The correct way to expose for SLOG3 when using 8bit cameras

[maxotics]

9 hours ago, mercer said:

 I feel shows a little of that Raw texture in the tree bark.

Agree, beautiful look!

[kidzrevil]

14 hours ago, maxotics said:

Yeah, and I just want to say, I love what you're doing artistically too!  But...

I think I understand.  I get bent out of shape because there's a fair amount of snake-oil sold out there in the way of color profiles, etc (which I know you're not doing) which actually hurts the footage that new filmmakers take.  This is a very interesting subject to me because, like you, I LUST after a low-contrast look.  Nothing beats a RAW workflow in my book.  It's just beautiful.  But I'm lazy, like everyone else, and want a short-cut.  So I've tried, like you, to make LOG work.  But I couldn't for most situations.  After some experimentations I figured out why it is impossible.  But in the back of my mind, I'm still hopeful.  So when I read your post I'm like, maybe he's done it!  But I need to see certain things addressed to be convinced.  So, all good.  I'm enjoying this!

I can only imagine how good a RAW workflow is. After seeing @Andrew Reid RAW footage and hearing @mercer love for the hacked 5D im really considering picking one up. If it's anything like the still images that would be a joy to grade ! Like you I do see some of the problems with LOG and more importantly the LUT's people use to grade them. Some of these are merely look LUT's and they completely destroy the image. Even some of the popular ones like IMPULZ lut's adds a ton of noise to the blue channel which is inherently noisy on most cameras. I most likely will not commit to a LOG workflow especially on important shoots because it is like rolling the dice on an 8 bit camera. Trying to change the white balance in post is a nightmare so exposure and WB is crucial (which most miss relying on auto exposure and auto white balance). I ran this test to dissect how the Sony A6500 processes the image but so far I can say the LOG is usable but only for advanced users. I for one will not commit to a LOG workflow and I am experimenting with the cinegammas which I will make a thread about when I run more tests with them.

 

@mercer I forgot to answer your first question. The RX10 ii is a great camera, I've always used mine with a tiffen pro mist diffuser when I had it. Have you seen the RX10 IV ? Looks promising

[jax_rox]

On 29/10/2017 at 8:13 AM, maxotics said:

I'm one of those people who are very critical of LOG profiles. 

 

Log allows you to capture a higher dynamic range than you could otherwise with REC709. It also apportions equal(ish) amounts of data to each stop, rather than the linear doubling you get in REC709. You therefore are able to capture 14 or more stops (depending on the range of the camera) whereas REC709 will only capture 7 or 8. 

To display that, you have to grade so that it will work on a REC709 monitor. That means throwing away information, sure - but you can pick where you place it. Log colour space is very good for versatility in the colour bay. If you don't have the time, knowledge or budget for a grade, you'll likely get results easier shooting straight to REC709. Many lower-end users jumped on log because it was this high-end camera thing that had some magical power to make your images look like a Hollywood film, so people still use it even when they have little idea why or how to.

On 29/10/2017 at 8:31 AM, Mattias Burling said:

My problem with slog3 was never the grading. It was the banding which is there in camera and after.
Nothing to do about it really. 

And I wouldn't mind if it wasn't for the A7sii I had costing more than my complete Red package, a RX1rii + bmpcc combo, way more than a Leica Q and even more than my complete DSLR setup  including great glass that I use today.
If the camera was cheap.. then it would be ok imo.
 

8-bit will do that to you. As for price, don't know where you're getting a 'complete' RED package for $2500, unless you're talking a secondhand RED One kit with a large number of hours on it.

19 hours ago, maxotics said:

Okay, so in LOG, it doesn't take 1,000 colors in that blue range say, it takes 250.  Normally, one doesn't notice.  But if you shoot a scene where there is a blue gradient sky and it shows it to you in 250 shades of blue, they will look like DIFFERENT COLORS next to each other.  That's BANDING!  Of course, they are different colors, but you don't want them so different that the brain says they're not connected!

You'll almost certainly run into banding with any 8-bit camera. Banding has more to do with bit-depth, and available colours to accurately display, say, the sky, than it does with log vs lin per se. Log in 8-bit though doesn't help.

3 hours ago, kidzrevil said:

 I most likely will not commit to a LOG workflow especially on important shoots because it is like rolling the dice on an 8 bit camera.

Yes and no. Depends what you're shooting, what tests you've done and how much knowledge you have of the curve and colour space you're using. It doesn't have to be a roll of the dice, but part of that is knowing when is the most appropriate time to shoot log.

[Shirozina]

8 bit is not the cause of banding - it's the high compression 4.2.0 internal codec. 8 bit HDMI out to an ext recorder won't band unless you really abuse it ( and that's the only way I use S-LOG on my Sony's). 8 bit (RGB)Tiffs generated from RAW stills are very hard to produce banding in unless you do very aggressive tonal changes  but output a highly compressed jpeg from the same RAW and it can have banding even without touching it. Also with a video camera we are not talking RGB but YCbCr which should be better than RGB in this regard as most of the data is in the Luma channel but due to the massive compression found on internal codec's it bands oh so easily. 

