Well explained. A handy tip for shooters that might grade their own stuff is to actually export a lut that is underexposed in the same fashion you show in the video. I always have a lut on my monitor that is 1 under as well as 2 under forcing me to increase light falling on the sensor. This is particularly handy for those that have cameras dual base iso (the higher one is typically a bit noisier). That would be a perfect scenario to use the underexposed luts. Hope this all makes sense!
@@tommyfilms2765 - I just made it myself. I'm using Cullens voyager pack so I picked out a look that I liked and printed that down one stop with linear gain. Exported it as monitor lut to use on my Atomos Ninja. Node 1: CST camera working space-->DWG Node 2: Linear Gain -1 stop Node 3: Lut from Cullens Voyager pack Node 4: DWG-->Rec709 Then just export that as a lut. Hope it all makes sense.
An important thing to keep in mind though is that by overexposing and gaining down your signal (be it wit ISO or in post) you are effectively losing dynamig range in the highlights by putting that into the shadows. To me that is a huge deal, as hard clipping highlights looks much worse than losing your blacks into the noise floor, which has a smoother falloff. I recently graded a feature in which they mixed film and digital (Alexa mini) and had to be matched (it was BW so it was easier). After some tests we decided that the best option both for dynamic range distribution and texture was to shoot at 1600 ISO.
@@MartiSmz yea but we're talking about overexposing it as long as you don't clip your highlights. If your highlights are hard clipping, don't ETTR. (Also shooting at 1600 ISO means very different things on different cameras. Your camera could have a second base ISO at 1600 which would mean that it's the best ISO to shoot at.)
@@jammaschan sure, I just said it changes the dynamic range distribution as something relevant to keep in mind. I hadn't specified it was shot on the Alexa mini (base ISO 800), just added that to my original comment.
That’s why I appreciate that BlackMagic gives us the exposure charts for their cameras, it shows how we are moving either away or towards the noise floor with each ISO. Thanks for another great video Cullen!
The way BlackMagic cameras incorporate ISO into their RAW format differs drastically from the way RED incorporates ISO. ISO is baked into the BMD RAW format, while RED doesn't apply ISO (or white balance) until post. Fiddling with ISO setting, in camera, is entirely different between BMD RAW and Red RAW.
@@billravens8136 You sure? I think ISO is not baked into BRAW with the exception of the dual native ISO cameras where (by necessity) the selection of gain is baked in, e.g. ISO 400 vs 3200. Other than that, the ISO is just metadata, best as I can tell.
@@robin5mi It's easy enough to verify whether ISO is baked in, or not. Setup a "normal" exposure. Then change the in-camera ISO. You can and will experience a change in the image brightness. But, then, check the change in the histogram. If the histogram changes to reflect the image brightness, ISO is being baked in. Visa versa, if the waveform does not change....ISO is not being baked in. And, after importing your image into your NLE, does the RAW image reflect the change in ISO?
@@billravens8136 Yeah I verified that for the BMPCC6K at the very least, the recorded data is identical (within the range of ISOs that effect the native ISO analog gain selection, so 400 and 3200).
@@robin5mi that's odd, because on my BMPCC6K, the waveform shifts when I change the ISO. That says ISO is baked in. AFAIK, only Red cameras don't bake in ISO in the Redraw.
If I were teaching someone from scratch, they'd start out shooting raw (BRAW) and log. They'd ignore the exposure charts and always shoot native 400 ASA outdoors and native 3200 indoors, so they see what the sensor sees. ND to control exposure. They'd ETTR but never clip any non-specular highlights. They'd set final exposure and colour temp in post. Since BRAW is raw at ProRes file sizes, they'd remain in raw thru-out. I'd tell them a camera is a light-recorder. Without photons exciting the sensor, even if there's an image, it's going to look dead. BRAW is simple and flexible. I will never go back to complicated and baked in. YMMV. Great video! Subscribed
I think the tip for using ISO to ETTR is correct (as Cullen usually is!) but incomplete: if you lower your ISO to below your sensor's native ISO and *are not shooting in RAW*, then you are simply clipping your top stop of highlights. The more complete piece of advice would be: Lower your ISO to half (1 stop) or a quarter (2 stops) of your camera's base ISO, set your lighting so that it looks good, then return to your base ISO and only then press record. Another less cumbersome option is to load up a -1 stop or -2 stop LUT onto your recording monitor.
Can happen even if you shoot raw. depends on the camera. Canon Raw, for example, is worthless and doesn't store ISO as metadata. ETTR is really not that complicated. Use false color or zebras or something and make sure you ain't clipping anything you can't afford to clip.
@@Mionwang yeah, these are pretty much camera specific scenarios. For example, I wouldn't ETTR Blackmagic pocket cameras in most cases. Only in "low light simulated" scenes with very soft shadows. Pocket's HL falloff is very steep. Highest 2 steps are both basically clipped and digging out detail from the highlights is a pain in the ass, especially if you try to make the curve more shallow. Also, I kinda think the false color in pocket cameras is already compensated a bit since exposing to the skin seems higher than typical. And what Tinsley said, can't really see a reason to dial back to the native ISO before hitting rec when ETTR, if your camera is recording ISO as metadata.
a bit of caution: Extreme ETTR(+2~3 or even more stops) is not a good technique. Even for a camera like Alexa which has a very very linear OETF(LogC) above mid grey, you are at the risk of losing color saturation and may potentially have channel clips, even in raw. There were guru colorists talked about this. Also from a workflow perspective, David Mullen has also mentioned the impracticality of shooting every shot ETTR cause that will result in workflow problem and also uneven base noise. The clever way, as a lot of experienced DP know, is to "rate" the camera with their light meter/false color at a lower ISO and light to it, combining with a pull down lut for preview on set.
Absolutely agree with this explanation of ETTR. Unfortunately the footage from camera people is not all of the caliber that you might be used to. And there are many people explaining and using ETTR incorrectly - especially the ISO part that you explained correctly. There are people that are using ETTR by setting up zebras for the highlights just before clipping and doing it on every angle and every situation. Which means that every angle and every clip is exposed differently - so the colorist is then setting exposure for every clip and consistency of the capture is all over the place. In cinematography, consistency and continuity between set-ups within a scene will trump exposure perfection in terms of giving each shot maximum exposure to lower noise. Plus how is anyone looking at the monitor on the set going to judge the mood if everything is just exposed brightly? There are also cameras that have a knee set that when the skin tones are pushed up into it that it creates a roll - off that is not very natural looking - even when the exposure is reduced. Also with ISO - lowering the ISO is supposedly lowering the middle grey and changing the ratio of stops below and stops above - which if you ever look at a BMD ISO chart, you can plainly see that you would get more detail in shadows with lower ISO and better highlight detail with higher ISO. I asked Walter Volpatto about middle grey and exposure on set and he told me: "If the 18% gray is coherent with the exposure in the set, the DoP will have to just use the light meter and get a predictable result."
Magic Lantern for Canon DSLRs (rest in pieces) had the best ETTR ever conceived. They figured out how to read raw data from the sensor and simply jacked up the exposure until right before clipping occurred. It even had a time-lapse mode that could be set to only move in one direction for sunrise/sunset use!
You’ve described this perfectly. I was trying to explain it the other day and couldn’t get it across. I should have just sent them a link for this video. Thank you.
That's like the best way to demonstrate how we should expose our images . It's great to know someone is trying to do the explanation part so that the colorist gets their jobs done more effectively
When this truth hit me, it' was sort of a "duh" moment. But , I admit, my mind has been opened. I was NOT connecting things in this way. I cannot unsee it now. Thanks!!!
Really outstanding and well explained video. When you talked about increasing iso to expose further to the right, I still think the consensus is that it's always best to expose properly using the camera, rather than increasing in post. Also, might have been good to mention that increasing the iso to hit the second dual native iso is a common tool used by videographers to further expose a scene but not be penalised too much for quality. Great video once again, learnt a lot!
so glad i just found this. I could not figure out why there was so much noise in my shadows when I set my base iso and "thought" i had proper exposure. Turns out I was completely off with my lighting and needed more light , not less. I am "just" a youtuber, but hey, the info is gold!
This is such a helpful video to re-align knowledge. I know, that I knew this...but that hasn't stopped me from falling into this exact trap every once in a while, being so caught up that a couple of notches in ISO couldn't hurt to expose to the right....Wrong...and hearing it explained so clearly and so concise really helps to hammer it home and remind me WHY, and keep it at the forefront of thought when hitting the record button. Always, just a fountain of insight Cullen. Appreciate your continued efforts and work. Just know that your voice will be booming in the minds of anyone exposing for an image, the voice of god...of colorists around the globe.
