Would anyone like to buy my A7S

[jcs]

Ebrahim- there are lots of 5D3 RAW (5D3R) vs. <insert camera here> tests out there. Here's a pretty good one comparing 5D3R to Kodak film and Red Dragon:

 

5D3R has fantastic skintones and a very filmic look. Skintones look better than Red Dragon (at the time of the test- Red is constantly improving their color science: skintones look better now).

 

Keep in mind ARRI and Canon colors aren't necessarily chart accurate (though ARRI is both more accurate and produces nicer skintones). Part of the secret to good looking people/skintones results in making other scene colors off a bit. Here Shane Hurlbut prefer's Canon's C500 'golden color bias' over Alexa's color (to my eye Alexa is more accurate and could easily be made to have the 'golden' look of the C500 if desired):

[Inazuma]

But surely if you have a colourist of any skill (and you probably do have the budget for one if you're using Arri's and RED's) then you can grade the colours to whatever you like?

[Tim Naylor]

I tried these settings from Kholi Hicks for my A7s. Definitely a big improvement for grading accurate skin tone. Restored much hope in my A7s.

 

Black Level: 0

Gamma Slog: 2
Black Gamma > Range: Middle , Level: -3
Knee > Mode: Auto , Auto Set > Max Point: 95 , Sensitivity: Mid
Colore Mode: Pro
Saturation: +8
Color Phase -2
Color Depth > R+1, G-1, B+2,, C-2, M-1, Y-3
Detail > Level, -5 Adjust > Mode; Manual, V/H Balance: -2, B/W Balance Type3, Limit: 0, Crispening:0, hi_light Detail: 0

White balance > Manual > Grid setting B 1 (one to the left)

3000-3600 = Tungsten
5000 - 5600 = Daylight
6100 - 7000 = late day or shade/shaded interiors in daylight

 

It gives a very juicy and gradeable flesh tone.

[jcs]

Inazuma- if it were that easy productions would use the cheapest cameras that get the job done- such as Red Epics, then fix the color in post. When the budget allows, they use ARRI cameras (there are exceptions, though interestingly most Oscar winners from last year were shot on ARRI. Looks like this year too.) as fixing Red (et al) color in post is expensive and time consuming. It turns out that getting great looking color from a camera is hard, perhaps there are patents impeding competition too. If it were as easy as color correcting in post, they could build the tech into cameras. With so many folks studying ARRI's color science, apparently reverse-engineering their technology is hard, as no one has matched ARRI yet. It's clear that both ARRI and Canon do quite a bit of special processing to make skintones look great over clinical accuracy. As an artistic tool, the immediate emotional reaction after reviewing footage can effect the edit moving forward.

 

For those shooting mostly landscapes, plants, animals, and tests, resolution and dynamic range can be more important. When shooting people, color is by far the most important. Human beings naturally detect when skintones are off: we evolved color vision in part to detect health and emotion in skin tones.

[jcs]

Tim, I tried Kholi's settings (basically PP7 with Color Mode Pro and minor tweaks) and other variants, however I'm finding PP6 with tweaks looks best for skintones in LED ~5000K light:

 

PP6

 

Black Level: -7

Gamma: Cine2

Black Gamma: Middle, Level 0

Knee: Auto

Color Mode: Cinema

Saturation: +3

Color Phase: 0

Color Depth:

R: -7

G: -1

B: 0

C: 0

M: 0

Y: -3

Detail, Level: -7

 

When shooting in other lighting conditions, I check a variety of picture profiles to see what the camera is doing with the available light. To make the A7S look its best, it's necessary to tweak setting for lighting conditions, including a custom WB (shot with gray card). This is also true for other cameras not shooting RAW, however Canon and Panasonic have better general purpose settings than Sony. Stock PP6 is pretty good as an all-around profile for the A7S. My tweaks were to make skin look it's best and most filmic at ~5000K (CREE LEDs) by viewing the subject live on an HDTV while making changes. I first cranked color saturation way up, then tweaked the Color Depth to make the oversaturated state look the most pleasing, then put saturation back at a practical level.

