GH4 Leeming LUT One – the best LUT for the Panasonic GH4

visceralpsyche

Well-known member
EDIT: Current version of the LUT is v323 and the Cine-D version is cross-compatible with the G7. New customers will get v323 from the beginning :)

EDIT2: Note that from v201 onwards, the recommended in-camera values are 0 -3 -5 0 0 as per the latest manual. In other words, Contrast 0, and not -5, as was the original recommendation for v126 of the LUT. The manual on the website and in your emails will always be the correct version to use with the latest LUT.


I'm very happy to announce the release of Leeming LUT One, the best LUT for the Panasonic GH4! :D

WHAT IS IT?

Leeming LUT One is a true Look Up Table (LUT) for the Panasonic GH4's Cinelike D profile, the best linear profile for dynamic range in the camera (and one that doesn’t cost $99 to enable). This is different to a Look, which is a set of creative colour values to make the image evoke a particular mood.

Leeming LUT One is designed to give your image the best starting point for further grading, especially recorded internally, while correcting all the deficiencies of the Cinelike D profile regarding colours and tones. By simply applying the LUT, your footage will be corrected for colourimetry and low end colour noise will be minimised when a natural curve is applied.

BUT, BUT, V-LOG L HAS BETTER DYNAMIC RANGE!

Actually, no, it doesn't. With the correct in-camera settings for Cinelike D (0 -3 -5 0 0 and 0 everything else), Cinelike D and V-Log L have the same effective dynamic range. What does effective dynamic range mean? It means that both profiles are feeding the MPEG encoder the largest range of luma values possible from the sensor. The big difference between the two is the way these values are mapped, and here's where the problems for V-Log L begin.

When V-Log L was released, I was excited to try it out. The promise of another 1-2 stops of dynamic range was not to be sneezed at. Thanks to a glaring oversight by Panasonic, they left the back door wide open to enabling V-Log L in camera without buying an activation key. Thus, testing began worldwide almost immediately.

Then I and others started to notice the problems. The most obvious one was weird magenta and cyan colour splotches throughout all internal footage (chroma smearing), leading to ugly macroblocking-like effects on any flat surface, as well as other areas of the image. It turned out that this was due to the less than 8 bit precision Panasonic were encoding their log values with in camera. Colour tonality was severely reduced, and the result was these ugly artifacts.

What had apparently not been addressed by all the beta testers nor Panasonic themselves (to this day), is that V-Log L really wasn't suitable for internal 8 bit 4:2:0 recording, due to this severe lack of colour and tonal resolution. For 10 bit 4:2:2 external recording (at an added cost of the recorder, batteries etc) V-Log L held up much better, giving a fair look into the extra dynamic range promised, but with other issues still evident, like a shift to green in a lot of footage at higher ISOs.

So back to which has better dynamic range – Cinelike D or V-Log L? Because of all the colour artifacts and limited bit distribution in the internal encoded footage, V-Log L, while holding a flatter profile, is limited in dynamic range by the usable low end, which manifests as very noisy and blocky, since it is forced to use less precision to render the blacks than the highlights. Cinelike D on the other hand uses the full 8 bits of precision to render from white to black, meaning less colour noise thanks to greater precision in the blacks. When exposed per ETTR (Expose To The Right) principles, Cinelike D offers effectively the same usable dynamic range as V-Log L, with the benefit of almost non-existent chroma smearing, meaning flat swathes of colour hold their precision between nearby tones much better, for cleaner colours.

OK, SO TELL ME ABOUT LEEMING LUT ONE.

The biggest problem to date with Cinelike D has been its less than optimal colour and tonal reproduction. Panasonic maximised the dynamic range to give the best output the sensor was capable of, but at the expense of making the colours all wrong. What they appear to have maximised is the Y, U and V component separation, to give the largest and cleanest range of discrete values to encode into the image with minimal noise floor.

Leeming LUT One uses this information to construct a far better, colour accurate image which retains the dynamic range but finally gets the colour and tone right. Now you have all the dynamic range the sensor is capable of, with the added benefit of true colour reproduction. Because the raw image information is encoded at maximum precision possible, it ends up giving a cleaner post-LUT image than V-Log L, due to the aforementioned deficiencies of that profile and LUT combination.

ART MEETS SCIENCE.

When I was building the LUT, I realised that I could use the above information to not only meet or exceed the dynamic range benefit of V-Log L, but also reduce the amount of colour noise in the post-LUT image. This is the part where art meets science. I have built the LUT to be visually accurate, but also considered the noise floor colours that are most obvious to the human eye, and compensated for them.