Fantastic video as usual! In my understanding of the tech I do think there is slightly more to the ETTR with ISO question though. Firstly, in dual gain cameras obviously there is a significant bonus in ETTR with ISO if it takes you to the higher base. The second, more nuanced point has to do with compression. In lossy format gammas (so non-raw, log gammas) the knee compresses the shadows in the recorded file more than the midtones. This extra compression is a little different to the noise floor but it stacks on top of it, adds blockiness and interacts with the noise grossly. ETTR with ISO is certainly much less effective at improving quality than increasing the light received. Using it to lift shadows out of the knee when there is enough highlight headroom and no other option, though, does give a cleaner image. The noise floor comes up, but it's less blocky and artefacty with less colour quantisation issues so it's definitely less intrusive. You have far more experience and knowledge than me though, so please let me know if any of this is way off base in practice
I wondered the same thing. With 8bit codecs, upping ISO definitely still nets a cleaner image. I assume because it’s better utilizing all the available “bits” of the codec. This is especially noticeable in saturated gradients. Maybe it’s just a semantic thing and not true ETTR. Would love to hear Cullen’s take. I guess I need to start attending grade school to ask these things 😅
@@JonPais this is just untrue. The same amount of noise is present, it's just brighter. ISO doesn't cause noise, it reveals it the same way as increasing exposure in post. If you push one stop with ISO then pull one stop in post the noise level will be identical to if you'd shot at that iso level, but there will be less compression in the shadows because they were lifted out of the knee when the image was recorded into log and compressed. This means less colour shifting and blockiness, more detail in the shadows. All log profiles use a knee, some also use a shoulder, where the shadows and highlights respectively are compressed more than the midtones. This results is more quantisation & a functionally lower bit depth per stop for the darkest parts of the image. ETTR with ISO doesn't reduce noise, the noise floor is fixed for each gain circuit, but it can lift the shadow information out of the knee and reduce compression on it - compression which can also make the noise significantly bigger and uglier.
@@LWMphoto TLDR; "With 8bit codecs, upping ISO definitely still nets a cleaner image." This is categorically false. Raising ISO increases the appearance of noise and increasing ISO is NOT ETTR. Correct exposure is about balancing noise and highlight protection.
Without going into technical stuff from simple tests on my Nikon when I first started shooting log, overexposing even if with ISO and then bringing it down just looked better. Otherwise it was a muddy mess.
It is counter intuitive but in cinema camera you want to use higher iso for high key scenes and lower iso for low key scenes as this is how you remap dynamic range. For example if you set the Alexa 35 at 3200 iso, you would get 11 stops of dynamic range above mid grey and only 5 stops below. This is ideal for bright scenes where you don’t care too much about the shadows details. And set iso to 200 if I’m shooting green screen as in most cases you want more dynamic below mid grey.
It’s somewhat of an oversimplification but I think of sensors as similar to traditional film. Film was rated at a particular ISO and anything you did that varied from exposing for that ISO was a compromise. Maybe it was worth it to over or under expose. Maybe it was worth it to push the processing one or two stops. But it was clear where the optimum performance was. Despite Red’s marketing, every sensor has an optimum ISO (or maybe a primary and secondary for dual ISO sensors) much like film. You can dial in a different number, but that bit if silicon is going to respond to the photons hitting it in a specific way regardless of the number you’ve dialed in on the screen. There’s no free lunch.
One thing you should mention tho is that lowering the ISO will ultimately lower your highlight rendition as the dynamic range will suffer a little. So what you should be aiming at is ultimately the native iso of your camera or if you need more highlights (in a brighter shot for example), raise it a little or if you habe less dynamic range in your image (as with a more low key shot), lower your ISO
also when you mentioned lighting the scene. You mentioned changing the brightness of the lights.. but the more important consideration is the contrast ratios. When the shadows are filled (lit up to be be visually too bright) and then crushed in post the file will be cleaner and give you more options. Of course if the 'visually too bright' is upsetting the DP or director a curve down LUT might get them to put a sock in it.
I never understood why you want to lower your ISO in a dark situation, but it's because there's when it's dark, and you crank the sensitivity, you get a noisy image that's magnified in intensity. It's like if you have someone who's bad at singing give them a mic... turning up the game or the mic sensitivity won't make them a better singer, you'll just get louder bad singing.
This can also change with Dual Iso systems. Glen is on the right track at looking at your where your cameras middle gray lands in what Iso. Love the blackmagic chart and is invaluable when trying to expose to the right.
Glad I stick with Sony native ISO 800 and going +1 stop to the right even on my camera display “709 viewing assist” shows an insane clipping on the highlight region. Going back home adding CST on Davinci introduce me to super clean image for me to start grading.
Very interesting! I've been satisfied by the images I got from my FX3, and I mainly sticked to the base ISOs. But I felt like Cine EI was lost on me (I don't do much fictionnal or narrative work). But by eliminating the ISO from the ETTR equation is what makes it clicks! I don't think I would use it in a real dirty run'n'gun, but if I want to do a shoot where I sit mostly at +1.0 I can put my EI around 500 ISO (800 base) and expose as neutral as I can. I've been eyeballing it to keep my image around +0.3, +1.7 with base ISO and monitoring with a LUT, but its hard to resist adjusting, since my eyes see +1 or +1.7 as overexposed on the monitor! Thanks for this!
Same here : I always stayed away of the ISO setting in CINE EI, but this video convinced me to give it a try in exactly the same maner as you. Many thanks, Cullen.
Quick question, I know I don’t make videos anymore but I did make one that very specifically experimented with iso levels I would link here but it would be flagged as spam. Based on your explanation every single one of my exposures adjusted in post should have looked the same as I only changed the ISO. But as you can see that’s not the case. I had a 30k iso image look way better than an 800iso image with the same amount of light getting to the sensor. I think what you fail to consider is the fragility of video codecs. Getting the gain in before it’s baked into the file is way better than adding gain onto an mp4 and especially a 4:2:0 bit depth file. I will comment the video link in a below reply in case it gets removed.
The one additional thing that I believe should have been mentioned in this video is that many cameras have dual native ISO's these days. In that case, cranking up the ISO may actually IMPROVE noise performance (and sometimes DR), if you're at or near those points. Example: the A7iii, with base ISO's of 100 and 640, shoots MUCH cleaner at ISO 640 than at ISO 500. So although you are technically increasing gain, the net result is much better.
Wirth dual native ISOs, once you hit the second ISO range, you're entering a new circuit with a new base exposure level (or Native ISO). When you go to the higher ISO circuit, you're no longer increasing the gain from the previous circuit, but entering a new gain range. For example, on the BMPCC6K cameras, you have ISO ranges of 100-1000 and 1250-6400. When you change the ISO from 400 to 1000 for instance, you're increasing gain. But if you change from ISO400 to 1250, you're entering a new circuit with a new Native ISO, which is why you can't change the ISO in post production from 400 to 3200. It's not a simple gain adjustment. It's also why there's less noise after entering the next Native ISO range. Hope this helped!
@@temporaryfilms1 yup, although I believe Sony's and some others work a tad bit differently. I could be wrong, but I think I remember reading about the difference. For example, on the A7iii ISO 640 never equals ISO 100, it's still a bit noisier and a little less DR
@@natepotter6911 Doesn't the Sony A7iii also have super good noise reduction? Although I think you can't use it when shooting RAW... however I'm not too well versed on that camera. Also just remembered the name for it, what I was referring to is "analog gain" vs digital gain which is within each of the ISO ranges
@@temporaryfilms1 not sure, I never use in-camera noise reduction. It's a very clean image, great low light performance like most Sony's. Wish it shot more than 8 bit, but I'm getting by until I get an FX3. I do equal photo and video, so it's handy. I have started looking at BlackMagic cameras, too.
11:20 true, but dual gain sensors (e.g. a7siii, but honestly lots of Sony sensors nowadays) can have different noise floors at base ISO and higher base ISO. The one option where changing ISO can actually make sense
Great video. It is important to match the ISO for the available light in your scene however if using Dual Native ISO and exposing correctly with the other variables. The dynamic range capture is mathematical so a good match is important to retain either highlights or shadows whichever are more prominent in the scene. Bumping the ISO for brightly lit scenes is important up to the top end of the lower signal chain for example. Not sure I'm comfortable with dropping the ISO below what is necessary for correct exposure of the maximum stops within the ability of the camera in this case, but everything else you explain here regarding ETTR makes perfect sense. Thank you!