[maxotics]

JCS I don't have near you experience with this stuff, but have a strong opinion nonetheless ;)  I believe the answer to Inazuma's question is that ANY image that has been debayered from RAW sensor data (like the ARRI and RED I assume) has applied algorithms that permanently set the color malleability of the image.  What Inazuma may not understand is that RAW sensor data is "RAW" in more ways than one.  That is, it isn't a simple matter of, say, combining 3 adjacent color values into one (R,G, B).  First, there are multiple types of debayering algorithms, as you pointed out, but it's more than that.  Each sensel is not equally sensitive to all brightness levels of light.  That is, the analog to digital converter "curves" the data to fit an expected range of values.  At each wavelength of light, it may make different adjustments. There are matrixes of constants that are applies to different ranges of values depending on ISO expectations, etc.  

 

One can get an idea of complexity involved in interpreting RAW data into a usable image here:

 

http://lclevy.free.fr/cr2/#interpol

 

The in-camera settings discussed above, is child's play to what you need to make really good images from RAW data.  Or what kinds of adjustments must be built into a camera to deliver good skin tones.

 

5D3 RAW gives you data before any assumptions are made.  That's why it can deliver an image to match the ARRI with the right expertise.  "The right expertise" is easier wanted than got ;)

 

As you say, this expertise if baked into the firmware of the ARRI, which is why that camera is preferable to the Sony.  Why the RED would be preferable to the SONY.  But again, with 5D3 RAW, you have all the data you need to create exactly the image you want!

[jcs]

maxotics- it's true that RAW is the most maleable in post, however even 5D3 H.264 has better skintones than the A7S and GH4. Also note the FS700, F5, F55, all shoot RAW. When RAW is compressed, it's typically wavelet-based, which won't really affect color (a slightly softer overall image at most- zero macroblocking). It's clear that the RAW data has already been shaped by the manufacturers' color science when we compare the RAW color output of all the RAW cameras.

 

In preparation for my next project I've been working with the A7S, GH4, and 5D3 (both H.264 and RAW) to find out how to make each camera look the most filmic (#1 being pleasing skintones followed by highlight/shadow behavior similar to a film response). For closeups, the softness of the 5D3 H.264 is actually a benefit.

 

The 5D2 and 5D3 are known for a great filmic look and superb skintones shooting H.264. If we search Shot on What for GH2,3,4, A7S, 5D etc. we see that the 5D has been used on quite a lot of films vs. the other cameras: http://shotonwhat.com/?s=5D

 

A cool 5D3 RAW video showing landscape and skintones:

Great example, the only thing missing is slomo (where the GH4 or A7S could fill in).

[maxotics]

maxotics- it's true that RAW is the most maleable in post, however even 5D3 H.264 has better skintones than the A7S and GH4. 

 Interesting video, but out in the bright sunlight it's not easy for me to see the difference between RAW and H.264.  Late in the day, cloudy, or indoors, is where I gravitate to the RAW look.  Anyway, so you think you can get better skintones with 5D3 H.264 than a good grading of s-log from the A7S?  Also, might it not be that you, and others, are just more used to working with Canon video?  That in a year or so the Sony will be on par, maybe better?

[jcs]

maxotics, here's a an early test with 5D3 RAW shooting very high contrast in bright sunlight (inside shooting outside to the street):

 

At the time of the test I didn't have all the tools to preserve skintones at the same time as taming the background (via masking, etc.)). Even so, as a beginning user of RAW we liked how filmic it turned out. For the interior, natural lighting was used and no makeup was applied to the manager. Skintones for both turned out well for the conditions. For the inside to outside shot, the A7S would likely do better (more DR). I'm curious how much as 14-bit RAW with ACR has a lot of gradations for shadow/highlight manipulation (16384 vs 256): as long as highlights and shadows aren't clipped, it would be able to match or exceed what is possible with the A7S.

