Adventures in the PSE

FilmBrute said:
The problem every picture style runs into is: 8 bit. All the styles have trade offs, like Samuel always points out. Flaat might be worse than some profiles in the shadows, but better in the midtones. And maybe some crazy profile will get you an extra 1.5 stops, but the midtones will suck. Personally, the fact that Act of Valor was shot on neutral kinda slapped me across the face. It made my snap out of this picture profile obsession. :grin:

The problem I faced when using some profiles is that I misused them. I think its an easy mistake to make; not understanding how to optimally use a particular profile. Like expsoing properly, or knowing what the profile is designed for. So, in a lot of circumstances you can do more harm than good. For instance: I see many people use technicolor for in studio facial close ups. And after grading, the skin looks awful. It makes so much more sense to use portrait or neutral, or prolost flat if you want to get fancy in that scenario.
Click to expand...
also, agreed

I end up using provost-like profiles mostly. While I liked my personal picture style, I never dialed it in, and then noise become an issue. Too many small color adjustments and being based on the wrong initial picture style maybe.

But the 1D-C has got me interested in pic styles again. Especially because prolost feels so wrong on it, and standard seems weird in the highlights.
 
I saw that Boris Jansch posted some 1D-C footage with Flaat10n. I've always thought 10n was a bit greenish and this gives the same impression, but I still love the colors in this one. I'd definitely be trying 10p on a 1D-C if I ever get to shoot with one.

http://vimeo.com/62968383

It would be interesting to see what you come up with for the 1D-C James.

Cheers,
Pete
 
c3hammer said:
I saw that Boris Jansch posted some 1D-C footage with Flaat10n. I've always thought 10n was a bit greenish and this gives the same impression, but I still love the colors in this one. I'd definitely be trying 10p on a 1D-C if I ever get to shoot with one.

http://vimeo.com/62968383

It would be interesting to see what you come up with for the 1D-C James.

Cheers,
Pete
Click to expand...

The 1D-C seems more clear, and doesn't seem like we can get away with quite as much in older dslrs (we are talking fractions here). I am just surprised the neutral and standard don't perform as well as they could. They look great at first, so there is potential there, imho.
 
Skin tone:

I have probably mentioned this before, but skin tone is still is quite limited in 8bit. Getting perfect skin tone for one type of skin color under a certain type of White Balance bias is difficult enough, let alone claiming to have "perfect" skin tone in general is a gimmick, and wrong. Even saying "better" is wrong.

Not only is skintone illusive in the camera, it is also a culturally influenced thing. DO you want to err on warm or white? Does the skin of your subjects tend to be more yellow or red? what do we wish it to be more like. It is not only what is the truth, but what we want something to be.
 
Some quick comments:

* I judge skin tone accuracy using the skin-tone boxes in my spydercheckr:
http://www.vimeo.com/47394255
When those squares are close to their target, people look nice. If, to achieve that, you have to push your image a lot, you have a problem. Additionally, if that push makes other colors look bad, you have a problem.

* I always run these tests for my own use, which means I use my best lenses: vintage Leica-R from the 60s and 70s; some of them happen to have a very slight green cast (which doesn't affect all tones, and which I don't find unappealing at all).

* On Canon DSLRs, you can control the green-magenta bias using the Tone parameter. I like it at 1, but if that's too green for you, you can dial it back to 0 (which I think it makes people a bit too magenta; maybe something like 0.6 or 0.8 would have all of us happy, but that's not an option).
 
It is important to note that none of the built-in picture styles is color correct. This includes Neutral and Faithful, which are generally supposed to be more "real world" than the others. This is quite wrong though. Not only are these two oversaturated, but also the colors are noticeably off. This is not a bad thing either (although I would love a color perfect profile): I actually like Faithful a lot, and even apply it to stills in post. Skin is slightly biased towards orange-green in Neutral and Standard, and towards pink in Portrait and Faithful, so as Samuel points out, using the Color Tone parameter at -1 or +1 may be useful (sometimes depending on the subject). Using higher (or lower values) is not a good idea as this also spreads a bit of color shift on the neighboring hues, and not just the skin tones as advertised. One other thing, the skin hue bias of the picture styles also appears to be non-uniform with respect to saturation. It seems to be somewhat more pronounced with more saturated skin tones.