Good information! But not all the cameras work like that. some cameras apply the gain in an analog manner. So, when you increase/decrease the ISO, you're actually changing the real amount of light which is hitting the sensor. (For instance, Canon C500)
Thanks Cullen, I love your explanation provided with real practical examples. It's great listening to a pro show us what is best delivered to a colorist. What is the best way in your opinion to create a for a -1 stop camera lut? Thanks in advance for helping me take my visual content to the next level.
Jim makes some great points - this is a really great explanation of the correct techniques but the issue of skin tone is one I wanted to ask about... I see this all time - flat, bright skin with no texture and I've often wondered if it's because it was exposed too bright and the skin is then in the high roll off part of the S-log curve, making it harder to grade and maintain any detail there? Any insights on this? Thanks again for an excellent video!
Im wondering about the same Things, espicially stuff shot with Sony cameras got more offen than not, pale skintones with no Detail. Honestly i always thought this is the Look they are going after, or they used to with this style of Shooting.
There are youtubers saying that you no longer need to overexpose that much on newer Sony cameras (A7s iii?) . Between 0 and 1 should be fine. I didn't understand why, to complicated to me. It would be greatly appriciated if you made a video either confirming or debunking this. Or explaining under which circumstances this might be true. I really like your explaing style. Thanks for the video. (I got the Sony ZV-E1 with built in lut and Premiere Pro automatically applying the lut. I really would love if it was fine to not ruin the preview by overexposing.)
Thank you so much, Cullen, this video is very informative and useful as always. I totally agree with you, to expose to the right you have to add more light hitting the sensor, that's the best way to get the best results. However, I wanted to share a few considerations about ISO and modern camera sensors and codec compression. When you say, changing ISO in the camera is the same as doing it in postproduction is not always true in my opinion. If you are working with a high-end cinema camera like Arri Alexa, RED or Sony Venice and you shoot RAW or with a codec that do minimal compression like Prores 4444 then yes, a gain processing in camera before compression or after in post-production would be somewhat equal. But what about consumer mirrorless cameras that shoot in 10-bit 4:2:2 or 4:2:0 codecs or less, which usually apply noise reduction and/or digital sharpness by default? In my experience, with these cameras applying a bit of gain before the compression usually results in an improvement over sticking with native ISO and only boosting the gain in post. I think compression makes a huge difference, the more the codec is compressed more you want to boost the ISO to get a decent exposure before the compression hits. Also, modern consumer cameras like Sony, Blackmagic and Panasonic use what is called dual native gain sensors. Panasonic for example have 640iso and 4000iso. In this case, changing the iso from 640 to 4000 equals boosting the lighting hitting the sensor in the same way changing the aperture or increasing lighting intensity would do. That is because these 2 ISOs (640 and 4000) have separate and parallel gain circuits, and both are native. All the information you shared was very handy and punctual, I just thought to add these concepts to the conversation for those who are working with more consumer-oriented cameras and codecs.
@@JonPais What amazing argumentation. Are you 14 years old? If you have the time to read and the urge to let me know I'm wrong, maybe you should also explain yourself, maybe? ...
I may be wrong...I thought on the newer Sony Cameras we no longer had to expose to the right. This summer I exposed everything in neutral. My night shot turned out excellent, however all my day shots have the highlights blown out...
The way ISO is described here I believe is accurate for most mainstream cinema cameras, where there's really only one (or sometimes two) native ISO (IE the analog gain is fixed to one or two specific levels). When you choose an ISO in the RAW panel, it is indeed equivalent to choosing the ISO at record time because in both cases, ISO is just a digital gain. HOWEVER, there are some cameras where the ISO control does control the analog gain. In those specific cameras, increasing ISO actually does give you cleaner shadows at the expense of dynamic range in the highlights. On those cameras, if they allow you to change the ISO in post, at that stage it's applying a digital gain and is not really equivalent to the at-capture ISO control. ISO in post is always just the exposure slider in the RAW panel, with fixed increments. That all being said, on these cameras you'll still get the best images by using a low or native ISO and putting more light on the sensor.
Blackmagic, Arri, and RED stick to the single ISO methodology. I know ZCam has analog gain in their cameras. Some older Canons also had ISO control analog gain, though I don't know what they're doing now, but presumably they still are if ISO can't be controlled in post.
Great explanations Cullen. Not sure if you can address this (if it is an actual issue). If you expose too far to the right, even though you have avoided clipping, are you not shallowing the shoulder of the curve thereby compressing the information in the highlights? I might be barking up the wrong tree but the histogram seems to imply more information being squeezed into less area into the highlights. Anyway… one for future video or Friday discussion maybe.
Yes, over exposing 1 stop will reduce the headroom of the recorded signal by one stop. Should not be a problem, since all professional cameras record more stops than can be (reasonably) displayed even in HDR, you don't need an Arri35 for that. But it of course does require the dop to control the light, which is thei job anyway.
I think ISO has gotten more complicated than just tuning up or down on modern cameras particularly. While it’s true on a general basis, the “dual ISO” and other stuff going on in the background makes it that if you want to preserve highlights better or shadows you will want to move your iso in your favor
Expose to the right is for low light situations though So highlight preservation wouldn’t be a need. Besides you can always change ND, shutter, or aperture instead. In raw formats You can always change iso in post anyways
@@liamhamilton124 changing iso is not the same, it’s still a post processed image, if you shoot, for example at 4000 iso when your base is 800, and then dial back to 800 in post, you will have more noise that if you shoot 800 iso from that start (and change it on post to whatever). It’s not a day and night difference always but it’s there.
@@Visethelegend my point was shoot low iso and if you need to up it in post you can. …I’m pretty sure that most cine cameras only records in their base iso. Meaning it won’t bake the iso into your shot. So changing it in post wouldn’t matter. You could confirm this by setting your exposure to the point where the peaking gates are almost on then changing the iso up. It won’t actually make the gates light up
Normally, we consider the brightness of an image is based on the exposure triangle of aperture, shutter speed and ISO - in combination with the physical brightness of the subject. General guidance is to set the shutter speed to 2x frame rate, so that removes one degree of freedom. If we want greater depth of field, we have to reduce the aperture, which removes another degree of freedom. You have assumed the ideal situation in which we can control the lighting. However, sometimes, we can't control the lighting and just have to shoot what is in front of us. At that point, I think the only alternatives are to crank the ISO in the camera, or raise the gain in post. Am I missing something?
ISO in terms of EI or Exposure index is different from in camera gain ISO. Exposing to the right in EI (less noise) means a lower ISO not a higher one. To capture more highlights at the expense of shadows use a higher ISO EI.
Hi Cullen, Thank you for a tutorial on exposing to the right. Does wrongling exposing to the right using the camera ISO apply when working with a dual ISO camera?
@cullenkelly. Help me here. Log format cameras have a suggested middle grey value. Example: clog2 middle grey should be 42 IRE. Exposing to the rightchanges this. I understand the benefits, especially in photography, but when you’re dealing with native iso and suggestions from the manufacturer, why would you want to deviate?
I will say the one time using ISO can help to shoot over is when a camera has dual native ISO. Its not perfect but if a camera has a native ISO of 800 and 3200 then the 3200 is technically much cleaner than it would typically be. Its shooting two stops over but doing so in a native way like shooting ISO 800 would be. Still not perfect however and I agree its best to get more light to the sensor vs electronically trying to make that light brighter. For me however this is one of the key benefits of dual native ISO. Its not just to shoot to less lights or no lights. Its a way to shoot over when adding more light is not possible. Its important to make sure the dual native ISO is actually cleaner however. Not all cameras handle dual native ISO the same. There is no camera where the second native ISO is ever as clean as the first. So there will be some loss of range and increase of noise. On my Canon R5 ISO 3200 is noisier than ISO 800 but using it to shoot 2 stops over is vastly cleaner looking than not shooting over. Using lights or aperture are better but that may not always be possible so its nice to fall back on another alternative option. Other cameras like the FX3 which has a second native ISO of 12,800 is 4 stops over which is likely way too much. You have to use some ND to compensate for the difference.
How would you expose to the right in a situation where your scene requires having both a bright element as well as a dark element? What would be the best method of finding the middle to not lose either the highlights and shadows?
