the input camera's color space important? For example, the GH4 can use either sRGB or aRGB. Picture profile settings like saturation don't change the color space...why does the latter matter but not the former?
- How is this easier? It's a minimum of 2 extra steps in FCPX for every single clip. For what gain?
- I don't understand how this would help rescue white balance any better than you can without it? Certainly not like Lightroom (or Premiere for that matter) does with a RAW file. The color error is baked into the non-raw file, and is going to be destructive no matter what.
- How is this different than the standardized LOG LUT's you released earlier?

Sorry if I sound skeptical, but the language around this "LUT system" is very opaque, and we've all been bitten by LUT's that promise to be miracle cures. I'm sincerely trying to understand why it's worth the extra work. If there are real, meaningful benefits, it would be great to know how to capitalize on them.

How do these LUTs make color correction better? Here's a concrete example. Say you underexposed a shot by 2/3 of a stop, by which I mean that if the ISO setting or the exposure would have been 2/3 of a stop higher, the shot would have turned out just how you wanted. Using normal color correction tools, only a specially shaped curve could fully correct that. Using lift/gamma/gain or some other shadow/mid/highlight tools you could come pretty close, except in the highlights near the clipping point. In Logarist, you just drag the offset slider up until the image is as bright as you want. It's a simple one-dimensional control, instead of a complex curve or separate low/mid/high adjustments. The result in Logarist will be nearly identical to increasing the ISO setting in the camera.

To correct white balance with traditional tools, you need to correct the shadows, midtones, and highlights separately if you want to achieve the same color balance across the whole image. In Logarist, it's just one color control for the entire image.

Color is baked into a non-raw file. But when you know how the camera did its transformation from raw sensor channel values to color values, that transformation can be inverted, taking the image from colors back to the raw domain. That's what I do here, though not all the way back to raw. I transform the camera's color values back to scene referred linear color, which is one step removed from raw sensor channel values. Of course the end-to-end transformation is not lossless. It's not the same as having a raw image. It's been de-bayered, sharpened, noise reduced, clipped, and quantized, and compressed. But that doesn't mean it's bad. A 10-bit uncompressed camera log recording isn't raw, but anyone would be hard pressed to tell the difference. In terms of color correction, a high precision log space with an extended range of highlights gives you nearly everything that the raw image had.

To use still photography as an analogy, would you shoot JPEG and bring that into Photoshop for color correction? It's far better to shoot raw, use Adobe Camera Raw as your first pass, and then make additional changes in Photoshop. Logarist isn't quite the same as shooting in raw, but it's closer.

Knowing the camera's color space is important in Logarist, and that's why I have dozens of different input transforms covering as many cameras as I could. On the GH4, the sRGB / Adobe RGB setting only affects still images. The color space of the GH4's video output is determined by the Photo Style setting, including all of the color and contrast adjustments.

Logarist is similar to my previous log LUTs, but this time I have explicit support for several different video apps, lots more cameras, and I've developed new methods for handling highlights.

In FCPX, after making corrections to one clip, you can save the effects as a preset so they can be applied to more clips in a single step. So the extra work is really just for the first clip, and after that it's less work.

No one is twisting your arm here. If your style of shooting lets you get the color and exposure almost perfect in the camera and you already have LUTs that give you the look you want, you're not going to benefit much from more advanced color correction tools. If you're comfortable working the way you have been, by all means keep doing it. I can't convert all the skeptics. I just know that for me, this new way of working has proven much better, and I wanted to share that with people.

 

[balazer]

So far I'm really impressed with how this handles V-log from the GH4. Cine-D seems to clip much easier than it does with manual grading or when using Leeming LUT. I know the shots are not clipped because I can recover them and other luts do not show this.

Are there recommended exposure levels you suggest for Cine-D to avoid this?

Same but opposite issue with V-log. When using these luts my shots get a lot darker than they do with the Leeming lut, Panasonic lut or my own manual grading. Are their recorded exposure levels that are different from normal V-log luts to shoot with or is it natural to be that much darker and I grade in between?

Finally why do you suggest only adjusting the global exposure setting in the FCPX color board and not the highs and shadows? I find this to be very limiting in grading by not tweaking the highs and shadows.

