Dan Keeble
Well-known member
new GH4 V-Log L footage : https://vimeo.com/138178172
GH4 GH4 V-log?
- Thread starter Kindredmachine
- Start date
joe12south
Veteran
Can you point out (or better yet, demonstrate) the fallacy in my statement? Sensors and monitors respond in a roughly linear fashion, but our eyes do not. If we were as sensitive to subtle gradations in highlights as we are in the midrange, the trick of log compression would not work at all.
joe12south
Veteran
Exactly. No disrespect to Lpowell, but he seems to be arguing as if we are discussing a linear image and not a logarithmic one. A LOG curve isn't magic, but recording with one isa very elegant "trick" that most certainly works, and it works because of how our eyes work.
joe12south
Veteran
I understand that it was an intentional aesthetic choice for this piece, but I shuddered thinking about the onslaught of grey, blah, flat images that are going to be unleashed upon the world after V-LOG L hits. ;-)
Also interesting to note that the GH4 does not lock you into the "native" ISO the way that the Sony a7 does when using S-LOG.
Color rendering of a fog?
Lpowell is actually arguing that VLOG-L works differently than other forms of LOG recording. Where C-Log and S-Log record "superwhites" between 64 IRE and 109 IRE, VLOG-L maxes out at 79 IRE, effectively reducing the bit depth of the recorded image.
Most of us are aware of the purpose of LOG recording to redistribute the light in a more efficient manner. But Lpowells argument is that VLOG-L's alteration of the Varicam's VLOG, which maxes out at 12 stops instead of 16, throws away bit depth. That's what he's getting at.
Now, Barry Green has apparently been trying to get Panasonic to remap V-LOG-L so that it has a higher clip point. Not only because it would give 8-bit recording more values to represent the image, but because it doesn't play nice with exposure tools like zebras and false color. It is counterintuitive.
I hope this happens.
joe12south
Veteran
1. I understand this theoretical concern, but other than that chart, and Barry's concern based on a different camera, I haven't yet seen a real world example of this. To the contrary, based on Nick Driftwood's sample files, it seems like this is not the case. And based on some anecdotal reports, it also sounds like it isn't the case. Unfortunately, none of the people who actually have it are speaking-up.
2. If part of the available data range is being wasted (let's hope not) equating that to "7-bit" is a bit disingenuous, I think. And even if that comparison were accurate (which it is not) it is still important to remember that the portion of the signal being sacrificed is hard to distinguish. All that is to say that blanket statements implying that you MUST use an external recorder to benefit from V-LOG L are premature at best, and damaging hyperbole at worst.
1. You are correct in that Barry's concern is with V-LOG-L on the DVX200, not the GH4. We will have to wait and see if the Gh4's implementation is the same.
But even though I haven't seen a real world example of problems with V-LOG-L, I have also not seen anything shot with V-LOG-L that looks better than carefully exposed footage shot with other profiles on the camera.
Personally, I have shot test footage on the GH4 with lifted pedestal and a flat tone curve and, upon adding an S-curve and pulling the blacks back down, have seen noticeable banding.
The argument here seems to be between rationalism and empiricism. Math vs. real-world experience. The math says V-LOG-L has only 150 values per channel and not 256. That is significant. Real world tests will see if this matters.
2. I don't think LPowell was implying that you MUST use an external recorder. I looked through his recent posts and nowhere does he say this. I think he's just pointing out the mathematical fact of how limiting the VLOG-L implementation is in 8-bit because it doesn't render anything above 79 IRE.
Is any of the sample video taken from internal recording? I remember reading that a few of them specifically listed an external recorder.
Maybe this debate should be put on hold for a bit while we wait to actual product to be shipped and installed. Then people can do 8 bit vs 10 bit recordings of the same scene and the truth of what is happening can be shown.
Maybe this debate should be put on hold for a bit while we wait to actual product to be shipped and installed. Then people can do 8 bit vs 10 bit recordings of the same scene and the truth of what is happening can be shown.
joe12south
Veteran
Here I would very strongly disagree. Cine-D in particular is a color train wreck, and in under all but the most controlled lighting situations the extra latitude it going to have a big impact.
As have I, and I too experienced the image falling apart.But "flattening" a linear profile like Cine-D is not comparable to expanding a logarithmic profile like V-LOG. More bits will be allocated in the areas that matter most.
Certainly some images are going to visibly suffer from a lack of color resolution. How often this will be a real world concern, and how often this concern will outweigh the benefits, is anybody's guess.
