Overexposing the footage as long as you are not clipping highlights. Raising the ambient level up if you can. Sometimes in uncontrolled environments (let say you are doing event coverage) you will just have to pick what you want preserved, highlights or shadows.
The idea is to over-expose in the camera and pull it down in post in an equal amount. In DSLR's there was the motto ETTR (aka Expose to the Right) referring to the histogram on most digital cameras where zero exposure is to the left and clipped is to the right and normal exposure is in the middle between the two. The reasoning is that if you make sure your exposure is bright enough, without clipping anything you want to preserve detail in. You can move the detail you want above the noise floor so much that when you compensate in post, you crush the noise down with the rest of the image. leaving you with a very clean and almost noise-free final image. It does rely on things like a very generous highlight rolloff and has a tendency to fall apart with low bitrate encoders like 8 bit per channel as you are asking the post production software to spread out basically 4 bit per channel over the range that would normally be covered by 8 bit of gradation. But with higher bit depth and RAW image recording, there usually was enough gradations up there to fill out what would end up as an 8 bpc jpeg in the final image. With video, we have 10-12 bpc video and RAW so the same tricks should be able to be used there as well... However... in the case of RAW, you need to consider something else. That most RAW formats only really record at one ISO level (ignoring multi-native ISO sensors) and push and pull it to make the other ISO settings. So, if the native ISO in the camera is 400, and you set it to ISO100 during shooting, you are essentially exposing the ISO400 sensitivity as if it was ISO100, aka pulling it 2 stops, and during playback the decoder does the -2 stop compensation automagically. So in that case, no extra LUT is needed. You set it to ISO100, expose it with 4 times the light (compared with ISO400) and it will look correct at playback with the noise floor clamping already done. You do loose a bit of highlight detail, but that's true with regular ETTR practices as well as it is essentially the same thing. With compressed video, I guess the same thing holds true. But the push-pull is done in camera, preferably with the raw signal, before encoding it with lossy compression. I guess there is some encoding artifacting that can be combated with ETTR-workflows with compressed video. But I am not entirely sure that clamping those to darkness is worth the enhanced banding in the highlights. You know... actually typing it all out... it makes me wonder if it's even a good idea at all to do it manually when manufacturers are already doing it far more accurately than I could.
Seems to me that using that process, that you would have to know as a colorist the intention of the photography before grading. What if the DP wanted to have a scene one stop over or two stops under? When you make a compensation LUT then the DP would have to have really good communication with the colorist on every clip on every scene. Wouldn't the DP just set to manufacturers middle grey, then light the scene to what they would like. Do you do these compensation LUTs for all your show luts? I would be afraid that the DP would want a 1 stop over look and I would then ruin it.
I would imagine that in a good/great working evironment... these things would be logged somewhere. For example, you can have a note in the clapper-board about this is 2 stops overexposed and should be corrected in post... but in some shorter form. Maybe the script-girl (or more modern and probably more gender-neutral equivalent) can note it just like in the olden days when you could have it written down that a film reel should be pushed or pulled during processing. In the case of BRAW, you can have the camera LUT included on the files nondestructively. That way you can turn that LUT on and see what the camera-person saw on set. It can also be noted in the metadata. Other camera manufacturers probably have similar workflow features. I have only just started with my BMPCC6K2G (it is so easy to say that camera-name...). But in the end. Good communication is always a good thing. If you are unsure. Ask the cinematographer or someone who may know.
You would have to be crazy meticulous in pre prod to make this work. Generally speaking, messing about with over or under exposing digital cameras is not ideal. If you go over or under you're changing the saturation/luminance relationship outside of the intended colour space. The whole ETTR is a left over from 35mm days because the highlight retention was more of a guarantee but people just kind of adopted that for digital thinking it magically gets rid of noise. Exposure is exposure, there's the intended exposure for most latitude and best colour rendition for each camera then there's everything else. When shooting, if your shadows are noisy then add light and if your highlights are clipping then ND and fill to taste. In my experience it almost always comes down to adding more light for the cleanest image. I'd say this video is much more about testing the camera latitude (breaking point) than creating useful LUTs for production.
