About low key lighting

[LucaM]

I’m following an online cinematography course for beginners and, while the teacher is giving a lot of interesting inputs on a lot of subjects, on one point caused me a bit of doubts, about histograms (in that lesson he was not using a lightmeter). If I did not misunderstand – and it’s possible i simply didn’t understand his point – he suggested to aim for a quite balanced histogram when shooting (not a perfectly flat one , of course), to give the sensor  a lot of informations in both shadows and highlights, then adjust the values in post production to get a low key or high key effect. According to the teacher the other way round (shooting from beginning  a low key scene for example) would make color correction and grading more complicated, since the sensor would have received informations in fewer areas in the shot and that would give less freedom in post production. While I see the logic behind this approach, I’d say that correctly exposing the scene and shooting it as it’s planned (low key or key etc) , in other words taking a decision from the beginning, would help in creating the desired look for the scene. I see myself aimlessy wandering during the grading step if the shot should allow me too much freedom…
What’s your point of view about this subject?

Topic

[Roger Deakins]

I’m totally with you here. I try to shoot the image the way I want it to be in the final film. Make the decision when you shoot not in post.

[Max A.]

In my opinion, and this is only my opinion, this is a topic with a lot of point of view.
The thing I met in my experience (not a lot at the moment, I have been DP for only 2 independent feature films, a few short films, and some local commercials) is that is a “battle” to gain more information from the file (of course you have to know the camera you use, the pros and con.) and so maybe overexpose “a bit” or get “the look” in camera.

In the last feature that I was involved in, a data wrangler guy who was also a DIT often came to me and “advice” for overexposing a bit or pushing information in the shadows, so maybe in post-production, the colorist can darken with tools.
In my opinion, and I repeat this is my opinion, this is something that I never understand 100% because, if it is true that some camera has noise problems in underexposure areas (I was using a Red Gemini) it’s also true that if I want a “dim” atmosphere it’s is strange to put “fill” everywhere, this can change the “look” I’m after.

So, what I consider (and I ask also to people in this forum, I remember there was an experienced and skilled DIT that often answered) is to think about the “contrast ratio” and try to go for it, and maybe overexpose a 1/3 or also 1/2 or even 1 STOP (depending of the characteristics of the camera) everything, and not adding bounces fill here and there that I constantly see in post-production and “alter my look”.

I don’t know if there is a good solution, this is my consideration and I’m glad to read other opinions.

I apologize for my English.
Have a nice Sunday,

Max.

[Stip]

When I shoot raw, I like to dial down the base ISO one stop (2 for extreme low light) when shooting a low key scene.

The only thing this actually does is it makes the preview on the monitor or EVF one stop darker. So I still create the look ‘in camera’ but I get cleaner shadows (because one stop of the dynamic range is moved from above to below middle grey).

Some cameras (Alexa) look great underexposed and don’t necessarily need that.

[Max A.]

Hello Stip,

I’m interested in your process. Tell me If I understand correctly: you lit your scene and exposed it for the native ISO of your camera (let’s say 800) then you dial down one or even two STOP of sensitivity to 400 – 200 before hitting rec.?

The result would be underexposed footage that you can “digitally restore” in post-production, but won’t this result in a loss of details?
Maybe I don’t understand correctly your process because my English is not good enough.

Have a nice Sunday,
Max.

[Stip]

Max,

I am bad at explaining. First, what I described is only valid when shooting raw, not when shooting compresses codecs like Prores!

Let’s say your camera shoots raw and has a base ISO of 800. And let’s say the camera has 10 stops of dynamic range. And that at the base ISO 800, dynamic range is evenly distributed at 5 stops below and 5 stops above middle grey.

If you now lower the ISO to 400, the image on your monitor gets darker, right? That’s all that happens when shooting raw and changing ISO – the preview image gets brighter or darker (except the camera has dual ISO). Changing ISO from 800 to 400 darkens the monitor image by 1 stop.

If you now expose and light according to that monitor image, you let 1 stop more of light hit your sensor compared to ISO 800. You now have 6 stops of dynamic range below and 4 stops above middle grey, so more in the shadows.

Again, this only works when shooting raw. By changing ISO, basically all you do is change the preview image and thus your behavior in lighting and exposure.

Maybe an easier example: You light a low key scene at ISO 400 and T2. Now you change ISO to 800. The image gets 1 stop too bright. To compensate, you close down the aperture to T2.8. Now exposure looks the same as at ISO 400 T2, but 1 stop less of light hits the sensor and the shadows get noisier.

Sometimes Visual Effects want a very clean image and then it makes sense to try and get a low key scene as clean as possible.

