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

dustylense said:
Well, you want that input LUT to be suitable for your desired output LUT. If yellow isn't yellow but more reddish and blue is not blue, but more purple or cyan....
Click to expand...
Gotcha. My assumption is that the output LUT was created to expect a proper REC709 image. I get a little anal about these things. I'm all for fun, creative "looks" so long as they are clearly labeled as such. In my mind, the Neumman log is either "wrong", or purposely creatively skewed. I wouldn't expect my input LUT to deviate from correct for the sake of any one output LUT.
 
David Del Real said:
Hi Joe, are you testing 0-255 and 16-235? Also, a big thank you for all these tests!
Click to expand...
There is no meaningful difference. Assuming your NLE can properly deal with a full range image, and you're not turning over unprocessed files for broadcast, there is no reason to shoot 16-235.
 
joe12south said:
Gotcha. My assumption is that the output LUT was created to expect a proper REC709 image. I get a little anal about these things. I'm all for fun, creative "looks" so long as they are clearly labeled as such. In my mind, the Neumman log is either "wrong", or purposely creatively skewed. I wouldn't expect my input LUT to deviate from correct for the sake of any one output LUT.
Click to expand...

The Neumann LUT takes the VLOG into a mapped R709 space. And the yellow is far more true. I guess I have to dig out the color card from the grip van.
 
dustylense said:
Like the yellow of the cup and blue of the sky.

Leeming
View attachment 105781

Neumann VLOG "Nothing Crazy"
View attachment 105782

Neumann VLOG "Basic Clean"
View attachment 105783

Now, if you are shooting a "hotdog on a stick" commercial, you better have the primaries pretty darn close on the input.
Click to expand...
The hues in an H.264 recording are highly sensitive to the color temperature selected in-camera. IMO, an input LUT designed for a built-in camera profile should strive to optimize its dynamic and tonal range while maintaining all hues as close to the original as practical. To evaluate this, I'd want to compare sample footage before and after LUT processing.
 
BONUS ROUND
Underexposure

V-LOG L > VLOG709 LUT
underexposure_vlog.jpg

CINE-D > Leeming One LUT
underexposure_leeming.jpg

V-LOG L > VLOG709 LUT (200%)
detail_under_noise_vlog.jpg

CINE-D > Leeming One LUT (200%)
detail_under_noise_leeming.jpg

The GH4 sensor sucks at low light. It's noise floor is very high. V-LOG L doesn't do anything to make this any better, but when you push it to normal levels in post, it does serve to make the problems even more noticeable.

The Leeming LUT deals with this by crushing the saturation in dark areas. In a properly exposed image this is a perfectly acceptable tradeoff. In an underexposed image, you're left with an almost monochrome image. I added back significant saturation, but it was not possible to regain natural colors.

Which is the worse of the two evils? I find it hard to make an empirical judgement. It's something of a matter of preference. Personally, I hate digital chroma noise, so I prefer Paul's band-aid.

WINNER: Nobody, we all lose. The real take-away, is ETTR! The GH4 is not forgiving of underexposure!
 
Last edited:
BONUS ROUND 2
Overexposure

V-LOG L > VLOG709 LUT
detail_highlight_vlog.jpg

CINE-D > Leeming One LUT (200%)
detail_highlight_leeming.jpg

Both profiles exposed exactly the same. The Cine-D>Leeming One LUT image not only has less color errors, but it holds onto more detail. V-LOG L clearly does nothing to help with accidental overexposure.

WINNER: Leeming One
 
dustylense said:
The Neumann LUT takes the VLOG into a mapped R709 space. And the yellow is far more true. I guess I have to dig out the color card from the grip van.
Click to expand...
I'm not passing judgement on which of those two is more accurate, only stating that accuracy is my preferred goal for an input LUT.
 
Joe, are you using the VLOG to Varicam 709 LUT? Because if you are, throw that LUT away. It's garbage for the Gh4 VLOG-L
 
dustylense said:
Joe, are you using the VLOG to Varicam 709 LUT? Because if you are, throw that LUT away. It's garbage for the Gh4 VLOG-L
Click to expand...
I am. Compared to anything else I have on hand I've found it the most accurate, if not the most pleasing. I know you prefer Neumann's, but do you find it accurate? Based on quick glances at the greyscales from some of images you've posted, it appears to have a cast.
 