This is the main reason it took so long to develop. Getting a colour accurate LUT is as simple as using a colour chart. But getting the noise floor to hide itself is not so easy, without potentially destroying the colours in the mids and highlights.

So I did a lot of night testing to see how colours react and iterated until I felt the colour balance was the best it could be, while not becoming visually distracting when the noise floor is lifted.

Of course, push the noise floor too much and noise is inevitable. But my idea was to build a LUT that both fixes the colour problems of Cinelike D, but also allows for what would be considered normal grading as well as shadow and highlight response. We don't generally lift blacks to an insane degree in normal shooting, so I used that knowledge to keep colours accurate in the mids while reducing the colour bleed that causes us to see colour noise in the shadows.

Luma noise is of course present, but as it's not contaminated by rogue colours it looks much more filmic and not at all distracting.

THE RESULT.

A single LUT that works across all white balances, provides the best colour, the best dynamic range and the least amount of colour noise for the GH4, which can be used with all GH4s, regardless of whether or not they have V-Log L installed, saving you $99 and giving you a superior image as well!

That's Leeming LUT One.

INSTRUCTIONS.

The most important points are to set up the camera correctly, white balance the shot, and use ETTR principles to retain all highlight information.

Don't clip anything, as it cannot be recovered! This is the biggest mistake people make when shooting.

Forget +1 EV etc. Expose to the right, hold the highlights you want using the zebras (explained in the manual), and in post-production, bring up the shadows and bring down the highlights as required to give the best, balanced picture with beautiful colours and tonalities and minimal colour noise.

From there you can grade for look with the best possible starting point.

FAQ.

Q: Why use 0 -3 -5 0 0 etc?

A: Here's the breakdown of each setting, and why it's the optimal one (established through extensive testing of each setting):

Contrast 0 gives the biggest dynamic range by evening out the distribution of tones across the histogram, meaning more latitude can be recorded while not compressing the highlight and shadow areas inordinately.

Sharpness -3 is where the image is sharpest without introducing ringing or other artificial artifacts.

Noise Reduction -5 effectively turns off noise reduction, which is optimal as even a small amount tends to lose luma detail from the image, resulting in ugly blocking artifacts. With the noise colour control of Leeming LUT One, filmic luma noise is retained without ugly colour noise, giving the nicest image possible.

Saturation 0 keeps as much of the true colour information as the sensor originally captured, without adding any (running the risk of super-saturating bright colours) or taking any away (losing colour precison and tonality, leading to colour errors and artifacts.

Hue 0 is the default setting of the camera which bypasses any colour shift processing. Hue is compensated for with Leeming LUT One, so changing this away from the default will only introduce errors.

Highlight/Shadow 0 prevents any inverted curve solarisation or posterisation artifacts.

Master Pedestal 0 keeps the full bit precision with no truncation at the low end. This results in the best colour bandwidth for the least amount of ugly colour noise in the shadows. Raising this away from 0 just compresses your colour bandwidth for no gain.

Luminance Level 0-255 give you the full bit precision of colour mapped values, for the best tonality.

SAMPLES.

Click on the link underneath each image to open the full 4K uncompressed PNG for pixel peeping in a new window:

Leeming_LUT_One_Sample_ISO400_1.jpg


http://www.visceralpsyche.com/misc/web_images/Leeming_LUT_One_Sample_ISO400_1.png

Leeming_LUT_One_Sample_ISO6400.jpg


http://www.visceralpsyche.com/misc/web_images/Leeming_LUT_One_Sample_ISO6400.png

Leeming_LUT_One_Sample_ISO1600_2.jpg


http://www.visceralpsyche.com/misc/web_images/Leeming_LUT_One_Sample_ISO1600_2.png

Leeming_LUT_One_Sample_ISO400_2.jpg


http://www.visceralpsyche.com/misc/web_images/Leeming_LUT_One_Sample_ISO400_2.png

Leeming_LUT_One_Sample_ISO1600_1.jpg


http://www.visceralpsyche.com/misc/web_images/Leeming_LUT_One_Sample_ISO1600_1.png

I look forward to your thoughts, footage and discussion :)

Cheers from Berlin,

Paul :)
 
Last edited:
I don't see any mention of compatibility with Final Cut Pro X. Is your LUT a 3D LUT (FCPX-compatible with the LUT Utility) or a 1D LUT (not FCPX-compatible)?
 
Bought , tried, falls short. Does not have more DR or even the equivalent of VLOG. Especially in the shadows. Set up exactly like your manual. Shot at 200 ISO which is noiser than VLog at 400 ISO.
 