 

[cpc]

Banding is a combination of bitdepth precision, chroma subsampling, compression and stretching the image in post. S-log3 is so flat (and doesn't even use the full 8-bit precision) that pretty much all grades qualify as aggressive tonal changes. S-log2 is a bit better, but still needs more exposure than nominal in most cases.

Actually, I can't think of any non-Arri cameras that don't need some amount of overexposure in log even at higher bitdepths. These curves are ISO rated for maximizing a technical notion of SNR which doesn't always (if ever) coincide with what we consider a clean image after tone mapping for display. That said, ETTR isn't usually the best way to expose log (or any curve): too much normalizing work in post on shot by shot basis. Better to re-rate the camera slower and expose consistently.

In the case of Sony A series it is probably best to just shoot one of the Cine curves. They have decent latitude without entirely butchering  mids density. Perhaps the only practical exception is shooting 4k and delivering 1080p, which restores a bit of density after the downscale.

[EthanAlexander]

I've started combining SGamut3.Cine with SLog 2 on 8 bit. SGamut3.Cine is much better suited both for 8 bit and the cheaper color filters that go into Sony's below F55, and S Log 2 makes better use of the entire range of code values. S Log 3 is designed for 16 stop sensors, and cuts off at 94IRE on current 14 stop sensors. Not a big deal on 10 bit but important on 8.

Have you tried this combo, @kidzrevil? One cool thing is that using standard SLog2 to 709 LUTs creates some very "cinematic" colors straight off the bat. I've started doing this even shooting 10-bit.

[kidzrevil]

@EthanAlexander funny you mention it because that was my next test ! It works REALLY well and doesn’t have that awful green tint of s.gamut. It may end up being my go to profile for the a6500  I just have to test it a little more

[Mattias Burling]

8 hours ago, jax_rox said:

8-bit will do that to you.

Not to that extent. Not even close.

[kidzrevil]

4 hours ago, cpc said:

ETTR isn't usually the best way to expose log (or any curve): too much normalizing work in post on shot by shot basis. Better to re-rate the camera slower and expose consistently.

In the case of Sony A series it is probably best to just shoot one of the Cine curves. They have decent latitude without entirely butchering  mids density. Perhaps the only practical exception is shooting 4k and delivering 1080p, which restores a bit of density after the downscale.

I would disagree with that first point. On average I shoot 3 stops over and the image is fine. No banding at all and thats with slog3. Slog2 3 stops over works even better. With any image you expose for the parts of the image you want to keep. You want more shadow detail ? Expose up. You want more highlight and midtone detail ? Expose one stop over. It’s very easy if you know how to grade (LOG). Delivering in 1080 makes compression artifacts unnoticable.

As far as the second point is concerned I do agree that it may be best to shoot in a CINE curve for most people. On a paid gig where I can not afford to experiment I would opt for a CINE curve. On a personal project I will shot SLOG for the flexibility

[EthanAlexander]

@kidzrevil I'm pretty certain what @cpc meant wasn't that you shouldn't ETTR - He's saying in order to keep exposure consistent,  rather than just ETTR as much as possible before clipping every time, it's better to overexpose by the same amount every time by rating the film speed lower. I agree - makes batch grading an actual possibility.

[Asmundma]

22 hours ago, mercer said:

Now as everybody knows, I am not that good of a colorist and sLog grading can be tricky for experienced colorists but here is my attempt at grading some sLog3 footage I shot with the a6500 last year. Grading aside, the shot at 1:30 in, I feel shows a little of that Raw texture in the tree bark. This is the only a6500 sample I have, so I apologize for the repost...

Again, my poor grading aside, here are a few screengrabs from my time with the D5500... I loved that camera... even though it was kind of a pain in the butt to shoot with in manual mode with old Nikkor lenses... the D750 looks even better to me but I never used one.

 

And finally, yes thank you... I doubt you remember but we had a long discussion about eos-m ml Raw a couple years back and I became obsessed with it but couldn’t get rid of this dreaded pink dots. Eventually I tried the 50D (later realizing I probably should have gone with the 7D instead... I didn’t realize how cheap they are) but the lack of audio on the 50D made it a novelty camera at best. And finally, 6 months later I sold a ton of stuff to afford a 5D3... I couldn’t be happier...

 

The C100 is a great choice, would love to see how you’re getting along with it.

 

Anyway, sorry @kidzrevil for the slight OT.

Sorry to say, but this film and pictures look bad to me. Maybe it’s distroyed putting them up here? The first one in the thread look better.

[kidzrevil]

29 minutes ago, EthanAlexander said:

@kidzrevil I'm pretty certain what @cpc meant wasn't that you shouldn't ETTR - He's saying in order to keep exposure consistent,  rather than just ETTR as much as possible before clipping every time, it's better to overexpose by the same amount every time by rating the film speed lower. I agree - makes batch grading an actual possibility.