Honestly this is surprisingly poor advice and will lead to many problems, for example the approach of using a LUT to compensate for exposure means the LUT won't fit into a color managed workflow, eg if you were using VFX shots, other cameras and so on. If you want to under or over expose the camera, use ISO/ASA instead because it will do exactly the same thing but be far less prone to serious problems in post or errors on set. If for example you were creating a LUT for Arri LogC, keep 18% grey at Arri's recommended 398 so that you maintain 7 stops under and 7 over. Then if you want to rate the camera for one stop more latitude in shadows, set the camera to 400 ISO or if working without a monitor rate your light meter to 400. Also, if you were to create a LUT for awgrgb LogC and you were working in that color managed space, having a LUT that is correctly mapped to LogC's 18% grey means you can transform other cameras / VFX shots into awgrgb logC, apply your LUT at the end and have everything remain consistent.
I think the main purpose of this workflow it's to create LUTs for the reference feed in set for the DP & Director, who previously determined which settings would be best for the desired look, the lut would be embedded on metadata for reference in the later-proper color grading process
@@alvarohipolito3625 Yes I understand that, but I don't think that makes sense. Why wouldn't you just use ISO in camera instead? It would be exactly the same thing and be much less prone to error or confusion and would also maintain a simpler post pipeline.
@@sebastianmorales6849 if you have played with davinci resolve, you will notice the offset amount does not use units that's is similar to exposure units (stops). you will have to do tests and color correct it yourself to find out how much you need to push the offset wheel for, say, 1 stop of exposure compensation.
@@danielhuang2488 Use HDR wheels instead and input your cameras color space. Then use exposure through stops in the Global wheel in HDR. This way you measure the number of stops scientifically
I am not an expert in this field but i learnt something of value from your video (exposure). Thank you for sharing your valuable knowledge.
Compensate exposure with color space aware tools or linear space in order to get mathematical correct expo comp.
Yep. Though using offset is not that bad mathematically (in a log space)
😍😍 Super insightful!
Awesome! Super helpful advice to know ahead of time before you shoot a scene/project
How can we avoid the noise??? Overexposing in camera or underexposing???? Greethings!!!
Overexposing the footage as long as you are not clipping highlights. Raising the ambient level up if you can. Sometimes in uncontrolled environments (let say you are doing event coverage) you will just have to pick what you want preserved, highlights or shadows.
Avoid noise by feeding the scene more light in the noisy/shadow areas ,or use an ISO that give less grain
@@kolecava thankyou mi friend!!!
@@karliemorris7318 thankyou my friend!!!
The idea is to over-expose in the camera and pull it down in post in an equal amount. In DSLR's there was the motto ETTR (aka Expose to the Right) referring to the histogram on most digital cameras where zero exposure is to the left and clipped is to the right and normal exposure is in the middle between the two.
The reasoning is that if you make sure your exposure is bright enough, without clipping anything you want to preserve detail in. You can move the detail you want above the noise floor so much that when you compensate in post, you crush the noise down with the rest of the image. leaving you with a very clean and almost noise-free final image.
It does rely on things like a very generous highlight rolloff and has a tendency to fall apart with low bitrate encoders like 8 bit per channel as you are asking the post production software to spread out basically 4 bit per channel over the range that would normally be covered by 8 bit of gradation. But with higher bit depth and RAW image recording, there usually was enough gradations up there to fill out what would end up as an 8 bpc jpeg in the final image. With video, we have 10-12 bpc video and RAW so the same tricks should be able to be used there as well...
However... in the case of RAW, you need to consider something else. That most RAW formats only really record at one ISO level (ignoring multi-native ISO sensors) and push and pull it to make the other ISO settings. So, if the native ISO in the camera is 400, and you set it to ISO100 during shooting, you are essentially exposing the ISO400 sensitivity as if it was ISO100, aka pulling it 2 stops, and during playback the decoder does the -2 stop compensation automagically. So in that case, no extra LUT is needed. You set it to ISO100, expose it with 4 times the light (compared with ISO400) and it will look correct at playback with the noise floor clamping already done. You do loose a bit of highlight detail, but that's true with regular ETTR practices as well as it is essentially the same thing.