Incidentally, I am currently using a Nex-6 for a small travel camera, and one of the two annoying things about the otherwise beautiful Nex-6 is the lack of decent blues (for my taste). Every single camera profile pushes them a bit towards magenta. This is in contrast with Canon cameras where every profile pushes blues towards cyan (which I pefer over the magenta bias, as it is a better complementary color to skin tones). Skin itself is ok though.
 
cpc said:
It is important to note that none of the built-in picture styles is color correct. This includes Neutral and Faithful, which are generally supposed to be more "real world" than the others. This is quite wrong though. Not only are these two oversaturated, but also the colors are noticeably off. This is not a bad thing either (although I would love a color perfect profile): I actually like Faithful a lot, and even apply it to stills in post. Skin is slightly biased towards orange-green in Neutral and Standard, and towards pink in Portrait and Faithful, so as Samuel points out, using the Color Tone parameter at -1 or +1 may be useful (sometimes depending on the subject). Using higher (or lower values) is not a good idea as this also spreads a bit of color shift on the neighboring hues, and not just the skin tones as advertised. One other thing, the skin hue bias of the picture styles also appears to be non-uniform with respect to saturation. It seems to be somewhat more pronounced with more saturated skin tones.

Incidentally, I am currently using a Nex-6 for a small travel camera, and one of the two annoying things about the otherwise beautiful Nex-6 is the lack of decent blues (for my taste). Every single camera profile pushes them a bit towards magenta. This is in contrast with Canon cameras where every profile pushes blues towards cyan (which I pefer over the magenta bias, as it is a better complementary color to skin tones). Skin itself is ok though.
Click to expand...

I just recently got to use a friend's NEX7 for a little while. It is nice to have the EVF and everything in such a small form. And I agree that cyan bias is nicer than magenta bias in the blues. Though, I dislike the general trend of cyan. Even though it looks nice, I dislike it in my own images to have this phenom. I like blue to be blue.

I like this one (ungraded), because it is closer to the real look, and also the near navy blue feels darker, colder, and more majestic than cyan.
k2.jpg
This altered one below looks nice too, but I do not like it, maybe different context.
The cyan is a nice color, but there is no weight or depth.
k2copy.jpg
Another thing with cyan for me, is it is difficult to build on recent nostalgia, because I do not see that color everywhere. But if I am going on pure fantasy and aesthetics (AKA transformers) then I can readily except broad brush applications of the cyan/skin phenom. Other wise it is situational use only.

I do not understand baking in this cyan bias, because it requires 1 extra minute to apply a hue shift to the blues in post. yet fixing the blues back can be near impossible without effecting cyans and greens.

Below, I have no idea what profile I used, or what the white balance is, or what my settings were, but the skin tone is fine for me. I like the slight cold/warm mood for this shot. (ungraded)
k1.jpg
It looks like Neutral -2 contrast to me.
 
Last edited:
It is certainly better to have correct blues. But if the camera will bias blues, I'd rather have the cyan bias, than the magenta bias, especially when there is skin as a counterpoint in the image. This certainly does not mean one should discard blues with a bit of magenta in them.
There is a deep connection between cinema and cyan and it goes back to film. Color film can have much deeper cyans than DCI, and DCI can have much deeper cyans than Rec709. This is probably one reason cyan shades appear thin in your example. It is also not necessary to go to in-your-face extremes ala Transformers, with their saturated orange skin countered with teal. A very pleasant and subtle combination is only very slightly tinted blues with neutral skin.
In your last example I believe the walls have an effect on the look of skin. They seem to color the ambient with their dark beige tint, which mght have affected skin.
 