By the way this lut is extremely slow on my 2012 i7 iMac with 24GB ram and a 2GB GTX680m GPU. Not slow in terms of playback but FCPX really locks up for a good 20 seconds when I apply the first lut. I assume there is a lot going on in the background vs other luts to make it run that slow.

Logarist isn't trying to match the highlight range of your source to the highlight range of the output. It's mapping the middle grey level. So if you just add the Logarist LUTs and do nothing else, you are likely to see clipping, and it may well be lighter or darker than you were used to seeing with other LUTs. You'll pretty much always want to apply some exposure compensation in Logarist. That's the first and most basic correction to apply.

How to set the exposure in the camera when shooting for Logarist is totally up to you. Whatever you decide, you can compensate in Logarist. So the exposure setting really comes down to the fundamentals of dynamic range:

• Increasing the exposure decreases the highlight range (lowers the clipping point)
• Decreasing the exposure increases noise

These are the two fundamental forces that you must decide how to balance. There is no single right or wrong answer. You must decide how important noise and highlight range are to you.

The Logarist LUTs are all high precision, with a minimum of 262,144 points. That's probably why they take so long to load in mLut. On my test machine, a 2012-era Macbook Air md232ll/a, it took about 10 seconds to load a LUT in mLut. After that, corrections were near real-time on the quarter-screen preview.

In FCPX, my instructions say to use the Global color and exposure control, because those act as offsets, which is mathematically how you make corrections in a log space that will be equal across the entire color range. The Shadows, Midtones, and Highlights adjustments won't work in the log space the way you are used to. You can add another Color Correction effect after the second mLut effect, and make your shadow/mid/highlight adjustments in the display color space just like you always did before.

 

[broughtonfilm]

What about the LUT's used on set? Where in this workflow does it get applied? After all I'd like the image to look the same as when monitoring on set. Having to bring them there from what Logarist pair of LUT's creates as a starting point is another step in the workflow. Or am I missing something?

 

[balazer]

What about the LUT's used on set? Where in this workflow does it get applied? After all I'd like the image to look the same as when monitoring on set. Having to bring them there from what Logarist pair of LUT's creates as a starting point is another step in the workflow. Or am I missing something?

If you want a monitor LUT for on-set monitoring that will match what Logarist produces, there's a way to create it with Resolve.

Put the Logarist input transform on one corrector node, the output transform one another corrector node, and then put whatever baseline corrections you normally apply in between. For example, if your preferred workflow is to over-expose by a stop in the camera and then bring it down by a stop in Logarist, you will need to apply one stop of negative exposure compensation in between the two Logarist LUTs. Then export the entire correction chain as a new LUT file. You can load that LUT in your monitor.

 

[Thomas Smet]

Logarist isn't trying to match the highlight range of your source to the highlight range of the output. It's mapping the middle grey level. So if you just add the Logarist LUTs and do nothing else, you are likely to see clipping, and it may well be lighter or darker than you were used to seeing with other LUTs. You'll pretty much always want to apply some exposure compensation in Logarist. That's the first and most basic correction to apply.

How to set the exposure in the camera when shooting for Logarist is totally up to you. Whatever you decide, you can compensate in Logarist. So the exposure setting really comes down to the fundamentals of dynamic range:

• Increasing the exposure decreases the highlight range (lowers the clipping point)
• Decreasing the exposure increases noise

These are the two fundamental forces that you must decide how to balance. There is no single right or wrong answer. You must decide how important noise and highlight range are to you.

The Logarist LUTs are all high precision, with a minimum of 262,144 points. That's probably why they take so long to load in mLut. On my test machine, a 2012-era Macbook Air md232ll/a, it took about 10 seconds to load a LUT in mLut. After that, corrections were near real-time on the quarter-screen preview.

In FCPX, my instructions say to use the Global color and exposure control, because those act as offsets, which is mathematically how you make corrections in a log space that will be equal across the entire color range. The Shadows, Midtones, and Highlights adjustments won't work in the log space the way you are used to. You can add another Color Correction effect after the second mLut effect, and make your shadow/mid/highlight adjustments in the display color space just like you always did before.

Awesome stuff. Thanks for the explanation.