[LucaM]

Thanks to you all for your answers!
I know that for some recording profile (s-logs for example) they suggest to overexpose a bit, and perhaps the idea expressed by the teacher was something in that line. But i’ve had the impression that he was meaning a general approach more than a requirement of some types of files, so i was curious about how diffuse is that idea among professional DOP and from your answers  my guess was correct: if you want a scene to look in some way, commit to the idea and manage to shot it in that way (and this may not be easy if the budget and the experience are limited, but this is a completely different problem).

[JakobGrasboeck]

(my original post was not finished and submitted too early)
I used the process that Stip describes on the Sony FS7. In CineEI-Mode the camera records with the native ISO of 2000. I usually rated it at 500, to get 2 stops overexposure. With the CineEI-mode the on-set monitors show the image at ISO500 then. In post i pulled it down two stops, thereby getting richer blacks with not too much noise. For me thats just a technical process (and really just an offset of your intended values – the ratios don‘t change with this process). So I would not consider that to be a creative process ie. it doesn‘t feel wrong to pull the image in post to the look I intended. I‘ll attach an image I shot a lighting exercise at university with this process. I hope it works.

[JakobGrasboeck]

ps.: looks a bit too dark to me now in contrast to the white background of the website, would do differently now but it serves the purpose. still learning how to use fill light. I think in the cinema it looked good back then (because it was dark around the image).

[Stip]

Looks great!

[Hongqi]

Hey Stip,

Could you explain why it has to shot on RAW? Will it work if i captured in Apple Pro RES?

thanks!

[Stip]

Edamame,

I haven’t shot to codecs in years. I think it’s very different than shooting raw and also camera dependent. It might work for low key scenes I guess? In general, using ISO when shooting raw works counter intuitive as opposed to shooting codecs because the former has no internal processing and sensitivity stays the same (sensor’s sensitivity) and the latter has internal processing and dialed in ISO values change the sensitivity.

One example, on a sunny day in the desert you may raise ISO when shooting raw to protect the highlights from clipping while raising ISO when recording to a codec would accelerate clipping.

[Max A.]

Hello Stip!

I’m sorry for the delayed answer. Now I understand your process, it is interesting and I never thought about it in that way. Usually I “pick” an ISO level that “looks fine” to me for the project and stick with it (of course depends on the different situations and how I can control the light).
I will try your method and shoot some tests with my BMPCC6K which I usually use for low-mid projects.

To answer also to Jakob, I like the result of your framing. It is dark but I like the tone. It has some grain but I think it is acceptable.

What I understand is that low-key scenes (especially for not high-end cameras) are often tricky, what I often noticed is that a lot of cameras struggle in the underexposure areas (like when there is -2 to -3 STOPS in the reflected light) but in areas falling in the black looks clear.
As Stips says probably cameras like Alexa don’t have this problem, I never shot on Alexa so I can’t say but if I also want to refer to Mr. Deakins movies there are a lot of scenes in Skyfall, Sicario, Prisoners, 1917, The Goldfinch and Empire of Light (to mention some) with low-key scenarios and those have zero noise and stunning details! The unparalleled talent that has Mr.Deakins is the principal aspect of those results but I think also a great camera can help a cinematographer to “take his risks” and underexpose areas with an “extra pinch of serenity”.

I apologize for my bad English and I wish you all a nice weekend.
Max.

[dmullenasc]

Raising the ISO for a non-raw recording format like ProRes 444 Log-C in the Alexa does not necessarily mean you will have more clipping compared to recording Arriraw with the high ISO just as metadata. It’s when your recording codec is in a display gamma like Rec.709 that you get in trouble with lopping off dynamic range.

The best way to control noise is to work at an ISO that gives you a noise level you like or can live with, and then expose consistently so there aren’t major adjustments in post (particularly to the shadows.)

Assuming your codec keeps all the dynamic range in log as a raw recording would, the problems you run into are more related to bit depth, compression artifacts and to color, which is now baked in.  So in that case, there can be color channel noise issues if you’ve baked in a certain color bias that you later try to remove in post.  With a debayered RGB recording, you have to work in post with the colors you see in the recording.

[Stip]

For Alexa’s I agree but especially mid and low end cinema cameras do not have such great latitude and dynamic range, or choice of (lossless) codecs.

[JakobGrasboeck]

Thank you, Max A. I forgot to mention: The grain you see (and referred to) was added in post, because I wanted some kind of grain texture. But not the texture that the Sony FS7 delivers when you expose it to it’s base ISO in dark scenes. I think the image, once pulled in post from ISO2000 to ISO500, looked super super clean (almost to clean for me).

I think the thing with all these processes is that you can try everything as long as it works for you.

I’m now on a new journey of rating (and exposing) the Alexa sensor to ISO400 because my colorist tells me I’ll get one stop more latitude then (and therefore I’ll be able to get more contrast and richer colors). Let’s see…

(For me this is a very new way of thinking, because I tend to not focus on technical/grading/post processing things to much. But then when I force myself to deal with these topics I find myself growing in things like exposing and lighting my frames differently (better?).)

[Author]

Replies