Last edited:
VLOG-L > VLOG709 LUT
skin2_vlog.jpg

CINE-D > Leeming One LUT
skin2_leeming.jpg

Okay, please excuse the crappy lighting. I'm abusing my poor family as "charts", not trying to make them look good. ;-)

The goal here was to show both in a typical, real-world, best-case scenario. A naturalistic scene in which V-LOG L's worst problems aren't really highlighted. I normalized the two images so that the slight exposure difference wouldn't bias judgement. In this scenario, the differences between the two is much more subtle than the more synthetic tests would suggest. IMO, if the red deficiency in Leeming One was nudged (most evident in Zack's shirt) the two images would be very close. Even as it is, it's hard to pick a clear "winner."
 
So Paul, where are you relative to changing the LUT based on Joe's tests? Making some more changes or not? Do you think current version is "final"?
 
Last edited:
joe12south said:
I am. Compared to anything else I have on hand I've found it the most accurate, if not the most pleasing. I know you prefer Neumann's, but do you find it accurate. Based on quick glances at the greyscales from some of images you've posted, it appears to have a cast.
Click to expand...
Well, the Varicam LUT is garbage because it amplifies the Magenta/Cyan issue.
As for my samples using Luke's LUTS. You have to take into consideration that what you saw was shot at 5600K during real late afternoon. So one would expect a cast. Luke's profiles, as well as some of James Millers, really take on the Magenta blocking issue and make internal VLOG quite useable.
 
dustylense said:
Well, the Varicam LUT is garbage because it amplifies the Magenta/Cyan issue.
As for my samples using Luke's LUTS. You have to take into consideration that what you saw was shot at 5600K during real late afternoon. So one would expect a cast. Luke's profiles, as well as some of James Millers, really take on the Magenta blocking issue and make internal VLOG quite useable.
Click to expand...
I'll give them a shot. With an external recorder that can both record 10-bit and preview LUTs, I might be much more tempted to use V-LOG L on an actual shoot.

I think that the Varicam LUT is simply accurately revealing the problem. I say this because charts hit nearly exactly (crazy good) and when I manually corrected a V-LOG L image via scopes I ended-up at nearly identical (but slightly inferior) results.
 
graded_leeming.jpg

Last image for awhile. This one just shows how the CINE-D/Leeming One image holds up to some pretty extreme creative color grading. It's a 10-stop 8-bit image, so there's only so much you can expect from it. IMO, holds up better than 8-bit V-LOG L, but this isn't exactly a torture test. An improperly exposed high DR outdoor shot would be a better test, but I think I've seen enough to know that it will not do significantly better or worse than anything else coming out of the GH4.
 
hue.jpg

I normalized values and saturation in order to compare hue between the two. I've noted when I can clearly state that one or the other is more accurate.

Paul, overall saturation is still a bit low. As it comes out of the camera at -3 is pretty accurate.
 
Paul has sent me version 1.8 beta. It's dangerously close to amazing. There are just a handful of discrete hue errors which I hope can be dialed out:
hues.jpg

V-LOG L/709
scene_vlog.jpg

CINE-D/LEEMING ONE 1.8 beta
scene_leeming.jpg

And as a reminder as to why this matters so much, let's not forget the ugly problem with V-LOG L:
detail_vlog.jpg

Which is a non-issn with Cine-D:
detail_leeming.jpg
 
Paul's up to version 109 beta. In the few areas the color isn't dead on, the difference is less than 5 degrees (not 5 percent) so it is very small. I'm pretty sure I'm going to switch back to shooting Cine-D with this LUT as is, but I remain hopeful that if he doesn't strangle me first, Paul might be able to get Leeming One even better.

I'm not going to post the recent changes because they are so subtle, but i'm going to shoot a different set-up and if Paul can crack the ceiling, I'll certainly share those results.
 
Last edited:
You must log in or register to reply here.
Back
Top