 

I purchased a Sony FS700 right after the SpeedBooster came out and have been grading Sony footage for a while now.  Whereas I can match cameras and color fairly well in tests with the 5D3, skintones were a whole different matter. With careful exposure and WB, the FS700 can do decent color, however it's fragile for skintones: typically magenta/green 'instability'. In some lighting conditions, the FS700 is quite a challenge for skintones (in the shade at sunset is one example).

 

Here's an example with 5D3 RAW at sunset/golden hour, then the rest is FS700 (interior).

While the FS700 skintones aren't bad, the 5D3 RAW skintones are much better (golden hour light also helped).

 

Reviewing this old footage while studying movies shot on film reminds me how much more filmic 5D3 footage looks compared to newer cameras (including the Canon Cx00 series). It makes sense why the 5D was used on so many feature films.

 

When using the A7S or GH4, having the 5D3 around to take shots for color reference is also helpful. Models and clients always love how the 5D3 makes them look: it's helpful to understand how to make other cameras look similar.

[jax_rox]

maxotics, here's a an early test with 5D3 RAW shooting very high contrast in bright sunlight (inside shooting outside to the street):

IMO, it looks very similar to what you could do relatively easily with an A7s.
 

Even so, as a beginning user of RAW we like how filmic it turned out. For the interior, natural lighting was used and no makeup was applied to the manager. Skintones for both turned out well for the conditions. For the inside to outside shot, the A7S would likely do better (more DR). I'm curious how much as 14-bit RAW with ACR has a lot of gradations for shadow/highlight manipulation (16384 vs 256): as long as highlights and shadows aren't clipped, it would be able to match or exceed what is possible with the A7S.

To my eye, it looks far from filmic - doesn't even really come close to an Alexa; maybe similar to a C100. It looks like a 5D with less compression!

In regards to DR vs 14-bit - you can have as many gradations as you want, it doesn't make up for dynamic range. If you light your scene to within your dynamic range (which is something that every DP has been doing for decades - one of the main tests you would do with film was to test your latitude to see what you have to work with), then you can get good results from any camera. But if you don't light it, then extra dynamic range will always be better. If you don't clip the highlights or shadows, then your scene doesn't have enough dynamic range in it for it to matter, and in effect is the exact same as lighting to within your dynamic range.

[jcs]

IMO, it looks very similar to what you could do relatively easily with an A7s.

Right, the A7S could possibly do better with more available DR, however the skintones won't look as good. Just as with real film, we light for faces/skintones and make sacrifices for hightlights/shadows as needed. IIRC, the A7S's sweetspot for skintones puts more useable DR in the shadows vs. the highlights, so it may not do better than the 5D3 for this condition.

 

To my eye, it looks far from filmic - doesn't even really come close to an Alexa; maybe similar to a C100. It looks like a 5D with less compression!

What elements don't look filmic? Do you have examples of your Alexa work you can share showing how and why it looks more filmic?

 

Regarding looking like an Alexa: there are many examples showing 5D3 RAW comparing well to Alexa footage in the exact same conditions. Do you have examples showing otherwise? (I would expect an inside to outside window shot to favor Alexa (DR)).

 

Here's a tutorial and example footage for converting 5D3 RAW into an Alexa rec709 look:

 

 

Regarding dynamic range- I think we both covered the same points in the same way (clipping, optimization), however mathematically 14-bits (16384 steps) provide for more color and luminance manipulation in post before artifacts appear due to color compression (from e.g. a curve slope) and quantization. Amazing things are possible with ACR highlight/shadows and 5D3 RAW:

There's more detail in the shadows- I crushed them to focus attention on the sunset. I'm not sure if the A7S could do better for this shot (perhaps a slightly smaller sun clip area. If the shadows needed more light, the A7S would do better, but again, was not needed for this shot).

 

Ideally we'd want to use all available DR for every shot, however that's not always possible (and rarely possible outside of a studio if we're optimizing exposure for skintones).