Awesome explanation as always! So when exposing to the right and monitoring via waveform, what approximate IRE would you suggest for the brightest highlights (specular highlights notwithstanding)? 90 IRE? 95 IRE? 🤔
8:20 this advice might not be applicable to ISO variant cameras. My camera has a much better result going from ISO 200 to 400 than manually increasing the exposure by 1 stop in post. The difference is actually quite huge. Analog Gain is much better than Digital Gain since digital gain increases the downstream noise of the camera much more than Analog does. Also shadow recovery becomes much better as I increase my ISO but that's for scenarios when I'm planning on raising the shadows My point is: Get to know your camera like you'd get to know your best friend. It really is one
What’s your thoughts on your shooting lut being -1 stop darker? Or is it maybe just an easier solution to monitor via ISO ? Also. Your thoughts on HIGH ISOs in bright scenes to bias your decisions to protect your highlights. Basically the inverse of this concept.
Also, this whole ETTR thing only really works if the dynamic range of the scene is less than the DR of your camera, right? E.g. my Sony a7siii has about 12.5 stops w/ Signal to Noise ratio of 2 using S-Log 3 according to Gerald and CineD measurements. If I'm shooting a studio scene that only has 8 or 9 stops of relevant DR, then I can freely do ETTR without clipping any lights and get great detail. If instead, I'm outside in the sunlight midday and I want to take a video of a car in a parking lot that is in the shadow (which I've done before), and I do ETTR such that the sky does not clip, then my car is probably going to be super noisy, because my overall exposure level will be lower than if I had exposed for the car ("exposed to middle" or whatever). That's because this outside daylight scene is probably more like 15-16 stops of DR, but I have only 12.5 in camera. Either you have to light the parking lot / car to reduce the DR, or you have to be OK with clipping (the sky to white or the car in parking lot to a black silhouette) In any case, generally you want the subject as far away from the noise floor as possible without clipping or encountering a weird desaturated / flat part of the tone curve. Relatedly, if you shoot not-RAW but some kind of compressed color profile like Log or some custom Cine gamma, then those gammas have a target middle gray value, and Color Space transforms / LUTs will expect the thing you want to be middle gray in your image to be that target value, otherwise the transform will look odd (the saturation might look over/undersaturated or it looks over/underexposed after the transform). I've been adjusting exposure via the HDR tool (global wheel) in Resolve (which is ok, but honestly still kind of confusing, and the histogram doesn't fully update in real-time). Does using the Gain wheel adjust the # of stops evenly across all values of the footage? Doesn't this depend on the tone curve (e.g. log vs 709)?
Just a question: from your experience, would you rather recommend the dop shooting at half the ISO, supplying them with a normal preview LUT - OR recommend using the camera manufacturer's native ISO and provide a preview LUT which is 1 stop lower?
The concept of ETTR was, first, proposed for still photography. In a still photograph, this concept is valid. I don't argue that ETTR isn't valid for video, with one caveat that is never mentioned. Many novice videographers set up their exposure without following through on a pan, or other motion effect. So, they ignore that the exposure changes as the capture device is panned, or otherwise moved. In the end, what was an ETTR strategy becomes a blown out highlight that wasn't accounted for, A pure ETTR that doesn't take into account the motion, is a fatal mistake resulting in blown out images. ETTR is a tactic that implies no headroom, no safety margin. I, wholeheartedly, disagree with the fundamental concept of ETTR, for this reason.
Thanks so much for this! I've been testing this approach, and I've got a couple of questions: 1) Why use Gain instead of Offset for exposure changes? 2) In my tests, reducing gain from 1 to 0.5 was considerably more than one stop. To get one stop under, I landed at more like 0.82 Gain. Any idea what accounts for the difference? (On Resolve 18.1.4). Thanks!
This might be true if you were bringing a completely uncompressed raw feed into post. In reality noise is also added in compression, and in pretty much every case a shot recorded at 3200 ISO will be cleaner than one recorded at 800 with two stops added in post. Not saying that would do the same as adding more light, but if you will need to add gain you should still do it in camera.
Is there benefit to ETTR for 8 bit footage if there is no Color grading in post - only correcting exposure and minor white balance? I am thinking yes but interested in your view.
So when shooting log my teacher told me to set the exposure at 42% on a grey card. Is this exposed to the middle or left? And I'd it's in the middle how many percent is one stop of light? Would it be 82%? Sounds to bright?
Thanks for this! A question when it comes to using ISO to ETTR. Like you said exposing to the right increases the noise in your image and isn’t an effective way to ETTR. However, does this change with cameras that have a dual-native ISO? From what i’ve noticed, selecting the higher native ISO in this case does not cause increased noise yet results in a brighter image.
Great question! Yes, dual-native ISO introduces another variable...when working with a camera that has this feature, you need to know which of the two "native" ISOs is the baseline for a given ISO selection. For example, if one native ISO is at 400 and the other is at 3200, you'd be better off going with 3200 as opposed to 1600, because setting at 1600 likely is based off the lower native ISO and will have more noise as a result. Hope that helps...this can be slippery stuff!
So if Im understanding Cullen correctly, you should set your ISO to native and then change the lights with keys/fills? And second, increasing ISO in a camera introduces noise, whereas changing the shutter speed will not do so? I had no idea about that last point
Thanks Cullen. Really helpful demystifying why everyone says that proverbial "expose to the right" phrase. But are there scenarios where you would recommend that you don't follow this rule? What about dual native ISO cameras? What about using false colours on your actors face? What about when you shoot outside in the Caribbean with metric tons of light? Are you supposed to expose to the right even when you risk burning your highlights and risk pushing everything beyond that 100% IRE cliff? Or is there something that I am missing in that logic? Cheers, mate! -Charles
I don’t understand something tho… If I want a dark look in my image, do I expose to the right anyways? Exposing everything up including the subjects face and background, and then get the look at post production? I don’t get it, when I over expose a point I tend to overexpose the brighter parts of the image more than I intend to..
How come the ETTR image doesn't blow the highlights on her forehead? If these highlights did start to clip a bit (like in my camera with poor dynamic range), would ETTR still be a worthwhile compromise?
when I see the footage in your resolve and your live recording frame at same time, I feel like you tried hard make you live recording session look exactly like, what footage looks like in resolve, same lamp, same position, almost same skin tone, the way you look to tour machine and the way she look slightly left, + the overall colour and look 😅😜
In a practical way this is like what the EI system in the Sony cameras do right? It's like you use it to show you a darker image so you can compensate and expose to the right, but with out changes in the ISO, so you can preserve the native ISO with the less amount of noise. I am right?
A million people need to see this.
You're the first person I've ever seen tie together the lighting process and grading process in such a concise way.
Well explained.
A handy tip for shooters that might grade their own stuff is to actually export a lut that is underexposed in the same fashion you show in the video. I always have a lut on my monitor that is 1 under as well as 2 under forcing me to increase light falling on the sensor. This is particularly handy for those that have cameras dual base iso (the higher one is typically a bit noisier). That would be a perfect scenario to use the underexposed luts.
Hope this all makes sense!
@@JonPais - How so?
@@JonPais Not at all. Makes perfect sense.
Is there somewhere I can find a lut like that?
@@tommyfilms2765 - I just made it myself. I'm using Cullens voyager pack so I picked out a look that I liked and printed that down one stop with linear gain. Exported it as monitor lut to use on my Atomos Ninja.
Node 1: CST camera working space-->DWG
Node 2: Linear Gain -1 stop
Node 3: Lut from Cullens Voyager pack
Node 4: DWG-->Rec709
Then just export that as a lut.
Hope it all makes sense.
This is also how some Red cameras work when shooting RAW. Their ISO control is just swapping in different strengths of offset LUT.
An important thing to keep in mind though is that by overexposing and gaining down your signal (be it wit ISO or in post) you are effectively losing dynamig range in the highlights by putting that into the shadows. To me that is a huge deal, as hard clipping highlights looks much worse than losing your blacks into the noise floor, which has a smoother falloff. I recently graded a feature in which they mixed film and digital (Alexa mini) and had to be matched (it was BW so it was easier). After some tests we decided that the best option both for dynamic range distribution and texture was to shoot at 1600 ISO.
ETTR and ISO are two separate things. ISO merely redistributes the latitude above and below middle gray.
@@JonPais of course, I'm talking about overexposing and gaining down the signal / underexposing and gaining it up.
@@MartiSmz yea but we're talking about overexposing it as long as you don't clip your highlights. If your highlights are hard clipping, don't ETTR. (Also shooting at 1600 ISO means very different things on different cameras. Your camera could have a second base ISO at 1600 which would mean that it's the best ISO to shoot at.)