 

[Thomas Smet]

How do these LUTs make color correction better? Here's a concrete example. Say you underexposed a shot by 2/3 of a stop, by which I mean that if the ISO setting or the exposure would have been 2/3 of a stop higher, the shot would have turned out just how you wanted. Using normal color correction tools, only a specially shaped curve could fully correct that. Using lift/gamma/gain or some other shadow/mid/highlight tools you could come pretty close, except in the highlights near the clipping point. In Logarist, you just drag the offset slider up until the image is as bright as you want. It's a simple one-dimensional control, instead of a complex curve or separate low/mid/high adjustments. The result in Logarist will be nearly identical to increasing the ISO setting in the camera.

To correct white balance with traditional tools, you need to correct the shadows, midtones, and highlights separately if you want to achieve the same color balance across the whole image. In Logarist, it's just one color control for the entire image.

Color is baked into a non-raw file. But when you know how the camera did its transformation from raw sensor channel values to color values, that transformation can be inverted, taking the image from colors back to the raw domain. That's what I do here, though not all the way back to raw. I transform the camera's color values back to scene referred linear color, which is one step removed from raw sensor channel values. Of course the end-to-end transformation is not lossless. It's not the same as having a raw image. It's been de-bayered, sharpened, noise reduced, clipped, and quantized, and compressed. But that doesn't mean it's bad. A 10-bit uncompressed camera log recording isn't raw, but anyone would be hard pressed to tell the difference. In terms of color correction, a high precision log space with an extended range of highlights gives you nearly everything that the raw image had.

To use still photography as an analogy, would you shoot JPEG and bring that into Photoshop for color correction? It's far better to shoot raw, use Adobe Camera Raw as your first pass, and then make additional changes in Photoshop. Logarist isn't quite the same as shooting in raw, but it's closer.

Knowing the camera's color space is important in Logarist, and that's why I have dozens of different input transforms covering as many cameras as I could. On the GH4, the sRGB / Adobe RGB setting only affects still images. The color space of the GH4's video output is determined by the Photo Style setting, including all of the color and contrast adjustments.

Logarist is similar to my previous log LUTs, but this time I have explicit support for several different video apps, lots more cameras, and I've developed new methods for handling highlights.

In FCPX, after making corrections to one clip, you can save the effects as a preset so they can be applied to more clips in a single step. So the extra work is really just for the first clip, and after that it's less work.

No one is twisting your arm here. If your style of shooting lets you get the color and exposure almost perfect in the camera and you already have LUTs that give you the look you want, you're not going to benefit much from more advanced color correction tools. If you're comfortable working the way you have been, by all means keep doing it. I can't convert all the skeptics. I just know that for me, this new way of working has proven much better, and I wanted to share that with people.

Completely agree about 10bit log getting very close to raw. The famous color engineer Charles Poynton even has stated that 10 bit log has about 99% of the useful information from 12bit raw.

 

[broughtonfilm]

If you want a monitor LUT for on-set monitoring that will match what Logarist produces, there's a way to create it with Resolve.

Put the Logarist input transform on one corrector node, the output transform one another corrector node, and then put whatever baseline corrections you normally apply in between. For example, if your preferred workflow is to over-expose by a stop in the camera and then bring it down by a stop in Logarist, you will need to apply one stop of negative exposure compensation in between the two Logarist LUTs. Then export the entire correction chain as a new LUT file. You can load that LUT in your monitor.

Thank you for your answer. Yeah, that scenario is clear to me but I was talking about using existing LUT's, let's say I have LUT's which I have been using and would like to continue using them. I guess there's a way to translate them into this new colorspace and then generating a new LUT for use on set would be straightforward but I can't quite wrap my head around how I would do translation part of it. BTW, I ask these questions without any sarcasm, I'm intrigued by this workflow (as well as ACES, which is something I have to read more about).

 

[Thomas Smet]

I once tried combining luts in Resolve to bake them together for a friend with a Flame and he said it didn't work. I only have a Ninja Blade which has built in crude luts and no ability to load custom luts so I haven't been able to test that workflow myself.

 

[balazer]

Thank you for your answer. Yeah, that scenario is clear to me but I was talking about using existing LUT's, let's say I have LUT's which I have been using and would like to continue using them. I guess there's a way to translate them into this new colorspace and then generating a new LUT for use on set would be straightforward but I can't quite wrap my head around how I would do translation part of it. BTW, I ask these questions without any sarcasm, I'm intrigued by this workflow (as well as ACES, which is something I have to read more about).

