Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Scarlet and F3 pointed at the same scene. Footage inside.

Status
Not open for further replies.
I can vouch for Timur, lol, the man knows his shit.
 
Timur, I suggest testing under very controlled lighting sources if you can. I would confirm color temperature of lights via a meter so that you can eliminate some variables. For initial tests perhaps avoid mixing color temperatures. That way, you will have a baseline.

Include in your frame the following items.

- A human being with skin tone
- A Macbeth chart or DSC chart
- objects/props that cover a wide tonal range

Use a calibrated monitor on set. Establish correct white balance on set. Nail the exposure.

Bring results into post and see where things are at. By eliminating as many variables as possible (for example mixed lighting) you can get your baseline. How do these cameras operate under controlled tungsten. Once you establish that, then from there introduce mixed lighting. The key is to establish how the cameras respond under ideal circumstances (in this case, ideal just means tungsten light that doesn't have any weird temp shifts and is controlled).

Now, when it comes to lighting, i've seen your work and that part you have well under control. I'm sure Sanjin didn't mean to offend. :-)
 
Sanjin Jukic said:
Taken from RECON:

2. Set White Balance BEFORE converting to RGB space for grading.
...

People who complain about RED's color usually find out that they have made this tactical error.

The easiest thing to do is hit AWB in camera and/or shoot a white or grey card in the scene

and hit it with the WB tool in REDCINE-X or other app that uses the SDK.


.
Click to expand...

Can this possibly be true? That's to say that setting WB in camera is qualitatively different to setting it in Redcine X, or any other system that deals with R3Ds (rather than burnt into a transcode). What is it about the meta data from the camera that's different to the meta data from Redcine X? Or am I missing something. Surely it's obvious that if you try grading a Prores (or similar) transcode with the wrong WB already burnt in you're asking for trouble. But that's comparing apples and oranges.
 
Sam Gordon said:
Can this possibly be true? That's to say that setting WB in camera is qualitatively different to setting it in Redcine X, or any other system that deals with R3Ds (rather than burnt into a transcode). What is it about the meta data from the camera that's different to the meta data from Redcine X? Or am I missing something. Surely it's obvious that if you try grading a Prores (or similar) transcode with the wrong WB already burnt in you're asking for trouble. But that's comparing apples and oranges.
Click to expand...


I don`t think this quote (Jim`s quote) is being interpreted correctly. What is said, is that you should White Balance you footage while you are still in RAW color space. For exemple, if you take un withe balanced footage in REDCINE X and export to DPX or ProRes for grading, as these two formats are in RGB color space. If you white balance in REDCINE X (or any other application that is using the actual R3D file) them you are still in RAW color space, and therefore, there is no diference what so ever between white balancing in post or in camera.

The only diference would be your monitoring, as all camera settings is juts metadata

Cheers
 
Rodrigo Prata said:
I don`t think this quote (Jim`s quote) is being interpreted correctly. What is said, is that you should White Balance you footage while you are still in RAW color space. For exemple, if you take un withe balanced footage in REDCINE X and export to DPX or ProRes for grading, as these two formats are in RGB color space. If you white balance in REDCINE X (or any other application that is using the actual R3D file) them you are still in RAW color space, and therefore, there is no diference what so ever between white balancing in post or in camera.

The only diference would be your monitoring, as all camera settings is juts metadata

Cheers
Click to expand...
Correct. Seeing proper white balance on set can help solve some issues on set, rather than in post. For example, while standing at the monitor one day during a setup I noticed a color shift happening. I asked the gaffer if he would check some of the sources for me to see if there were some bad bulbs. Sure enough, one of the sources had a bad bulb with a spike in the spectrum, a new bulb at that. Got it replaced and the images on the monitor looked nice after that. :-)
 
Timur has been my friend for years, and he did this test with the available time and tools he had. the time factor is always at a minimum cause he's always working so much :)

I'm probably gonna dive into helping him out with the post processing on the next set of tests. He doesn't own a color suite, or monitors I deem trustworthy for critical color. keep in mind he's only doing these corrections on a mac screen too. so he's surrounded by tools that lie to him right now, and throwing off results everybody else is seeing. Trying to do these kind of tests and help the community with his knowledge and experience is a daunting task being a one man band. So next set of tests will be


sony f3 recorded 444 dpx in s log.
Scarlet to RAW r3d (developed in redlogfilm. gonna have to be redcolor2 cause redcolor3 isn't avail in resolve yet) and locked on WB before any additional corrections are added

both will be graded in Resolve, grading monitor will be a calibrated sony pvm-2541 oled, and i think for both objective and subjectively, several results should show

a simple delog to see where the colors naturally fall with just a simple contrast expansion with white balance predetermined. in the case of f3, straight out of camera, in the case of scarlet, wb defined in the RAW. (for most accurate results we'll do a white card, and wb in rcx first with color picker and redo those values in resolve, wb/tint)




next set will be colored to match, and see how the highlight rolloff and shadow detail differs and if they can intercut seamlessly.


and finally a pushed creative look and see which one falls apart first. this will be the most subjective portion, as there are a lot of ways to manipulate the RAW data first before going to primary and secondary color. but in this case since the f3 has no raw functionality, i feel like the only fair comparison is just to set them to matching points neutrally, and than try to do a creative look pushing the image far and equally doing it with the other one and seeing how robust it can be.


we won't be diving into 422 recording either, maybe another day if there is a demand for it, but i feel the starting point should be the highest quality acquisition the camera is capable of first.
 