@@jammaschan sure, I just said it changes the dynamic range distribution as something relevant to keep in mind. I hadn't specified it was shot on the Alexa mini (base ISO 800), just added that to my original comment.
@@MartiSmz that's fair
That’s why I appreciate that BlackMagic gives us the exposure charts for their cameras, it shows how we are moving either away or towards the noise floor with each ISO. Thanks for another great video Cullen!
The way BlackMagic cameras incorporate ISO into their RAW format differs drastically from the way RED incorporates ISO. ISO is baked into the BMD RAW format, while RED doesn't apply ISO (or white balance) until post. Fiddling with ISO setting, in camera, is entirely different between BMD RAW and Red RAW.
@@billravens8136 You sure? I think ISO is not baked into BRAW with the exception of the dual native ISO cameras where (by necessity) the selection of gain is baked in, e.g. ISO 400 vs 3200. Other than that, the ISO is just metadata, best as I can tell.
@@robin5mi It's easy enough to verify whether ISO is baked in, or not. Setup a "normal" exposure. Then change the in-camera ISO. You can and will experience a change in the image brightness. But, then, check the change in the histogram. If the histogram changes to reflect the image brightness, ISO is being baked in. Visa versa, if the waveform does not change....ISO is not being baked in. And, after importing your image into your NLE, does the RAW image reflect the change in ISO?
@@billravens8136 Yeah I verified that for the BMPCC6K at the very least, the recorded data is identical (within the range of ISOs that effect the native ISO analog gain selection, so 400 and 3200).
@@robin5mi that's odd, because on my BMPCC6K, the waveform shifts when I change the ISO. That says ISO is baked in. AFAIK, only Red cameras don't bake in ISO in the Redraw.
If I were teaching someone from scratch, they'd start out shooting raw (BRAW) and log. They'd ignore the exposure charts and always shoot native 400 ASA outdoors and native 3200 indoors, so they see what the sensor sees. ND to control exposure. They'd ETTR but never clip any non-specular highlights. They'd set final exposure and colour temp in post. Since BRAW is raw at ProRes file sizes, they'd remain in raw thru-out. I'd tell them a camera is a light-recorder. Without photons exciting the sensor, even if there's an image, it's going to look dead. BRAW is simple and flexible. I will never go back to complicated and baked in. YMMV. Great video! Subscribed
I think the tip for using ISO to ETTR is correct (as Cullen usually is!) but incomplete: if you lower your ISO to below your sensor's native ISO and *are not shooting in RAW*, then you are simply clipping your top stop of highlights.
The more complete piece of advice would be: Lower your ISO to half (1 stop) or a quarter (2 stops) of your camera's base ISO, set your lighting so that it looks good, then return to your base ISO and only then press record.
Another less cumbersome option is to load up a -1 stop or -2 stop LUT onto your recording monitor.
Can happen even if you shoot raw. depends on the camera. Canon Raw, for example, is worthless and doesn't store ISO as metadata.
ETTR is really not that complicated. Use false color or zebras or something and make sure you ain't clipping anything you can't afford to clip.
This is something I've been trying to work on. Ie compensation LUT.
Any tips on how to create them?
I was just thinking the same thing about losing the lut. I’ll give that a try!
If your shot is bright enough to clip, then you need to use ND filters.
@@Mionwang yeah, these are pretty much camera specific scenarios. For example, I wouldn't ETTR Blackmagic pocket cameras in most cases. Only in "low light simulated" scenes with very soft shadows. Pocket's HL falloff is very steep. Highest 2 steps are both basically clipped and digging out detail from the highlights is a pain in the ass, especially if you try to make the curve more shallow. Also, I kinda think the false color in pocket cameras is already compensated a bit since exposing to the skin seems higher than typical.
And what Tinsley said, can't really see a reason to dial back to the native ISO before hitting rec when ETTR, if your camera is recording ISO as metadata.
a bit of caution: Extreme ETTR(+2~3 or even more stops) is not a good technique. Even for a camera like Alexa which has a very very linear OETF(LogC) above mid grey, you are at the risk of losing color saturation and may potentially have channel clips, even in raw. There were guru colorists talked about this. Also from a workflow perspective, David Mullen has also mentioned the impracticality of shooting every shot ETTR cause that will result in workflow problem and also uneven base noise.
The clever way, as a lot of experienced DP know, is to "rate" the camera with their light meter/false color at a lower ISO and light to it, combining with a pull down lut for preview on set.
Absolutely agree with this explanation of ETTR. Unfortunately the footage from camera people is not all of the caliber that you might be used to. And there are many people explaining and using ETTR incorrectly - especially the ISO part that you explained correctly. There are people that are using ETTR by setting up zebras for the highlights just before clipping and doing it on every angle and every situation. Which means that every angle and every clip is exposed differently - so the colorist is then setting exposure for every clip and consistency of the capture is all over the place.
In cinematography, consistency and continuity between set-ups within a scene will trump exposure perfection in terms of giving each shot maximum exposure to lower noise. Plus how is anyone looking at the monitor on the set going to judge the mood if everything is just exposed brightly?
There are also cameras that have a knee set that when the skin tones are pushed up into it that it creates a roll - off that is not very natural looking - even when the exposure is reduced.
Also with ISO - lowering the ISO is supposedly lowering the middle grey and changing the ratio of stops below and stops above - which if you ever look at a BMD ISO chart, you can plainly see that you would get more detail in shadows with lower ISO and better highlight detail with higher ISO.
I asked Walter Volpatto about middle grey and exposure on set and he told me:
"If the 18% gray is coherent with the exposure in the set, the DoP will have to just use the light meter and get a predictable result."
A+ agree. This is why I still use a light meter and intentionally choose my lighting ratios - consistency.
In HDR, middle gray can be almost anywhere, as can diffuse white. Award -winning cinematographer Erik Messerschmidt exposes to the right.
Magic Lantern for Canon DSLRs (rest in pieces) had the best ETTR ever conceived. They figured out how to read raw data from the sensor and simply jacked up the exposure until right before clipping occurred. It even had a time-lapse mode that could be set to only move in one direction for sunrise/sunset use!
same on red cameras
You’ve described this perfectly. I was trying to explain it the other day and couldn’t get it across. I should have just sent them a link for this video. Thank you.
This single best explanation, on any topic, I have ever seen. Ever.
That's like the best way to demonstrate how we should expose our images . It's great to know someone is trying to do the explanation part so that the colorist gets their jobs done more effectively
When this truth hit me, it' was sort of a "duh" moment. But , I admit, my mind has been opened. I was NOT connecting things in this way. I cannot unsee it now. Thanks!!!
Really outstanding and well explained video. When you talked about increasing iso to expose further to the right, I still think the consensus is that it's always best to expose properly using the camera, rather than increasing in post. Also, might have been good to mention that increasing the iso to hit the second dual native iso is a common tool used by videographers to further expose a scene but not be penalised too much for quality.
Great video once again, learnt a lot!
so glad i just found this. I could not figure out why there was so much noise in my shadows when I set my base iso and "thought" i had proper exposure. Turns out I was completely off with my lighting and needed more light , not less. I am "just" a youtuber, but hey, the info is gold!
Glad to hear it Josh!
This is such a helpful video to re-align knowledge. I know, that I knew this...but that hasn't stopped me from falling into this exact trap every once in a while, being so caught up that a couple of notches in ISO couldn't hurt to expose to the right....Wrong...and hearing it explained so clearly and so concise really helps to hammer it home and remind me WHY, and keep it at the forefront of thought when hitting the record button. Always, just a fountain of insight Cullen. Appreciate your continued efforts and work. Just know that your voice will be booming in the minds of anyone exposing for an image, the voice of god...of colorists around the globe.
the voice of god? smh
Fantastic video as usual! In my understanding of the tech I do think there is slightly more to the ETTR with ISO question though.
Firstly, in dual gain cameras obviously there is a significant bonus in ETTR with ISO if it takes you to the higher base.
The second, more nuanced point has to do with compression. In lossy format gammas (so non-raw, log gammas) the knee compresses the shadows in the recorded file more than the midtones. This extra compression is a little different to the noise floor but it stacks on top of it, adds blockiness and interacts with the noise grossly.
ETTR with ISO is certainly much less effective at improving quality than increasing the light received. Using it to lift shadows out of the knee when there is enough highlight headroom and no other option, though, does give a cleaner image. The noise floor comes up, but it's less blocky and artefacty with less colour quantisation issues so it's definitely less intrusive.
