GH3 Flat Color Profile to match Marvel for AF100

Kevin Sawicki

Kevin Sawicki

Member
I'm leaving for Korea to shoot a documentary next week and would like to prep my equipment by this weekend. I currently own a AF100, and just purchased a GH3 to use as a 2nd camera. I will be coloring the final film in DaVinci.

On the AF100, I'm using Marvel's flat profile: http://marvelsfilm.wordpress.com/2011/08/08/panny-af100-af101-flat-picture-style-for-cinematic-look/

Can anyone recommend a flat color profile for the GH3 that would match or be similar to the Marvel Flat Profile?

Thank you!
 
robcollins said:
I've read elsewhere on this forum that:

"The flattest picture you can get is NATURAL at:

-5
-4
-1
-5"

from (http://www.dvxuser.com/V6/showthread.php?303818-Panasonic-GH3-Novice-Help-Please)

I have both cameras and hope to do a test soon.
Click to expand...

I would caution to test first, especially in documentary-available light situations. My experience with settings similar to this is that the in-cam codec wouldn't hold up to the grade needed to pull it back. Just going as flat as possible is not the solution, unless you are really confident in your tools to clean up the results.
 
LyleGorch said:
I would caution to test first, especially in documentary-available light situations. My experience with settings similar to this is that the in-cam codec wouldn't hold up to the grade needed to pull it back. Just going as flat as possible is not the solution, unless you are really confident in your tools to clean up the results.
Click to expand...
Interesting. I always assumed that the flattest profile, reflected as straight a signal from the sensor as possible, which in turn minimized artifacting that results from the DSP applied in 'prettier' profiles. Correct me please.
 
Zephyrnoid said:
Interesting. I always assumed that the flattest profile, reflected as straight a signal from the sensor as possible, which in turn minimized artifacting that results from the DSP applied in 'prettier' profiles. Correct me please.
Click to expand...

The DSP may introduce artifacting, but not as much as may be introduced in post adding the requisite contrast curve to the flat footage. The issue is that the GH3 uses an 8-bit codec. Shooting "flat" and then adding contrast in the grade later can actually cause histogram gaps because of the 8-bit codec. This can appear as banding and other artifacts that might not appear if shot with the contrast already "dialed in". "Flat" makes more sense with a 10-bit or 12-bit codec, where there are so many values per channel that there's plenty of headroom and no perceivable banding. But when you only have 256 values per channel (in an 8-bit codec), even crushing shadows or adding contrast can make the image look bad.
 
Last edited:
dvbrother said:
The DSP may introduce artifacting, but not as much as may be introduced in post adding the requisite contrast curve to the flat footage. The issue is that the GH3 uses an 8-bit codec. Shooting "flat" and then adding contrast in the grade later can actually cause histogram gaps because of the 8-bit codec. This can appear as banding and other artifacts that might not appear if shot with the contrast already "dialed in". "Flat" makes more sense with a 10-bit or 12-bit codec, where there are so many values per channel that there's plenty of headroom and no perceivable banding. But when you only have 256 values per channel (in an 8-bit codec), even crushing shadows or adding contrast can make the image look bad.
Click to expand...

I concur.

To be more specific: yes, you can flatten the hell out of the GH3 picture but it can introduce more problems than it solves. Sharpening and NR can probably be dialed down at will, but contrast and sat need to be adjusted carefully according to conditions. I'm am also pretty certain everyone is really really looking forward to the first firmware upgrade.
 
Last edited:
dvbrother said:
The DSP may introduce artifacting, but not as much as may be introduced in post adding the requisite contrast curve to the flat footage. The issue is that the GH3 uses an 8-bit codec. Shooting "flat" and then adding contrast in the grade later can actually cause histogram gaps because of the 8-bit codec. This can appear as banding and other artifacts that might not appear if shot with the contrast already "dialed in". "Flat" makes more sense with a 10-bit or 12-bit codec, where there are so many values per channel that there's plenty of headroom and no perceivable banding. But when you only have 256 values per channel (in an 8-bit codec), even crushing shadows or adding contrast can make the image look bad.
Click to expand...

+1.

Noise will increase the more you futz with the image in post. IMHO, it's best to get the profile as close to grade as possible in camera. I think all this craziness over flat profiles was also influenced by the 5Dii hackers who realized that the stock codec was sacrificing shadow data for increased detail in the mids. The GH3 codec doesn't seem to have this problem.
 
Zephyrnoid said:
256X 3 (RGB) Is that correct? Can't be that low can it? Then again with native 4:2:0 capture, sounds about right. Yuck!
Click to expand...

No it isn't low until you start pushing those color channels. Of course it depends how much you push your color. If it is really flat however you will be pushing it a lot. Most of us will deliver 8bit anyway but if you start with 8bit and push it you could up delivering 7bit or 6bit or even 5bit if you go nuts. By starting with a flat 10bit or higher image you can adjust color a lot and still end up delivering 8bit.

In camera profiles also apply those curves and colors to the raw dsp image which I believe is 14bit. Not only is it higher bits but it is pre compression. Raw uncompressed 4:4:4. No matter how good compression is it always breaks down your image into blocks. Even jpeg which is actually a very good compression format can start to show those blocks as you adjust in post. To the naked eye you may not see them before you grade but start to push them and you will see them easier. The raw dsp can be pushed rather far and not show these artifacts in the recorded image since the profiles happen pre compression.

Flat can give an interesting look but it is only really practical from an image quality perspective when shooting raw or at the very least 10bit. Even then 10bit with HD can also fall apart if you go nuts. Many 10bit formats are also compressed with dct based compression including ProRes so they can still have the artifacts issue.
 
Zephyrnoid said:
256X 3 (RGB) Is that correct? Can't be that low can it? Then again with native 4:2:0 capture, sounds about right.
Click to expand...
The H.264 codec converts the camera's RGB data into 4:2:0 YUV format and encodes it in 8-bit precision. The encoded luma channel is a full-resolution 8-bit signal, but the chroma channels are downsampled to half the vertical and horizontal resolution. When a decoder converts the compressed YUV data back into RGB, it combines the half-rez 8-bit chroma data with the full-rez 8-bit luma data to produce RGB pixels. The results fall short of matching 8-bit RGB resolution, but in well-exposed areas of the footage, it's hard to discern any differences.
 
You must log in or register to reply here.
Back
Top