We are probably more in agreement, than not. I may have been trying to counter the overall broad acceptance of the cyan/skin rule, as well as the whole notion of 'getting great' skin tone.
cpc said:
It is certainly better to have correct blues. But if the camera will bias blues, I'd rather have the cyan bias, than the magenta bias, especially when there is skin as a counterpoint in the image. This certainly does not mean one should discard blues with a bit of magenta in them.
There is a deep connection between cinema and cyan and it goes back to film. Color film can have much deeper cyans than DCI, and DCI can have much deeper cyans than Rec709. This is probably one reason cyan shades appear thin in your example.
Click to expand...

Yes, this is digital. No use talking about film. So few places have film projection these days. Regardless of what format is used, cyan cannot give the emotional context of navy blue. Though, I am the same, I rather the blue to err on cyan than magenta. Even still, I much prefer to have a blue and cyan represented separately. And adjust in post as needed. And when cyan is needed, actually photograph cyan. But that again is my preference. So many rules, and ideas floating around. They are good, but get to the point they distract.

cpc said:
It is also not necessary to go to in-your-face extremes ala Transformers, with their saturated orange skin countered with teal. A very pleasant and subtle combination is only very slightly tinted blues with neutral skin.
Click to expand...
I like a bit of variation, and occasionally I really like the over the top color grading. Though most of the time I like the same as you, pleasant and subtle combinations.


cpc said:
In your last example I believe the walls have an effect on the look of skin. They seem to color the ambient with their dark beige tint, which mght have affected skin.
Click to expand...
Yes, the room was warmer, but still that cold warm. I guess my point about that was skin tone is culturally subjective within context. If it feels right, is as important as any other rule. In North America, the trend is to make skin more warm and less magenta when working with certain white skin types. But if one picture style helps those skin types, it will also apply the same warming to other skin types that do not need it, or that would be better with more red/magenta added. Skin tone is not going to be perfect for all types of people with 8bit h.264. It isn't even perfect with film.

The other thing, is if one picture style warms the skin, it can produce lively looking skin with even a cooler white balance. While another picture style might cool the skin allowing a warmer white balance. It is all subject to the end goal. Some picture styles look like crap without a WB shift, and kids with T2i's are using them. (The t2i doesn't have a proper white balance shift in video mode). The result is nasty.
 
I've started playing around with picture styles I'm the 1dc, but won't be able to upload any video for a month. Limited internet access.
 
I was shocked to find that a user found Flaat useful on the 1DC. I thought once you had a true log profile, all other options would be obsolete. As it turns out, the 8-bit codec creates banding, and something like Flaat is still useful.
I think he posted a couple of tests on the Shot on Flaat vimeo group.
 
Samuel H said:
I was shocked to find that a user found Flaat useful on the 1DC. I thought once you had a true log profile, all other options would be obsolete. As it turns out, the 8-bit codec creates banding, and something like Flaat is still useful.
I think he posted a couple of tests on the Shot on Flaat vimeo group.
Click to expand...

Good picture styles whether in or out of camera will always be useful and appreciated, no matter 8bit, 10bit, RAW or compressed. ;)
 
The PSE has been updated. Seems a bit better than before.

I recently started using the PSE again. Trying to make something for the 1D-C, since all of the default Canon picture styles are not great. C-log is good, but not for everything.

Has anyone been using the new version of PSE? Any comments?
 
Samuel H said:
Interesting. I'll have a look.
Click to expand...

I've only been exploring the curve adjustments, so far. Haven't had time to do any serious color adjustments, so left them for another day, and just focusing on curves.

The curve adjust tool is better, since small adjustments are far easier to make, but still has much of the same quirks as the original.
 
(EDIT: C-Log is underexposing to get those highlights back. As a result is losing almost an equal amount in the shadows. Still exploring the pros and cons)

Started playing around with picture styles again. Just some preliminary "in-the-wild" tests.

Trying profiles that I had left alone before...
Flaat2
LightformN
Pancake

Pancake, I have not done much with.


DYNAMIC RANGE:
EDIT: With C-Log highlight recovery is the farthest reaching. But this was not a shadow test. And for those that like "Portrait", the color palette is quite different. Things tend to be more golden and at times slightly on the green, rather than the rosy extra warm plus cyan look of portrait.