You have far more experience and knowledge than me though, so please let me know if any of this is way off base in practice
I wondered the same thing. With 8bit codecs, upping ISO definitely still nets a cleaner image. I assume because it’s better utilizing all the available “bits” of the codec. This is especially noticeable in saturated gradients. Maybe it’s just a semantic thing and not true ETTR. Would love to hear Cullen’s take. I guess I need to start attending grade school to ask these things 😅
On the contrary, raising ISO increases the appearance of noise in the image. Increasing ISO is NOT ETTR.
@@JonPais this is just untrue. The same amount of noise is present, it's just brighter. ISO doesn't cause noise, it reveals it the same way as increasing exposure in post.
If you push one stop with ISO then pull one stop in post the noise level will be identical to if you'd shot at that iso level, but there will be less compression in the shadows because they were lifted out of the knee when the image was recorded into log and compressed. This means less colour shifting and blockiness, more detail in the shadows. All log profiles use a knee, some also use a shoulder, where the shadows and highlights respectively are compressed more than the midtones. This results is more quantisation & a functionally lower bit depth per stop for the darkest parts of the image.
ETTR with ISO doesn't reduce noise, the noise floor is fixed for each gain circuit, but it can lift the shadow information out of the knee and reduce compression on it - compression which can also make the noise significantly bigger and uglier.
@@LWMphoto TLDR; "With 8bit codecs, upping ISO definitely still nets a cleaner image." This is categorically false. Raising ISO increases the appearance of noise and increasing ISO is NOT ETTR. Correct exposure is about balancing noise and highlight protection.
Without going into technical stuff from simple tests on my Nikon when I first started shooting log, overexposing even if with ISO and then bringing it down just looked better. Otherwise it was a muddy mess.
it make sense I was doing wrong now I know thanks for the tip Cullen
It is counter intuitive but in cinema camera you want to use higher iso for high key scenes and lower iso for low key scenes as this is how you remap dynamic range. For example if you set the Alexa 35 at 3200 iso, you would get 11 stops of dynamic range above mid grey and only 5 stops below. This is ideal for bright scenes where you don’t care too much about the shadows details. And set iso to 200 if I’m shooting green screen as in most cases you want more dynamic below mid grey.
oh shit. where can I read more about this. you kind of just blew my mind...
Thank you for the clarity on ISO. I learned it in film photography, but it seems like the digital rules are different.
I would always try to use more light or less ND filtering than change the aperture, but sometimes that's no an option.
It’s somewhat of an oversimplification but I think of sensors as similar to traditional film. Film was rated at a particular ISO and anything you did that varied from exposing for that ISO was a compromise. Maybe it was worth it to over or under expose. Maybe it was worth it to push the processing one or two stops. But it was clear where the optimum performance was. Despite Red’s marketing, every sensor has an optimum ISO (or maybe a primary and secondary for dual ISO sensors) much like film. You can dial in a different number, but that bit if silicon is going to respond to the photons hitting it in a specific way regardless of the number you’ve dialed in on the screen. There’s no free lunch.
When I hear or see the ETTR acronym, I immediately translate it to "expose for the shadows, process for the highlights".
I would like to say - been testing your luts for 2 weeks. And I am blown away by the quality. Thank you!
Detailed and clear, a rare combination. Thanks.
Best explanation of ETTR I've seen, yet. Many thanks.
Thanks so much, now I finally thing i grasp this well.
One thing you should mention tho is that lowering the ISO will ultimately lower your highlight rendition as the dynamic range will suffer a little. So what you should be aiming at is ultimately the native iso of your camera or if you need more highlights (in a brighter shot for example), raise it a little or if you habe less dynamic range in your image (as with a more low key shot), lower your ISO
Simple way to explain something that people seem to forget all the time.
also when you mentioned lighting the scene. You mentioned changing the brightness of the lights.. but the more important consideration is the contrast ratios. When the shadows are filled (lit up to be be visually too bright) and then crushed in post the file will be cleaner and give you more options. Of course if the 'visually too bright' is upsetting the DP or director a curve down LUT might get them to put a sock in it.
I never understood why you want to lower your ISO in a dark situation, but it's because there's when it's dark, and you crank the sensitivity, you get a noisy image that's magnified in intensity. It's like if you have someone who's bad at singing give them a mic... turning up the game or the mic sensitivity won't make them a better singer, you'll just get louder bad singing.
This can also change with Dual Iso systems. Glen is on the right track at looking at your where your cameras middle gray lands in what Iso. Love the blackmagic chart and is invaluable when trying to expose to the right.
Glad I stick with Sony native ISO 800 and going +1 stop to the right even on my camera display “709 viewing assist” shows an insane clipping on the highlight region. Going back home adding CST on Davinci introduce me to super clean image for me to start grading.
Very interesting! I've been satisfied by the images I got from my FX3, and I mainly sticked to the base ISOs. But I felt like Cine EI was lost on me (I don't do much fictionnal or narrative work). But by eliminating the ISO from the ETTR equation is what makes it clicks! I don't think I would use it in a real dirty run'n'gun, but if I want to do a shoot where I sit mostly at +1.0 I can put my EI around 500 ISO (800 base) and expose as neutral as I can. I've been eyeballing it to keep my image around +0.3, +1.7 with base ISO and monitoring with a LUT, but its hard to resist adjusting, since my eyes see +1 or +1.7 as overexposed on the monitor! Thanks for this!
Same here : I always stayed away of the ISO setting in CINE EI, but this video convinced me to give it a try in exactly the same maner as you. Many thanks, Cullen.
A very important note there regarding ISO
Brilliant, clear and concise.. Thank you
Invaluable information. Love this
Fantastic. Thank you for sharing your knowledge.
My pleasure Sean!
Great and simple explanation. Thanks and subscribed.
Quick question, I know I don’t make videos anymore but I did make one that very specifically experimented with iso levels I would link here but it would be flagged as spam. Based on your explanation every single one of my exposures adjusted in post should have looked the same as I only changed the ISO. But as you can see that’s not the case. I had a 30k iso image look way better than an 800iso image with the same amount of light getting to the sensor. I think what you fail to consider is the fragility of video codecs. Getting the gain in before it’s baked into the file is way better than adding gain onto an mp4 and especially a 4:2:0 bit depth file. I will comment the video link in a below reply in case it gets removed.
The one additional thing that I believe should have been mentioned in this video is that many cameras have dual native ISO's these days. In that case, cranking up the ISO may actually IMPROVE noise performance (and sometimes DR), if you're at or near those points. Example: the A7iii, with base ISO's of 100 and 640, shoots MUCH cleaner at ISO 640 than at ISO 500. So although you are technically increasing gain, the net result is much better.
Wirth dual native ISOs, once you hit the second ISO range, you're entering a new circuit with a new base exposure level (or Native ISO). When you go to the higher ISO circuit, you're no longer increasing the gain from the previous circuit, but entering a new gain range.
For example, on the BMPCC6K cameras, you have ISO ranges of 100-1000 and 1250-6400. When you change the ISO from 400 to 1000 for instance, you're increasing gain. But if you change from ISO400 to 1250, you're entering a new circuit with a new Native ISO, which is why you can't change the ISO in post production from 400 to 3200. It's not a simple gain adjustment. It's also why there's less noise after entering the next Native ISO range. Hope this helped!
@@temporaryfilms1 yup, although I believe Sony's and some others work a tad bit differently. I could be wrong, but I think I remember reading about the difference. For example, on the A7iii ISO 640 never equals ISO 100, it's still a bit noisier and a little less DR
@@natepotter6911 Doesn't the Sony A7iii also have super good noise reduction? Although I think you can't use it when shooting RAW... however I'm not too well versed on that camera.
Also just remembered the name for it, what I was referring to is "analog gain" vs digital gain which is within each of the ISO ranges
@@temporaryfilms1 not sure, I never use in-camera noise reduction. It's a very clean image, great low light performance like most Sony's. Wish it shot more than 8 bit, but I'm getting by until I get an FX3. I do equal photo and video, so it's handy. I have started looking at BlackMagic cameras, too.
Brilliant! Thanks for making quality content brother, it's appreciated
Game. Changer. Thank you!
Excellent explanation, Cullen.
11:20 true, but dual gain sensors (e.g. a7siii, but honestly lots of Sony sensors nowadays) can have different noise floors at base ISO and higher base ISO. The one option where changing ISO can actually make sense
There is also a thing called ISO Invariance which can be added to the mix/confusion.
Usually only applies to RAW stills?
These are incredible worth informations!
I feel both dumb and enlightened 🤣🙌🏽🙌🏽you just saved my ”moody dark” movie. 👀
Great video Cullen 👍
Great great discussion (as always!)