If you have existing LUTs that you want to integrate with Logarist, it really depends on what that LUT was doing. LUTs can do three things, broadly:

• change the colors
• change the color space
• display rendering

If you have a "look" LUT that you apply to a display-ready image, you can still use that with Logarist. Just apply the LUT after the Logarist-to-BT.709 LUT. Such a LUT operates in the display color space and is only changing the colors.

If you have a LUT that goes from a camera log space to a display color space, it's doing color space transformation, display rendering, and maybe also changing the colors. There's no good way to integrate that kind of LUT into the Logarist workflow. Logarist depends on having an extended highlight range and scene-referred color. Once you're in the display color space, you lose those.

 

[broughtonfilm]

If you have a LUT that goes from a camera log space to a display color space, it's doing color space transformation, display rendering, and maybe also changing the colors. There's no good way to integrate that kind of LUT into the Logarist workflow. Logarist depends on having an extended highlight range and scene-referred color. Once you're in the display color space, you lose those.

This is exactly the case and what my worry was when I asked the question. I guess one thing that would solve the problem would be a special Logarist LUT replacing the existing BT.709 LUT that you provided (second of the pair) which would reverse the first Logarist LUT into for example Sony LC709 instead of BT.709. In other words, one could still get benefits of Logarist without having to use the result of BT.709 reverse LUT as a starting point.

 

[balazer]

The problem with all those camera log to display LUTs is that they do nothing to help with the color correction process. The camera log space and the display color space are both really sub-optimal for color correction. There's nothing magic about the look of the Sony LC-709 or LC-709TypeA LUTs. They're just BT.709-ish, with reduced contrast and saturation. They're really only good for previewing. I think you'll forget all about them after you get used to Logarist.

 

[broughtonfilm]

The problem with all those camera log to display LUTs is that they do nothing to help with the color correction process. The camera log space and the display color space are both really sub-optimal for color correction. There's nothing magic about the look of the Sony LC-709 or LC-709TypeA LUTs. They're just BT.709-ish, with reduced contrast and saturation. They're really only good for previewing. I think you'll forget all about them after you get used to Logarist.

I'm actually testing it now. I guess I'm probably used to my usual workflow which uses one of the Sony LUTs you mentioned and need to test Logarist more extensively but so far I can't get the highlights to have a gentle rolloff using Logarist, not without additional adjustments and still can't get them to look as good as in my usual workflow. They also tend to look oversaturated. Not sure why but noise seems to be more visible despite a more contrasty look of the base starting point as compared with a straight LC709A LUT. I do see how adjusting exposure with the offset wheel works linearly. This will be especially an advantage when dealing with Slog 2 footage which requires lots individual lift/gamma/gain adjustments, more so than Slog 3.

 

[joe12south]

How do these LUTs make color correction better? Here's a concrete example. Say you underexposed a shot by 2/3 of a stop, by which I mean that if the ISO setting or the exposure would have been 2/3 of a stop higher, the shot would have turned out just how you wanted. Using normal color correction tools, only a specially shaped curve could fully correct that. Using lift/gamma/gain or some other shadow/mid/highlight tools you could come pretty close, except in the highlights near the clipping point. In Logarist, you just drag the offset slider up until the image is as bright as you want. It's a simple one-dimensional control, instead of a complex curve or separate low/mid/high adjustments. The result in Logarist will be nearly identical to increasing the ISO setting in the camera.

How does this differ from the exposure control in Resolve or Lumetri?

To correct white balance with traditional tools, you need to correct the shadows, midtones, and highlights separately if you want to achieve the same color balance across the whole image. In Logarist, it's just one color control for the entire image.

Same question?

Color is baked into a non-raw file. But when you know how the camera did its transformation from raw sensor channel values to color values, that transformation can be inverted, taking the image from colors back to the raw domain. That's what I do here, though not all the way back to raw. I transform the camera's color values back to scene referred linear color, which is one step removed from raw sensor channel values. Of course the end-to-end transformation is not lossless. It's not the same as having a raw image. It's been de-bayered, sharpened, noise reduced, clipped, and quantized, and compressed. But that doesn't mean it's bad.

But it actually is kind of bad. As you state, we're STARTING with a mangled 8-bit image. The majority of the data has been thrown away, it can't be recovered. (It can be synthetically reconstructed to varying degrees of perceptual success, but that's not what you're suggesting.)

