FS7: slog3 color correction breakdown video

starcentral

starcentral

Veteran
Last edited:
Thanks Dennis for posting this. I can't tell though from the video what grade is being used for each swipe. Do you have info on that? Thanks.
 
Im sorry to say I think their skin looks wooden. Actually worse than wooden.

There seem to be 2 'key lights' that hit the faces, maybe it is just overlit ie those lights pushed the skin into the 'knee' area of the file.
 
skimmel - I hope to do a follow up video with that. I did this quite fast, all by hand, no plugins, etc.. except Kodak 2383 LUT and 35mm_G3_DIRTY.MOV. There is a small write up on this on my blog if you click my name below.

sam - the scaled down image as it appears in the post on DVX does seem much more contrasty, hence perhaps emphasizes any crunchy wood. Do you feel its as bad on the full sized still images from the flickr link? If you look at the full size image on flickr and use arrow keys to go back and forth between the slog grade and colored one you can see that I practically kept the exact tonality on the faces in the slog version to the final version. So if the wood is present in the original slog version, maybe I could have tried to reduce that in the grade better. Maybe the lighting was a bit too harsh on-set.

This is my favorite still with the grain, although looking at it like this I think I have too much skin color.

https://www.flickr.com/photos/starcentralfilms/15785080502/in/set-72157648881013078/lightbox/

cropped:

Untitled_zpsc8c46637.png


All and all remember that this is only 10-bit. Maybe just not enough gradations where it matters most for skin and highlights..
 
starcentral said:
All and all remember that this is only 10-bit. Maybe just not enough gradations where it matters most for skin and highlights..
Click to expand...

Ive seen the bigger files.

10bit has plenty of graduations, 4* more than the 709 delivery space.

The problem occurs when the log curve is so flat that it is less than 1/4 the steepness of linear.. at that point 10bit does not have enough graduations.

Well the problem really occurs when the skin is exposed in a portion of the log curve that has 1/4 or less of the steepness.

S
 
Yes but any such problem as described in essence makes 10-bit unusable, and also means the full 14 stops is also unusable. Hence why I would tend to suggest not enough gradations to render better looking less crunch-wooden skin tones.

Is this function of XAVC codec however? Would SR or ProRes be marginally or drastically improved? The jury is still out, but if this was shot on the R5 RAW in 16-bit I think with the exponential increase in data values above the mid-tones it would definitely make for much smoother gradations.
 
My minimal experience with XAVC says that it compresses the shadows but should not really be a problem - just can go a bit gritty.

SRcodec 444 = IRO 4000 levels (instead of 1000) should have enough flex, again my minimal experience with that codec is that it is Sweeet

Of course Slog was designed for 444.

Can we 'make' 444 1080 from 10bit 4k?

My experiments with cine1 are all about getting more levels than Slog.

Again with absolutely spot on exposure it might be possible to avoid wood, Ive been revisiting some of my F3 footy and finding that some of the clips sing with the Arri Lut and some just go horrid, I guess that is varying exposure from me. My exposure doesnt vary cos I am a muppet, just that Im firing slightly blind (DP6 has minimal tool-age)

S
 
Im hoping to test the FS7 and 80% am not expecting log to work, but if I can get any sweet baked (with a bit of flex) looks then the camera might work for me.

S
 
For critical exposure stuff (Slog) I plan on using false colour to nail exposure. I personally prefer it to waveforms as I can see where each value lies on the image exactly. It is by far my fav exposure tool.
 
Yes false color is handy dandy.

Unfortunately however if you shoot in slog because the slog values are scaled down on the full IRE scale most 3rd party monitors won't work well with it. For example a monitor with false color will maybe show you values over 90% in red or yellow, but in slog your 90% white should be going over 65-68%.

If you enable a REC709 (800%) LUT on top of slog, then of course it works much better.

Does the FS7 offer false color via body or VF?
 
I was mostly just thinking of using it to verify that skin specifically was in the right range and then use WFM for brightest and darkest values. Wouldn't I be able to throw false colour on my external monitor and make sure skin is in the right range? Sort of like I do with zebras on Rec 709.
 
starcentral said:
Yes with rec709 you'd be ok. I just meant false color directly on slog signal.
Click to expand...

Would it not show the skin range on Slog? If I sort out where I like my skin tone with Slog 3 can I not just look for the colour associated with the IRE value on the persons face? Or are you saying the Slog3 skin IRE values wouldn't show up? The only log I've used false colour on before was Clog and it worked perfectly.
 
But in rec709 skin should be in the 45-55 range let's say. In slog that's in the 30's...

Look how SmallHD maps out the IRE ranges to their colors:

SmallHD-False-Color-IRE-Scale.jpg


Purple has a huge range, so monitoring an slog signal wouldn't work that well, unless you monitor on rec709 signal, or the particular monitor you're using let's you charge the range scale for each color...

Hope that makes more sense.....
 
Also if your monitor only supports HDMI, as I think the ikan-vh8 does, you might want to see if your monitor supports Sony's HDMI output in 24p mode. On the F5/F55 the HDMI port puts out a crazy signal when the camera is in 24p mode, which only works on 2-3 monitors I am aware of.

An ikan monitor I've used which clearly says it supports 24psf over 1920x1080 never worked with my Sony F55. I opened a case with ikan about it and never heard back from them despite numerous follow ups.
 
Last edited:
You must log in or register to reply here.
Back
Top