Great video. It is important to match the ISO for the available light in your scene however if using Dual Native ISO and exposing correctly with the other variables. The dynamic range capture is mathematical so a good match is important to retain either highlights or shadows whichever are more prominent in the scene. Bumping the ISO for brightly lit scenes is important up to the top end of the lower signal chain for example. Not sure I'm comfortable with dropping the ISO below what is necessary for correct exposure of the maximum stops within the ability of the camera in this case, but everything else you explain here regarding ETTR makes perfect sense. Thank you!
Good information! But not all the cameras work like that. some cameras apply the gain in an analog manner. So, when you increase/decrease the ISO, you're actually changing the real amount of light which is hitting the sensor. (For instance, Canon C500)
Thanks Cullen, I love your explanation provided with real practical examples. It's great listening to a pro show us what is best delivered to a colorist. What is the best way in your opinion to create a for a -1 stop camera lut? Thanks in advance for helping me take my visual content to the next level.
Great demonstration. I hope there are DOPs out there watching this.
Jim makes some great points - this is a really great explanation of the correct techniques but the issue of skin tone is one I wanted to ask about... I see this all time - flat, bright skin with no texture and I've often wondered if it's because it was exposed too bright and the skin is then in the high roll off part of the S-log curve, making it harder to grade and maintain any detail there? Any insights on this?
Thanks again for an excellent video!
Im wondering about the same Things, espicially stuff shot with Sony cameras got more offen than not, pale skintones with no Detail. Honestly i always thought this is the Look they are going after, or they used to with this style of Shooting.
There are youtubers saying that you no longer need to overexpose that much on newer Sony cameras (A7s iii?) . Between 0 and 1 should be fine. I didn't understand why, to complicated to me. It would be greatly appriciated if you made a video either confirming or debunking this. Or explaining under which circumstances this might be true. I really like your explaing style. Thanks for the video.
(I got the Sony ZV-E1 with built in lut and Premiere Pro automatically applying the lut. I really would love if it was fine to not ruin the preview by overexposing.)
Thank you so much, Cullen, this video is very informative and useful as always.
I totally agree with you, to expose to the right you have to add more light hitting the sensor, that's the best way to get the best results.
However, I wanted to share a few considerations about ISO and modern camera sensors and codec compression.
When you say, changing ISO in the camera is the same as doing it in postproduction is not always true in my opinion.
If you are working with a high-end cinema camera like Arri Alexa, RED or Sony Venice and you shoot RAW or with a codec that do minimal compression like Prores 4444
then yes, a gain processing in camera before compression or after in post-production would be somewhat equal.
But what about consumer mirrorless cameras that shoot in 10-bit 4:2:2 or 4:2:0 codecs or less, which usually apply noise reduction and/or digital sharpness by default?
In my experience, with these cameras applying a bit of gain before the compression usually results in an improvement over sticking with native ISO and only boosting the gain in post. I think compression makes a huge difference, the more the codec is compressed more you want to boost the ISO to get a decent exposure before the compression hits.
Also, modern consumer cameras like Sony, Blackmagic and Panasonic use what is called dual native gain sensors.
Panasonic for example have 640iso and 4000iso.
In this case, changing the iso from 640 to 4000 equals boosting the lighting hitting the sensor in the same way changing the aperture or increasing lighting intensity would do.
That is because these 2 ISOs (640 and 4000) have separate and parallel gain circuits, and both are native.
All the information you shared was very handy and punctual, I just thought to add these concepts to the conversation for those who are working with more consumer-oriented cameras and codecs.
Your comment is complete and utter nonsense.
@@JonPais What amazing argumentation. Are you 14 years old?
If you have the time to read and the urge to let me know I'm wrong, maybe you should also explain yourself, maybe? ...
Thanks for the explanation!
In Resolve, how are you evenly adding or removing a photometric stop in the Color tab?
Oh, you're setting gain to 2x. That's an even x2 to all the values?
I may be wrong...I thought on the newer Sony Cameras we no longer had to expose to the right. This summer I exposed everything in neutral. My night shot turned out excellent, however all my day shots have the highlights blown out...
The way ISO is described here I believe is accurate for most mainstream cinema cameras, where there's really only one (or sometimes two) native ISO (IE the analog gain is fixed to one or two specific levels). When you choose an ISO in the RAW panel, it is indeed equivalent to choosing the ISO at record time because in both cases, ISO is just a digital gain.
HOWEVER, there are some cameras where the ISO control does control the analog gain. In those specific cameras, increasing ISO actually does give you cleaner shadows at the expense of dynamic range in the highlights. On those cameras, if they allow you to change the ISO in post, at that stage it's applying a digital gain and is not really equivalent to the at-capture ISO control. ISO in post is always just the exposure slider in the RAW panel, with fixed increments. That all being said, on these cameras you'll still get the best images by using a low or native ISO and putting more light on the sensor.
Blackmagic, Arri, and RED stick to the single ISO methodology. I know ZCam has analog gain in their cameras. Some older Canons also had ISO control analog gain, though I don't know what they're doing now, but presumably they still are if ISO can't be controlled in post.
Great explanations Cullen. Not sure if you can address this (if it is an actual issue). If you expose too far to the right, even though you have avoided clipping, are you not shallowing the shoulder of the curve thereby compressing the information in the highlights? I might be barking up the wrong tree but the histogram seems to imply more information being squeezed into less area into the highlights. Anyway… one for future video or Friday discussion maybe.
Yes, over exposing 1 stop will reduce the headroom of the recorded signal by one stop. Should not be a problem, since all professional cameras record more stops than can be (reasonably) displayed even in HDR, you don't need an Arri35 for that. But it of course does require the dop to control the light, which is thei job anyway.
Thank you for explaining ETTR. Could you, one day, cover on stream deck + panel setup wheb grading!?
Just by virtue of giving folks just starting out the opportunity...to waste less time 🤘🤘🤘⛳
Would a happy medium be exporting your on set monitoring lut to be one stop under? Great video once again!
I think ISO has gotten more complicated than just tuning up or down on modern cameras particularly. While it’s true on a general basis, the “dual ISO” and other stuff going on in the background makes it that if you want to preserve highlights better or shadows you will want to move your iso in your favor
Expose to the right is for low light situations though So highlight preservation wouldn’t be a need. Besides you can always change ND, shutter, or aperture instead.
In raw formats You can always change iso in post anyways
You preserve your highlights by using ND filters.
@@liamhamilton124 changing iso is not the same, it’s still a post processed image, if you shoot, for example at 4000 iso when your base is 800, and then dial back to 800 in post, you will have more noise that if you shoot 800 iso from that start (and change it on post to whatever). It’s not a day and night difference always but it’s there.
@@Visethelegend my point was shoot low iso and if you need to up it in post you can.
…I’m pretty sure that most cine cameras only records in their base iso. Meaning it won’t bake the iso into your shot. So changing it in post wouldn’t matter.
You could confirm this by setting your exposure to the point where the peaking gates are almost on then changing the iso up. It won’t actually make the gates light up
Which is the better option if the scene is dark? 1. Raise the ISO in camera OR 2. Keep the ISO lower but raise the gain/offset in post? For video.
Keep the ISO low. If you can control it, open the aperture wider or use more powerful light sources.
Great videos, great topic.
Hi would this also apply to the new Sony censors like the one in the Sony FX3? I've been told I no longer need to expose to the right.
Been using ETTR since 2009. Still have to explain this to people.
Hey Cullen, this is a great demo of ETTR, but HOW did you actually change the exposure, with slower shutter, less ND, lower aperture?
Normally, we consider the brightness of an image is based on the exposure triangle of aperture, shutter speed and ISO - in combination with the physical brightness of the subject. General guidance is to set the shutter speed to 2x frame rate, so that removes one degree of freedom. If we want greater depth of field, we have to reduce the aperture, which removes another degree of freedom. You have assumed the ideal situation in which we can control the lighting. However, sometimes, we can't control the lighting and just have to shoot what is in front of us. At that point, I think the only alternatives are to crank the ISO in the camera, or raise the gain in post. Am I missing something?
A super insightful one !
ISO in terms of EI or Exposure index is different from in camera gain ISO. Exposing to the right in EI (less noise) means a lower ISO not a higher one. To capture more highlights at the expense of shadows use a higher ISO EI.
Hi Cullen,
Thank you for a tutorial on exposing to the right.
Does wrongling exposing to the right using the camera ISO apply when working with a dual ISO camera?