This image shows the highlight capture between three different picture styles. LightformN is recovering practically the same as Neutral (0,-2,-2,0), but Lightform does not have true white. Its highlights are artificially crushed, meaning the info is there, but darker than it should be on a gradient, so there is less of a ramp to blow out. (i can explain the exact process if needed). It will look like it is preserving more highlight detail, at first, but once you start grading the files next to others, you'll find it different. Also, despite it being "Lightform Neutral" it is pushing the yellows and reds torwards a pleasing skin tone, so your yellows will be warm, and your reds a bit more orange. Just an inherent trade off. Lightform could be better liked, just the marketing language is misleading.

But this test seemed hand picked for C-Log, making it the clear winner of highlight recovery.
9184964818_f2df42f8d2_z.jpg


Flaat2 did not recover any more highlights than the others, but it digs deep into the shadows. But this test didn't really have a good shadowy subject of interest, and Lightform claims to recover highlights.


Neutral with contrast at -2, and saturation at -2, is one of my go to picture styles. Allows some room for grading, but still has enough contrast to not grade. I call it "half lost" after the "Prolost" Nuetral (0-4,-4,0). It always seems to get close enough to the "flat" picture styles in the highlight recovery.
 
Last edited:
(This was a highlight test. Turns out C-log was underexposing to recover these amazing highlights, but in turn losing some shadow detail. See below)

Comparing Half-lost with Lightform in photoshop....


9183017261_48f76a024f_z.jpg



Keep in mind, that Lightform is at -4 contrast, while the neutral profile is at -2 contrast. So, it should have a slight edge in highlight recovery. But they are very close. But look at how the tonal range of Lightform is limited. It is because the highlights are hard smashed to look darker, and it is throwing away gradation in favor of 'pixel information'. While lightform is not the best on a technical level, there are some that use it well. I will compare it with Prolost another time to see Neutral at -4.

Points of interests:

- Look at the gradation of the couch top. The top is blown out. On the Neutral image, it is a smoother ramp into the clipped top. Whereas the Lightform is very abrupt. Little gradation and gray discolored areas.

- The shirt has only slightly more info on the highlights, but again, less gradation and more blocks and discoloration.

But no one would grade this way, this is a torture test. So, this would not be the end result, it is just showing how these picture styles work.


Here is C-Log:
9185230744_c9f697bbbf_z.jpg


I didn't quite match the contrast with this and the other two, but it is easy to see that there is real detail there, and much smother gradation into the clipped part of the image.
 
Last edited:
Recently, I have been on a quest to find a picture style with the best highlights, and hopefully preserve some shadow detail where possible, without messing with gradation too much.

Initial testing showed that C-log was recovering some amazing amount of highlight detail. But as it turns out, C0log was underexposing. Meaning the shadow information was being lost. Not noise, not just compression, but actually not there. Just like the other picture styles did not have the same highlight information that c-log magically had.

9297031029_4460b3bb14.jpg

In the picture above, I shot the same scene with and without C-log, then I pushed the images in post to see what kind of shadow detail was there. C-log at the same exposure as Neutral, has very little detail in the hairline, whereas Neutral has just enough.

What this means is, that other picture styles may actually not be that far behind C-log in latitude.

Not sure all the actual technicalities, but proof is in the pudding.

Sorry, for any incomplete information prior, as no one seems to be mentioning that c-log is underexposing compared to default picture styles (and the actual raw!), and I was primarily interested in the highlights since that is where most digital cameras suffer. I wrongfully assumed that c-log was only weaker in the shadows, but didn't realize I was losing so much as to have to overexpose by at least a stop to get that shadow information back.
 
Last edited:
The Cook got the first Perk on Indiegogo, now the campaign will be online until the end, because after the first Perk the campaign keeps online and the final goal cannot be changed.

You can do anonymous perk, nobody will se your name or email. Also you do not need to inform shipping address.

There will be no individual sales if the campaign does not reach the goal. It is just 200 perks to reach the final goal, so, if you want it, do your perk and trust other people will do also.

Thanks
 
You must log in or register to reply here.
Back
Top