Leeming CineD lut applied
Leeming.jpg

VLOG with Neumann "nothing crazy" LUT applied
Vlog.jpg

VLOG graded (gain, gamma, lift adjustments). More shadow detail, more highlight separation.
VLOG graded.jpg

Notice shadows of trees on the right. Highlights in bricks and rocks near fence.
 
Last edited:
Paul

thank you very much for this. I know the pragmatic quickly thinkers (non thinkers actually) will be appalled by this but would you consider three more luts optimized for each of the idynamic settings. I suspect (pragmatic quickly) if you set idynamic to high you can drop your ISO by 2 stops and get the same 18% grey data data level as you had before the idynamic and ISO change. Idynamic reaches into the shadows pushing the dynamic range to the left and lifting the mid tones, highlights at the clip point remain unmoved. The reduced ISO helps with the noise problem associated with idynamic, and gets you 2 more stops of highlight head room.

If if you go to http://www.dpreview.com/reviews/panasonic-lumix-dmc-gh4/10 you can see what I am talking about. I suspect the idynamic shown is being applied to the standard or natural profiles, it is not specified in the text.

I would certainly be willing to buy a pro extension pack from you that would be include luts for each of the idynamic setting, or even just the high setting.

again thank you Paul and I hope you will consider this.
 
@Dustylense, Thanks for the tests, looks like I've saved a total of $115.00 between the both. Your test clearly shows a pretty big difference between the two.
 
@Dustylense, Thanks for the tests, looks like I've saved a total of $115.00 between the both. Your test clearly shows a pretty big difference between the two.

in spite of dusty's post I just purchased Paul's lut, and look forward to checking it out myself and comparing to vlog, and pre and post lut d-log.
 
Leeming CineD lut applied
View attachment 105718

VLOG with Neumann "nothing crazy" LUT applied
View attachment 105719

VLOG graded (gain, gamma, lift adjustments). More shadow detail, more highlight separation.
View attachment 105720

Notice shadows of trees on the right. Highlights in bricks and rocks near fence.
You need to lift the shadows to match the other LUT. The LUT is designed not to artificially lift anything but it takes one second to nudge shadows up to match levels, and you'll see they are the same :)

Thanks for buying and testing. I'm confident my LUT holds up better than your initial test.

Cheers,

Paul :)
 
Bought , tried, falls short. Does not have more DR or even the equivalent of VLOG. Especially in the shadows. Set up exactly like your manual. Shot at 200 ISO which is noiser than VLog at 400 ISO.
Would love to see some full size stills of those tests you posted. Because no doubt the V-Log has the same magenta blocking it always has. That YouTube video has it all over (even the thumbnail for it shows it all over the water).

I actually think (read bias into it if you like, but I'm serious) that the still of my LUT applied to your image looks much nicer than the Neumann one. Your graded V-Log one looks more like mine from the thumbnail, but please do show us the full size ones for comparison, as I would especially like to lift the black levels on mine to show you the same look as the Neumann LUT :)

As always, I appreciate criticism too! That's how we improve :)

Thanks,

Paul :)
 
BUT, BUT, V-LOG L HAS BETTER DYNAMIC RANGE!

Actually, no, it doesn't. With the correct in-camera settings for Cinelike D (-5 -3 -5 0 0 and 0 everything else), Cinelike D and V-Log L have the same effective dynamic range.

Can you provide some proof of this?
 
I don't see any mention of compatibility with Final Cut Pro X. Is your LUT a 3D LUT (FCPX-compatible with the LUT Utility) or a 1D LUT (not FCPX-compatible)?
Good question. I'm Windows based so unfortunately I don't have Final Cut to test with. Can someone chime in and let me know? Then I can add that to the official list of supported apps!

Cheers,

Paul :)
 
You need to lift the shadows to match the other LUT. The LUT is designed not to artificially lift anything but it takes one second to nudge shadows up to match levels, and you'll see they are the same :)

Thanks for buying and testing. I'm confident my LUT holds up better than your initial test.

Cheers,

Paul :)

Already did that. And they are not the same. Look at the color channel clipping in the shadows.

Leeming Lut lifted shadow to match VLOG
Lemming Flower.jpg


VLOG Nuemann Lut.
Vlog Flower.jpg

Look at the color channels.

Not the same.

The only thing I see in your LUT is an improvement to the color of the blue channel for CineD for things like skies. Your greens however are super fluorescent looking.

Enjoy your $15.
 
Can you provide some proof of this?
Look at the images and you can see it. I know it's hard to take my word for it.