@cullenkelly. Help me here. Log format cameras have a suggested middle grey value. Example: clog2 middle grey should be 42 IRE. Exposing to the rightchanges this. I understand the benefits, especially in photography, but when you’re dealing with native iso and suggestions from the manufacturer, why would you want to deviate?
I will say the one time using ISO can help to shoot over is when a camera has dual native ISO. Its not perfect but if a camera has a native ISO of 800 and 3200 then the 3200 is technically much cleaner than it would typically be. Its shooting two stops over but doing so in a native way like shooting ISO 800 would be. Still not perfect however and I agree its best to get more light to the sensor vs electronically trying to make that light brighter. For me however this is one of the key benefits of dual native ISO. Its not just to shoot to less lights or no lights. Its a way to shoot over when adding more light is not possible. Its important to make sure the dual native ISO is actually cleaner however. Not all cameras handle dual native ISO the same. There is no camera where the second native ISO is ever as clean as the first. So there will be some loss of range and increase of noise. On my Canon R5 ISO 3200 is noisier than ISO 800 but using it to shoot 2 stops over is vastly cleaner looking than not shooting over. Using lights or aperture are better but that may not always be possible so its nice to fall back on another alternative option.
Other cameras like the FX3 which has a second native ISO of 12,800 is 4 stops over which is likely way too much. You have to use some ND to compensate for the difference.
How would you expose to the right in a situation where your scene requires having both a bright element as well as a dark element? What would be the best method of finding the middle to not lose either the highlights and shadows?
Awesome explanation as always! So when exposing to the right and monitoring via waveform, what approximate IRE would you suggest for the brightest highlights (specular highlights notwithstanding)? 90 IRE? 95 IRE? 🤔
Interesting! 👍
Wonderfully explained. Are there any downsides to using ETTR?
Thank you.
8:20 this advice might not be applicable to ISO variant cameras. My camera has a much better result going from ISO 200 to 400 than manually increasing the exposure by 1 stop in post. The difference is actually quite huge.
Analog Gain is much better than Digital Gain since digital gain increases the downstream noise of the camera much more than Analog does.
Also shadow recovery becomes much better as I increase my ISO but that's for scenarios when I'm planning on raising the shadows
My point is: Get to know your camera like you'd get to know your best friend. It really is one
What’s your thoughts on your shooting lut being -1 stop darker? Or is it maybe just an easier solution to monitor via ISO ?
Also. Your thoughts on HIGH ISOs in bright scenes to bias your decisions to protect your highlights. Basically the inverse of this concept.
Exposing to the right by changing ISO *can* work, if you are going up from a low native ISO to a high one.
Also, this whole ETTR thing only really works if the dynamic range of the scene is less than the DR of your camera, right?
E.g. my Sony a7siii has about 12.5 stops w/ Signal to Noise ratio of 2 using S-Log 3 according to Gerald and CineD measurements.
If I'm shooting a studio scene that only has 8 or 9 stops of relevant DR, then I can freely do ETTR without clipping any lights and get great detail.
If instead, I'm outside in the sunlight midday and I want to take a video of a car in a parking lot that is in the shadow (which I've done before), and I do ETTR such that the sky does not clip, then my car is probably going to be super noisy, because my overall exposure level will be lower than if I had exposed for the car ("exposed to middle" or whatever).
That's because this outside daylight scene is probably more like 15-16 stops of DR, but I have only 12.5 in camera. Either you have to light the parking lot / car to reduce the DR, or you have to be OK with clipping (the sky to white or the car in parking lot to a black silhouette)
In any case, generally you want the subject as far away from the noise floor as possible without clipping or encountering a weird desaturated / flat part of the tone curve.
Relatedly, if you shoot not-RAW but some kind of compressed color profile like Log or some custom Cine gamma, then those gammas have a target middle gray value, and Color Space transforms / LUTs will expect the thing you want to be middle gray in your image to be that target value, otherwise the transform will look odd (the saturation might look over/undersaturated or it looks over/underexposed after the transform).
I've been adjusting exposure via the HDR tool (global wheel) in Resolve (which is ok, but honestly still kind of confusing, and the histogram doesn't fully update in real-time).
Does using the Gain wheel adjust the # of stops evenly across all values of the footage? Doesn't this depend on the tone curve (e.g. log vs 709)?
Just a question:
from your experience, would you rather recommend the dop shooting at half the ISO, supplying them with a normal preview LUT - OR recommend using the camera manufacturer's native ISO and provide a preview LUT which is 1 stop lower?
The concept of ETTR was, first, proposed for still photography. In a still photograph, this concept is valid. I don't argue that ETTR isn't valid for video, with one caveat that is never mentioned. Many novice videographers set up their exposure without following through on a pan, or other motion effect. So, they ignore that the exposure changes as the capture device is panned, or otherwise moved. In the end, what was an ETTR strategy becomes a blown out highlight that wasn't accounted for, A pure ETTR that doesn't take into account the motion, is a fatal mistake resulting in blown out images. ETTR is a tactic that implies no headroom, no safety margin. I, wholeheartedly, disagree with the fundamental concept of ETTR, for this reason.
Thanks so much for this! I've been testing this approach, and I've got a couple of questions: 1) Why use Gain instead of Offset for exposure changes? 2) In my tests, reducing gain from 1 to 0.5 was considerably more than one stop. To get one stop under, I landed at more like 0.82 Gain. Any idea what accounts for the difference? (On Resolve 18.1.4). Thanks!
This might be true if you were bringing a completely uncompressed raw feed into post. In reality noise is also added in compression, and in pretty much every case a shot recorded at 3200 ISO will be cleaner than one recorded at 800 with two stops added in post. Not saying that would do the same as adding more light, but if you will need to add gain you should still do it in camera.
Depends on the camera sensor.
Is there benefit to ETTR for 8 bit footage if there is no Color grading in post - only correcting exposure and minor white balance? I am thinking yes but interested in your view.
So when shooting log my teacher told me to set the exposure at 42% on a grey card. Is this exposed to the middle or left? And I'd it's in the middle how many percent is one stop of light? Would it be 82%? Sounds to bright?
Thanks for this! A question when it comes to using ISO to ETTR. Like you said exposing to the right increases the noise in your image and isn’t an effective way to ETTR. However, does this change with cameras that have a dual-native ISO? From what i’ve noticed, selecting the higher native ISO in this case does not cause increased noise yet results in a brighter image.
Great question! Yes, dual-native ISO introduces another variable...when working with a camera that has this feature, you need to know which of the two "native" ISOs is the baseline for a given ISO selection. For example, if one native ISO is at 400 and the other is at 3200, you'd be better off going with 3200 as opposed to 1600, because setting at 1600 likely is based off the lower native ISO and will have more noise as a result. Hope that helps...this can be slippery stuff!
So if Im understanding Cullen correctly, you should set your ISO to native and then change the lights with keys/fills? And second, increasing ISO in a camera introduces noise, whereas changing the shutter speed will not do so? I had no idea about that last point
Yep, all correct!
@cullenkelly how to create a node that does photometric 1 stop in davinci?
Thanks Cullen. Really helpful demystifying why everyone says that proverbial "expose to the right" phrase. But are there scenarios where you would recommend that you don't follow this rule? What about dual native ISO cameras? What about using false colours on your actors face? What about when you shoot outside in the Caribbean with metric tons of light? Are you supposed to expose to the right even when you risk burning your highlights and risk pushing everything beyond that 100% IRE cliff? Or is there something that I am missing in that logic? Cheers, mate! -Charles
Great question, Charles! It's a bit of a nuanced answer that might be best suited for a Grade School discussion!
I don’t understand something tho… If I want a dark look in my image, do I expose to the right anyways? Exposing everything up including the subjects face and background, and then get the look at post production? I don’t get it, when I over expose a point I tend to overexpose the brighter parts of the image more than I intend to..
How come the ETTR image doesn't blow the highlights on her forehead? If these highlights did start to clip a bit (like in my camera with poor dynamic range), would ETTR still be a worthwhile compromise?
when I see the footage in your resolve and your live recording frame at same time, I feel like you tried hard make you live recording session look exactly like, what footage looks like in resolve, same lamp, same position, almost same skin tone, the way you look to tour machine and the way she look slightly left, + the overall colour and look 😅😜
In a practical way this is like what the EI system in the Sony cameras do right? It's like you use it to show you a darker image so you can compensate and expose to the right, but with out changes in the ISO, so you can preserve the native ISO with the less amount of noise. I am right?
You are correct yes