A 10-bit uncompressed camera log recording isn't raw, but anyone would be hard pressed to tell the difference. In terms of color correction, a high precision log space with an extended range of highlights gives you nearly everything that the raw image had.

No doubt, but again, that's not what we're starting from with these cameras. And if we were, we'd just work in that space, not yours.

To use still photography as an analogy, would you shoot JPEG and bring that into Photoshop for color correction? It's far better to shoot raw, use Adobe Camera Raw as your first pass, and then make additional changes in Photoshop. Logarist isn't quite the same as shooting in raw, but it's closer.

To extend your analogy, you seem to be suggesting that one could take that compromised JPEG, transform it into a different color space, and somehow get back to the RAW image. You can't. Too much has already been lost.

No one is twisting your arm here. If your style of shooting lets you get the color and exposure almost perfect in the camera and you already have LUTs that give you the look you want, you're not going to benefit much from more advanced color correction tools. If you're comfortable working the way you have been, by all means keep doing it. I can't convert all the skeptics. I just know that for me, this new way of working has proven much better, and I wanted to share that with people.

I want to understand what real world benefits I could gain from this workflow. I think there are some minor benefits...i.e. certain transforms are more straightforward depending on the NLE...but you're clouding those benefits by using language that seems to claim much more. Maybe that's just an unfortunately byproduct of trying to keep the explanation simple, but it does make the skeptic in me come out.

Here's a real world example that will work with a RAW file, and will fail with Logarist:

1. Use a camera supported by Logarist and that can shoot a RAW still,
2. Set-up a color chart under low tungsten light. (Necessitating an ISO high enough to introduce noise.)
3. Underexpose by a stop
4. Set your white balance to daylight.

(I use this example because I can't count how many times I've had to work with event footage shot this way by accident.)

5. Record video using a normal profile, as recommended in the Logarist guide.
6. Shoot a RAW still. (Again, high ISO, underexposed, wrong white balance.)

If you do this, you WILL be able to recover the RAW file with reasonably good results. Most of the colors on the chart will be able to be pushed into place close enough to perceptually match a daylight shot. You will NOT be able to do so with the video, whether you use the Logarist "system" or not. The baked-in color errors are too severe. The color information was destroyed, there is nothing to recover.

Yes, this is a somewhat extreme example. I would not expect it to work, no more than I would expect it to work with a still JPEG. I mention it here because when you make comparisons to Lightroom and RAW, I think you implicitly invite such hope.

 

[balazer]

so far I can't get the highlights to have a gentle rolloff using Logarist, not without additional adjustments and still can't get them to look as good as in my usual workflow.

I have 8 different highlight handling LUTs there, besides plain BT.709. Try them. If you don't like the way they do the highlights, of course you can roll them off however you want with a curve.

 

[balazer]

joe12south, I never claimed that 8-bit compressed video will be as good as raw. I only claim that the color correction can be easier and more accurate and more like what you can achieve with a raw image.

Logarist doesn't care if you record in 8 bits or 10 bits, compressed or uncompressed. Whatever you record in, that gets decoded to 32-bit floating point, and that's what Logarist works with. And yes, it's still subject to the limitations of your camera's recording, with its noise and quantization and clipping. Logarist doesn't care about any of that. It's just color values to Logarist, just numbers representing colors.

 

[joe12south]

Balazer,
My encouragement is to be careful about how you describe your product. From your website:

Logarist enables the same types of corrections that you can achieve using raw images in Adobe Lightroom, but without the need to shoot in raw, and right inside your video editing application.

A reasonable person could read that the wrong way.

I think you would also benefit from being more straightforward in describing the characteristics of the LUT. After all this back and forth, it sounds like:
1. Linear response curve
2. Wide gamut
3. Some amount of camera specific adjustments to get them all to the same point, but I'm unclear about which differences you are accounting for?

 

[Thomas Smet]

Balazer,

Any chance Logarist would work with Fusion or could be adapted to work in Fusion? I tend to use that for compositing and fine tune processing of material.

 

[diegocervo]

I gave it a try and I find it pretty straightforward to use it in DaVinci, expecially when you put all the nodes that you need in a powergrade.

 

[balazer]

I don't know about Fusion. I've never used it.