Dynamic range is NOT highlight retention. Digital sensors either have highlight info, or they are clipped. "Rolloff" is more a function of where middle grey sits relative to how many stops above that still hold information clipping.

Dynamic range is: Set the exposure to JUST beneath clipping range at the highlight end (99.9% if you will). Now look down into the shadows. How far down can you see usable information? This is what I refer to by dynamic range. If you compare V-Log L and Cinelike D with -5 -3 -5 0 0 (forget my LUT portion, the original uncorrected file will do) you will see that if both are properly exposed to hit 99.9% of highlight value, you'll be able to match the levels between them and find the same amount of shadow detail visible. This is what I refer to when I say they have the same dynamic range. The difference is, V-Log L gives a bunch of extra noise and colour smearing that Cinelike D doesn't, hence you can extract a cleaner shadow level from Cinelike D if you match levels.

If I had one I'd shoot a Stouffer chart, then there would be no more discussion. Unfortunately I don't have one. If anyone has one, please let me know. I'll throw a freebie your way in order to get a definitive answer.

Good discussion!

Cheers,

Paul :)
 
Paul

thank you very much for this. I know the pragmatic quickly thinkers (non thinkers actually) will be appalled by this but would you consider three more luts optimized for each of the idynamic settings. I suspect (pragmatic quickly) if you set idynamic to high you can drop your ISO by 2 stops and get the same 18% grey data data level as you had before the idynamic and ISO change. Idynamic reaches into the shadows pushing the dynamic range to the left and lifting the mid tones, highlights at the clip point remain unmoved. The reduced ISO helps with the noise problem associated with idynamic, and gets you 2 more stops of highlight head room.

If if you go to http://www.dpreview.com/reviews/panasonic-lumix-dmc-gh4/10 you can see what I am talking about. I suspect the idynamic shown is being applied to the standard or natural profiles, it is not specified in the text.

I would certainly be willing to buy a pro extension pack from you that would be include luts for each of the idynamic setting, or even just the high setting.

again thank you Paul and I hope you will consider this.
iDynamic makes mush of the image though. It's a terrible setting that really shouldn't be used if you want to retain your image quality and be certain of what you're going to get.
 
Would love to see some full size stills of those tests you posted. Because no doubt the V-Log has the same magenta blocking it always has. That YouTube video has it all over (even the thumbnail for it shows it all over the water).

I actually think (read bias into it if you like, but I'm serious) that the still of my LUT applied to your image looks much nicer than the Neumann one. Your graded V-Log one looks more like mine from the thumbnail, but please do show us the full size ones for comparison, as I would especially like to lift the black levels on mine to show you the same look as the Neumann LUT :)

As always, I appreciate criticism too! That's how we improve :)

Thanks,

Paul :)

Not getting horrible magenta blocking using the Neumann Luts or a handful of James Miller LUT's. And what does show up is the same about of artifacting I see using your LUT. Just being honest man.
 
Already did that. And they are not the same. Look at the color channel clipping in the shadows.

Leeming Lut lifted shadow to match VLOG
View attachment 105721


VLOG Nuemann Lut.
View attachment 105722

Look at the color channels.

Not the same.

The only thing I see in your LUT is an improvement to the color of the blue channel for CineD for things like skies. Your greens however are super fluorescent looking.

Enjoy your $15.
Greens can be pulled back if you think they are too much.

It's weird. I'm not seeing hard noise floors like you are. If I lift shadows in dark shots, they roll upwards like your V-Log example.

Original Cinelike D direct out of camera (underexposed - this was a torture test clip):

shadows_raw_1.jpg


Leeming LUT One applied only, no further adjustments (image is actually made darker, due to correcting for tonality):

shadows_LUT_Only_1.jpg


LUT applied and shadows/blacks lifted. Note there is information there, not a flat line:

shadows_lifted_1.jpg


Hopefully someone else can chime in with their own tests as well. I truly want this to be the best LUT that's non V-Log oriented, so it's good to get critical feedback to make it even better :)

Can you post the Cinelike D clip without my LUT applied and shadows lifted, to see if it makes a difference?

Cheers,

Paul :)
 
Not getting horrible magenta blocking using the Neumann Luts or a handful of James Miller LUT's. And what does show up is the same about of artifacting I see using your LUT. Just being honest man.
I wouldn't ask for anything else :)

You're recording internally, right? Just checking to be sure.

Would also really love some 4K uncompressed PNG files to pixel peep on your stills. Because in my experience there's no way to avoid the magenta unless you're recording external 10 bit.

Cheers,

Paul :)
 
Back
Top