With compressed video, I guess the same thing holds true. But the push-pull is done in camera, preferably with the raw signal, before encoding it with lossy compression. I guess there is some encoding artifacting that can be combated with ETTR-workflows with compressed video. But I am not entirely sure that clamping those to darkness is worth the enhanced banding in the highlights.
You know... actually typing it all out... it makes me wonder if it's even a good idea at all to do it manually when manufacturers are already doing it far more accurately than I could.
Is there any way to contact David if we have further questions?
Seems to me that using that process, that you would have to know as a colorist the intention of the photography before grading. What if the DP wanted to have a scene one stop over or two stops under? When you make a compensation LUT then the DP would have to have really good communication with the colorist on every clip on every scene. Wouldn't the DP just set to manufacturers middle grey, then light the scene to what they would like. Do you do these compensation LUTs for all your show luts? I would be afraid that the DP would want a 1 stop over look and I would then ruin it.
I would imagine that in a good/great working evironment... these things would be logged somewhere. For example, you can have a note in the clapper-board about this is 2 stops overexposed and should be corrected in post... but in some shorter form. Maybe the script-girl (or more modern and probably more gender-neutral equivalent) can note it just like in the olden days when you could have it written down that a film reel should be pushed or pulled during processing.
In the case of BRAW, you can have the camera LUT included on the files nondestructively. That way you can turn that LUT on and see what the camera-person saw on set. It can also be noted in the metadata. Other camera manufacturers probably have similar workflow features. I have only just started with my BMPCC6K2G (it is so easy to say that camera-name...).
But in the end. Good communication is always a good thing. If you are unsure. Ask the cinematographer or someone who may know.
You would have to be crazy meticulous in pre prod to make this work.
Generally speaking, messing about with over or under exposing digital cameras is not ideal. If you go over or under you're changing the saturation/luminance relationship outside of the intended colour space. The whole ETTR is a left over from 35mm days because the highlight retention was more of a guarantee but people just kind of adopted that for digital thinking it magically gets rid of noise.
Exposure is exposure, there's the intended exposure for most latitude and best colour rendition for each camera then there's everything else.
When shooting, if your shadows are noisy then add light and if your highlights are clipping then ND and fill to taste. In my experience it almost always comes down to adding more light for the cleanest image.
I'd say this video is much more about testing the camera latitude (breaking point) than creating useful LUTs for production.
Honestly this is surprisingly poor advice and will lead to many problems, for example the approach of using a LUT to compensate for exposure means the LUT won't fit into a color managed workflow, eg if you were using VFX shots, other cameras and so on. If you want to under or over expose the camera, use ISO/ASA instead because it will do exactly the same thing but be far less prone to serious problems in post or errors on set. If for example you were creating a LUT for Arri LogC, keep 18% grey at Arri's recommended 398 so that you maintain 7 stops under and 7 over. Then if you want to rate the camera for one stop more latitude in shadows, set the camera to 400 ISO or if working without a monitor rate your light meter to 400. Also, if you were to create a LUT for awgrgb LogC and you were working in that color managed space, having a LUT that is correctly mapped to LogC's 18% grey means you can transform other cameras / VFX shots into awgrgb logC, apply your LUT at the end and have everything remain consistent.
I think the main purpose of this workflow it's to create LUTs for the reference feed in set for the DP & Director, who previously determined which settings would be best for the desired look, the lut would be embedded on metadata for reference in the later-proper color grading process
@@alvarohipolito3625 Yes I understand that, but I don't think that makes sense. Why wouldn't you just use ISO in camera instead? It would be exactly the same thing and be much less prone to error or confusion and would also maintain a simpler post pipeline.
This was information lacking
In what way?
@@filmmakers_academy how does exposure relate to the offset amount, for example
@@sebastianmorales6849 if you have played with davinci resolve, you will notice the offset amount does not use units that's is similar to exposure units (stops). you will have to do tests and color correct it yourself to find out how much you need to push the offset wheel for, say, 1 stop of exposure compensation.
@@danielhuang2488 Use HDR wheels instead and input your cameras color space. Then use exposure through stops in the Global wheel in HDR. This way you measure the number of stops scientifically