Highly disagree with no shooting ETTR and is not my method of exposure. I optimize every scene for how I plan on processing in post. Some scenes I may even go beyond the clipping point. Batch processing anything is out of the question and is kind of lazy when you think about it. LOG isn’t something you can afford to be lazy with and batch process your work. If this is the mentality behind shooting LOG then no wonder SLOG is nototorious for giving crap results with an 8 bit camera. Not an efficient way to shoot and grade LOG in my experience...slam that histogram as far to the right as possible ! 107 ire has the highest code values 

[EthanAlexander]

It's not like you can't change your exposure during a shoot by consistently rating the film lower... it's the amount you over expose by that should stay consistent. It's way more sustainable for any kind of large project, and it keeps skin at the same IRE between shots, which I'd say is the most important part.

That way you can apply one single starting LUT such as an "S Log 2 - minus 2 EV to Rec 709" to everything and then grade from there. It's a much quicker way to get a matching starting grade, and then you can tweak from there.

Consistent exposure is the most efficient way to deal with LOG, by definition ; )

[mercer]

28 minutes ago, Asmundma said:

Sorry to say, but this film and pictures look bad to me. Maybe it’s distroyed putting them up here? The first one in the thread look better.

No need to apologize, I prefer honesty. I come here to get better and kidzrevil is definitely a better operator and colorist than I am. I agree that my a6500 Video doesn’t look great but it wasn’t really the point of posting it, I posted it to show how sLog3 can capture the detail that a lot of other cameras will smooth over. As far as the photos/screengrabs, I actually like them... especially the last two that were shot with the 5D3 and ML Raw... can you elaborate what you do not like about them in a constructive way, so I can take your thoughts and learn from your experience?

[kidzrevil]

@EthanAlexander again I highly disagree. I’ve done plenty of shoots where my exposure varied shot to shot but always fell around the 2-3 stop range. Im sure keeping the exposure consistent like lets say +2 over will give you great results but there are many situations where the in camera metering will get confused. Thats why I don’t follow the meter I pay attention to the zebras and histogram. Also I do not depend on LUT’s for my grades because I am a control freak so that’s kinda irrelevant to my workflow..plus you can change the exposure before you apply the LUT so you are not limited by the exposure range the LUT was built for. Please keep in mind some of these LUT’s are POORLY designed and I do not reccommend them for consistent results. Again I do not use a batch workflow because I aim for precision not speed when I am working. 

this was shot in SLOG3 and the exposure varies. If you know what you are doing in post without the dependency of LUT’s you can normalize anything. I highly reccommend FILMCONVERT.

I will experiment with your suggestion of keeping the exposure consistent and see if it gives me better results on a shoot. 

@Shirozina absolutely ! The compression has a bigger impact than 8 bit so its key you optimize the image before compression. I’ve adjusted my detail settings to suppress the in camera sharpening as much as possible. I also use “invisible” diffusion filters like the tiffen black diffusion fx to limit the amount of high frequency data the sensor is able to see before compression. I get next to 0 artifacting especially when downsampling to 1080p 10bit 444

[EthanAlexander]

I'm seeing some places where I haven't explained myself very well:

The way I expose 9 times out of 10 is by using false color and keeping skin consistent and/or using a middle grey card (zebras can be set to work like this). This means there's no way for the camera to mess things up.

As for LUT's, I'm talking LUTs that provide a starting point, and for me, I've created my own in LUTCalc for the specific scenarios I've shot, such as the "S Log 2 - minus 2 EV to Rec 709" I mentioned. Sometimes I even leave the color space change out and only use the exposure compensation. I consider myself highly skilled with grading, so please understand that I don't use these because I need them, I use them because of the time savings. I will still go in for a creative grade at the end of a project.

Since you said you're open to experimenting, I highly recommend at least trying this method. So for instance, with S Log 2, to get a +2EV you'd put middle grey at 55% and skin tones in the 67-72% IRE range, and the grade will be repeatable between shots. If anything, all I'm really advocating is that instead of using zebras for clipping, set them for skin exposure

[mercer]

1 hour ago, kidzrevil said:

 I forgot to answer your first question. The RX10 ii is a great camera, I've always used mine with a tiffen pro mist diffuser when I had it. Have you seen the RX10 IV ? Looks promising

Yeah, I liked the RX10ii from my brief time with it. For most projects I have my 5D3, but I would like a 2nd camera to shoot impromptu videos with and family type stuff. However, I don’t need the 4K, so for the money the original RX10 would really fit my needs and be more than enough as an all in one, bare bones pseudo-cine camera. So my question is, what regular profiles do you recommend ... autumn leaves, standard, portrait, neutral? Do you dial everything down?

[kidzrevil]

@mercer standard with -2 saturation and -3 sharpness works well. Neutral with -3 sharpness is amazing for skin tones. The other two profiles I am interested in is “deep” and “light”. I read that those two profiles mirror minolta’s low key and high key profile settings. If you have some time I need help testing what’s the clipping point of the standard gammas. Not sure if it’s 100 or 109 ire

[mercer]

Is that Neutral with contrast and saturation at default 0 and sharpness at -3, or with contrast dialed all the way down and saturation at -2? If I get the RX10, I’ll be glad to help, just unsure right now since I’m in no rush to get a second camera until I can find the right price.

@kidzrevil

[kidzrevil]

@mercer everything at default except sharpness