 

From all the tests I have viewed, 5D3 RAW and Alexa can perform similarly except in extreme conditions for color (Alexa can record 2.8-3.4K resolution (Amira will do a scaled up "4k")). 5D3 RAW even compares well to film (see the Atlantis video in this thread). At $40+k for an Amira, no one is arguing that 5D3 RAW (or A7S) matches or exceeds ARRI, however in many conditions it's amazing that it comes close enough to easily cut between cameras.

[Inazuma]

I really think skin tones is another of those things that DOPs fuss over but no one else notices really. I wasn't even aware of such an issue until the gh4 came out earlier this year and everyone started talking about the subject. I think Canon RAW gives really nice tones but as I said in an earlier post, the tones from h264 seem to have a plasticky look and makes it easy to point the footage out as being Canon, which is distracting. It's just a matter of taste I suppose. I mean I've seen several RED v canon videos and always prefer the tones from the red.

[sunyata]

Inazuma- if it were that easy productions would use the cheapest cameras that get the job done- such as Red Epics, then fix the color in post....

 

Human beings naturally detect when skintones are off: we evolved color vision in part to detect health and emotion in skin tones.

 

jcs- Inazuma makes a valid point, color is also not the only or primary difference between those examples.

 

I believe the answer to Inazuma's question is that ANY image that has been debayered from RAW sensor data (like the ARRI and RED I assume) has applied algorithms that permanently set the color malleability of the image.  

 

You can alter de-bayered Alexa footage in post quite a bit because of the image quality and film-like sensor noise: it grades quite nicely. Also, because of the log-c profile, it's analogous to working with film, I'm sure that was one of Arri's goals when designing the Alexa. Realistic or idealized skin tones are also rarely the goal in developing the look for a show. I'll post some archived uncompressed Alexa examples to illustrate as soon as I get a chance, just finished doing some crap for Sons of Anarchy, which coincidentally was shot on an Alexa (with other cameras mixed in, even a GoPro this season) and very not-glam. 

 

Our eyes evolved to sense color for health and emotion in skin tones what??? Please continue. 

[jcs]

 

Our eyes evolved to sense color for health and emotion in skin tones what??? Please continue. 

 

http://www.sciencedaily.com/releases/2006/03/060320221839.htm

[sunyata]

Very interesting article.. "Also, ecologically, when you're more oxygenated, you're in better shape."  Maybe I should start hyperventilating when I'm around someone I'm trying to impress?

[hmcindie]

it wouldn't surprise me if you showed a side by side clip to a regular audience member or client, they may very well prefer the 5d3. Most humans don't care about resolution. They do sense when colors are off.

 

That's actually whats happened to me. A couple of friends picked the 5d from two different films I did saying "A7s looks really good!" while I was playing back 5d material. 

 

Placebo man, it's really strong. Wait till you go to hifi- forums like head-fi. 

 

Though I did finally do a spot where I really liked the a7s. It was a parody of product commercials and the slightly off skintones of the a7s fit it perfectly xD

 

And I was using Kholis settings!

[hmcindie]

doesn't even really come close to an Alexa; maybe similar to a C100. It looks like a 5D with less compression!

 

As someone who has assistant edited a feature film shot with AlexaRAW, I can say that the 5d RAW is not that far off. I actually think it's absolutely amazing that 5d RAW is available for people (and that it gets so close to AlexaRAW) for the price it is.

 

AlexaRAW takes a humongous amount of space and needs separate capture devices (we used Codex). 5d RAW records into 1000x CF cards. THAT'S CRAZY! It's also funny how fast USB3 can be compared to the Thunderbolt / Codex combination. We get material just as fast into our harddrives from both cameras.

 

Yeah the 5d is worse especially in highlights but man... the difference is way, way less than people say in forums. It's like being in Hifi- forums. People are saying things like "blows away!" and "night and day difference!". Then when you actually listen side-to-side, the difference is... very small. Something like that happens with the 5d RAW vs Alexa. Yeah, there is a difference... but in most situations it's moot.