Fun thread :emote_popcorn:


I like Timurs tests, though I would like them more if there was something to listen to, as well as a side-by-side. Looking forward to see Tom n Timurs creation =)
 
Timur Civan said:
there are two different bulbs in those two lamps, one is a GE Reveal and the other is a standard bulb.

I'm going to reshoot the test from scratch, and try that work flow that Sanjin Suggested from the Recon thread. See if it makes a difference. I have to reshoot it because i trashed the R3D's by accident. Will re upload the file in the video too, reflecting any new findings.

I imagine the RED should perform a bit better in terms of color. IT was very magenta.
Click to expand...

Having revised my opinion of the F3 footage, and noting that they are not in fact the same, it's obvious that the R3D's are amplifying the magenta case dramatically due to their color science. I would be curious as to what you white balanced off in this scene, simply to figure out if it's purely to Color Science or an interaction between what you were using to white balance and the color science (I know Graeme's preference seems to be White Balancing off a neutral grey card to not pick up reflected light - as such if he is also the designer of the color science then this workflow may yield better white balance results for a given scene...)

It's certainly a performance difference, and I would imagine that ultimately, given they are different bulbs, the scene in reality was somewhere in between the two.. The Delogged F3 looks more natural, but a little sickly. It's more natural because skin can go that color. The over saturated Magenta on the RED footage looks more healthy, but is an impossible tint towards magenta for natural skin, which makes it feel worse. Looking forward to further testing.
 
Bruce Allen said:
Thanks Timur for doing this.

It takes a brave man to put up a tungsten skin tone test on REDUSER.

Your results are the same as what I'm getting - it's really really difficult to get filmic and natural skin tones (yes you can have both at the same time) on RED yet.

I'm not saying F3 is perfect. And I'd love to see if you can get the two a bit closer matched in terms of color balance - F3 is a little too green, Scarlet is too magenta.

But I feel that I can add some warmth to the F3 shot in the grade at it'll shine nicely. The Scarlet shot feels like there is some lack of color detail in the face. Like there is some kind of channel crosstalk or IR pollution or... I don't know! It just doesn't look right!

Look forward to the new tests! I haven't really played with RedColor3 so maybe it is possible?

Otherwise... yes, hope Dragon is better in this regard. But then for a long time I was the only guy talking about this on REDUSER so I'm not too sure if RED took my mutterings seriously.

Bruce Allen
www.boacinema.com
Click to expand...

I know I'm rather new around here, but agree with you 100% on the skintones.. I've read all the threads around here on the topic, seen all the examples, and I feel like its definitely one of the things that can use some improvement. I haven't seen as much with RC3 so I'm curious to see how that changes things.. should have my Scarlet back tomorrow to start running my own tests. But yeah, your mutterings should be taken seriously haha.
 
Can you put your fill light on a squeezer, and show how the skin tones are affected while light is dimmed?
 
Ok couldnt do the Fil light thing, mainly cause i used no fill. I wanted a lot of contrast. See how skintones hold up in shadow as well as direct light.

This was an interesting test. The results are coming in a couple days, i'll make a new thread.
 
okay, maybe its just me - but looking at the footage on a calibrated screen - the sony color looks, well, like Sony color - which ALWAYS to my eyes looks FALSE (its not green, its neutral, but its like clinically NOT REAL) whereas I imagine her skin tone probably actually is that Reddish in the light you guys are in. But that could be my bias.
 
that's the thing about skin tones too, it's completely subjective. hollywood has their trend on what a skintone should look like, many DP's i meet like the "golden" skin tone, and now everybody here is digging more red in it.

so this next set of tests is gonna be more objective, and we are scrapping out lot of the original things we are gonna do. instead this is gonna more of a, you take a look, you decide what looks better and we'll leave it at that. it's not a winner loser thing, it's just about the kind of look you dig, and the possibility of intercutting one another on shoots that might want mixed cameras.



what will be presented, hopefully by this weekend,

f3 de logged, from dpx 444. de log meaning just contrast and saturation. only adjustments to luminance, and increase in saturation to something more poppy.

scarlet footage, shown in rc3, rg3, plus an additional de-logged, redlogfilm version in rc3. to show what a more custom version would look like in a advanced pipeline.

I also was monitoring everything through a leaderscope, that had a self screen capture option. I will be posting the captures of the waveform/vectorscope representing each image. in log and out of log. in all the settings I just mentioned. everything shot has a macbeth chart in it, and everybody can see for themselves how the cameras "naturally see"

this is about where the camera just naturally settles into, everybody knows you can color correct all day and adjust this and adjust that. this is all about what the camera wants to do right off the bat.

there will also be shots of just the macbeth chart with exaggerated saturation so you can see it on the vector better.


lastly, if anybody wants to donate up some server space, we will be posting 5 seconds of each segment. the original dpx files from f3, and the trimmed r3d. and the LUT i created as well for the f3 so people can play around with it themselves.

pm me if you got server space u wanna offer up...
 
I have to say, what's clear to me is the incredible amount of time and energy both Timur and Tom are putting into this second round of tests. Speaks volumes about both of their characters, and I salute them both!

Also reminds me of that wonderful proverb, how does it go? "Those who can do, those who can't..." :lol:
 
Last edited:
Dan Kanes said:
okay, maybe its just me - but looking at the footage on a calibrated screen - the sony color looks, well, like Sony color - which ALWAYS to my eyes looks FALSE (its not green, its neutral, but its like clinically NOT REAL) whereas I imagine her skin tone probably actually is that Reddish in the light you guys are in. But that could be my bias.
Click to expand...


She is actually rather yellow.
 
Status
Not open for further replies.
Back
Top