I agree. Especially the color clipping that happens in highlights. I'm referring to the Natural profile, which is much better.
I said nothing about "allocating bandwith" or "linear versus logarithmic" images. You might consider reading my posts rather than misrepresenting them.
As far as your claim of "no disrespect to Lpowell" is concerned, why change course now? You were doing so well with your insulting tone and derailing pontifications.
joe12south
Veteran
No intent to offend, but I do think that you're (intentionally or not) spreading FUD. The use of VLOG-L does not mandate that the user limits the exposure as might be inferred from your statements. Unless I seriously misunderstand what is trying to be said, and the files I worked with are dramatically different from those produced by the production version, then Barry Green's exposure recommendation is only valid if attempting to interchange with legacy V-LOG.
Barry_Green
Moderator
My comments were about the DVX200's implementation of VLOG-L. I do not know what the GH4's implementation will be, although I would assume that it will be similar.
VLOG-L is designed to interchange with VLOG. You don't get a chance to "attempt" to interchange it or not, it is the same. On the DVX200, the fact is that VLOG-L will not generate a signal hotter than 79 IRE. At full brightness, pointing the camera at the sun with the shutter wide open, the most that the exposure tools will register is 79 IRE. It's not a problem for the dynamic range being captured (as it is capturing a full 12 stops), it is only a concern when trying to use legacy monitoring and exposure tools. You will have to understand and adapt your exposure expectations when using VLOG-L on a DVX200, that's just a solid fact. It doesn't mean you need to expose necessarily differently (autoexposure, for example, works properly to set the middle gray where it belongs and to keep the highlights within range) but it does mean that you can't rely on your existing false color or zebras at their existing levels to advise you about proper exposure. You will have to adapt those tools, you'll want to use zebras at 60% instead of at 90% to catch 90% white for example.
Brandon Freeman
Well-known member
I'm assuming my workflow would be to set exposure with the profile I use now (Portrait, default settings), and then switch over to V-LOG L before recording. Gonna test it when my firmware arrives and see if that in fact will work for me.
Here's what Barry Green (and by implication, Panasonic) recommend when using V-Log-L:
Barry_Green said:
While you're free to ignore those recommendations if you want, overexposing your V-Log-L footage will make a mockery of the precise calibration required to use V-Log LUT's properly in a post production workflow. If you don't use a LUT compatible with V-Log specifications, you'll need to calibrate exposure and modify the footage tone curve to match a different type of LUT. If you don't use a pre-calibrated LUT at all, you're winging it with your eyeballs like any other enthusiast.
Last edited:
There's the core fallacy that lurks beneath your misleading conclusions. Panasonic's Cine-D tone curve is not "linear", it's a Rec 709 curve with a gamma of around 2.2. No Rec 709 video camera records the linear output of the image sensor, they all use gamma-corrected tone curves to roughly match the gamma curves of consumer TV's and monitors. See page 10 of this manual for more details on how gamma curves work in Panasonic video cameras:
https://eww.pass.panasonic.co.jp/pro-av/support/content/guide/EN/AG-HMC150handbook_e.pdf
In practice, the log curve used in V-Log-L is shaped in a similar manner as the gamma curve of Cine-D. Both have a non-logarithmic rolloff knee in the shadow region, a flat midtone region where gain is proportional to f-stop, and little or no rolloff knee in the highlights. The main difference is that log profiles are calibrated with precisely logarithmic response in the flat region, along with a precise rolloff curve in the shadow region. Those precise specifications (at a specific exposure level) are what enables you to match the log footage to a pre-calibrated LUT in post. While you can produce similar results with a tweaked Cine-D profile, its lack of precise calibration will most likely produce crude results when processed with a LUT designed for a log profile. But for many folks, eyeballing it in a video editor is good enough for their purposes.
rogar-BlackEye(DBG:angry::nads:joe12south
Veteran
I think you're being pedantic to the point of obfuscation. All of the existing profiles encode the image for **display.** They have a very gentle curve, even Cine-D. V-LOG L is not meant for display, it is meant for capturing the most information for post processing. It doesn't look or behave like Cine-D.
joe12south
Veteran
Us poor misguided "enthusiasts" "mocking" "proper" workflows.
Sorry, I'm finding it so hard to not fan this flame. Input LUT's are convenient, but you and I both know that a good colorist doesn't **need** them. The final image is all that matters.