[/p/]

I just shoot in PP4 for the best skin tones

[mojo43]

I have been shooting in cine4, pro colour mode at a higher saturation and my skin tones have started to come out well now. It's really just a matter of getting the white balance right I think. The 5d raw stuff that has been posted is hit and miss. I think we all have a different idea of what a "good" skin tone is. Some of the stuff posted looks off to me...

[jax_rox]

What elements don't look filmic? Do you have examples of your Alexa work you can share showing how and why it looks more filmic?

It looks nothing like film. It looks to my eye like a C100. If someone showed me the footage, I would assume it had been shot on a C100 or C300.

I tried posting some images, but the forum won't let me? :/

Anyway, there's a roll of with Alexa highlights, it's not straight clip - on the 5D, even the highlights that are close to clipping are not rolled off in that way, and you end up with slabs of pure white, or near pure white. Plus, I find the Alexas colours to be more filmic - maybe I wouldn't say 'more accurate' but they're more accurate to how some film stocks would see the colours in the shot. The skin tones on Alexa are 100x better than skin tones on the 5D. I find skin tones (and colour in general) on 5D to be way too warm, and often at the expense of the other colours in the scene.
That wildlife footage is quite nice, but in general, I've never seen colours from a 5D that I've absolutely loved..
They're better in raw than H.264, but not that much better
Plus, I would say the Alexa has more colour gradation - it might not have the same bit-depth, but there's more colour information there, and there seems to be more shades (comparing, for example, hair).

To my eye, the Alexa just looks much, much better. That's not to say the 5D isn't useable, or that it doesn't create good looking images. There are a number of issues in the 5D footage that would not be present
Just that in general, the Alexa is a better camera with a better image... which is to be expected as it costs a lot more.
 

Regarding looking like an Alexa: there are many examples showing 5D3 RAW comparing well to Alexa footage in the exact same conditions.

In your video above where the 5D is compared to what I assume is 5213 and the RED Dragon, the 5D easily comes off as third best. Keep in mind, that the test is lit to within those cameras' dynamic ranges.
If you remember back to the Zacuto shootout, where a whole bunch of cameras were compared, you can see that if you light to a cameras dynamic range, you can get similar looking images out of any camera.
That doesn't mean a 5D is the same as an Alexa.
 

Regarding dynamic range- I think we both covered the same points in the same way (clipping, optimization), however mathematically 14-bits (16384 steps) provide for more color and luminance manipulation in post before artifacts appear due to color compression (from e.g. a curve slope) and quantization.

Yes - but that is not the same as dynamic range. In practice, it's all about useable DR, and it's theoretically possible that a camera with less DR but more information captured can have the same useable DR as a camera with more DR but less information captured. I'd be interested to see someone compare useable DR on both cameras.
 

Amazing things are possible with ACR highlight/shadows and 5D3 RAW:

There's more detail in the shadows- I crushed them to focus attention on the sunset. I'm not sure if the A7S could do better for this shot (perhaps a slightly smaller sun clip area. If the shadows needed more light, the A7S would do better, but again, was not needed for this shot).

That sunset is beautiful. But man are the colours off here. Not sure if that's a result of grading or not..
 

Ideally we'd want to use all available DR for every shot, however that's not always possible (and rarely possible outside of a studio if we're optimizing exposure for skintones).

Why would we? I certainly don't want to use all available DR! I want to light my scenes to optimise them for the dynamic range of the camera - which is part of why I do camera tests. I don't necessarily want something exposed in every step of the DR! It depends on the scene - perhaps you mean that we want to keep the lighting contained within the available DR of the camera, but doing so is not always possible..?
 

in many conditions it's amazing that it comes close enough to easily cut between cameras.

See, I don't personally think they do cut all that well and don't believe they ever have - though there are many who can't pick the difference even when comparing H.264 5D footage in big movies.

I think the most important point to make is that any camera will give you a pretty good image these days if you know how to work it..
So get out there and shoot, rather than sitting talking about specs all day on the internet ;)

Use what you like, what you have, what you can afford. The less time you spend attempting to justify your choice and attempting to prove that a 5D is as good as an Alexa, the more time you have to actually shoot :)