POST



Autor: Tobias Wiedmer


HDR in shooting and grading 2 / 3

In-Depth and Best Practice for HDR.

When shooting for HDR, you should pay attention to a few things to ensure there are no problems later in post production.

The camera should record either RAW or a logarithmic RGB image. Rec. 709 with Gamma 2.4 or a burned-in look should be avoided. The sensor of the camera and the codec should work in a quantization of at least 10 bits. Using inferior cameras, such as many DSLRs, GoPros and most drones, can lead to banding and compression artifacts that can be seen much more clearly in HDR than in SDR. Ideally you should work with the Arri Alexa, a RED, Panasonic Varicam, Sony F55/F65/Venice, Blackmagic Ursa Mini, Canon C700/C300II or classic 35mm film.

Exposure

When exposing, it is important to protect lights and shadows. This means using exposure and lighting to ensure that there is no clipping in the signal. 



Pan Speed

Due to the perceived image sharpness, pans can be visually more jerky. The differences are particularly noticeable in the brighter image areas, which do not attract as much attention with SDR because there are no longer as many differences in brightness.

Light setting

Lighting for HDR is a very interesting topic. In particular, the fact that in the next few years you will have to continue to use the SDR format in addition to HDR and thus inevitably make compromises. In general, lighting cameramen for HDR can light up in the same way as for SDR. This is a good sign when it comes to the compatibility of the two standards. In detail, however, there are differences that lead to considerable effects. These details and how to handle them in grading will be explained later in the article.


Grading in HDR

The basic prerequisite for grading in HDR is access to an HDR monitor. Without this device, which is relatively expensive to purchase, you cannot do HDR grading.

Since both Hybrid Log Gamma (HLG) and the PQ-based standards (e.g. HDR10) have to be used as standards, one should first consider for which colour space the final programme be coloured. HDR10 has the big advantage that it uses Dolby's PQ curve, which provides a natural gradation of the lights.

HLG has a simpler design. Fortunately, you can visually convert from HDR10 to HLG without loss, so that you can take advantage of the good characteristics of the PQ curve, even if you deliver an HLG delivery.
Rec. 2100 is usually the prescribed delivery gamut. However, you should avoid creating colours outside the P3 gamut, as these can lead to unpredictable clipping on TV sets whose gamut is too small. This can be achieved with good colour management.

For PQ Deliveries, the metadata MaxFALL (max. Average Light Level), MaxCLL (max. Content Light Level), Gamut, EOTF (Electro Optical Transfer Function), max. Luminance and min. Luminance must be calculated and specified after grading so that the display between the mastering display and the end device can be converted correctly.

Room brightness

Grading suites are divided into different room brightness levels. Cinema grading is performed in a dark environment - the only light source is the projected image on the screen. SDR TV grading is done on a monitor in a suite with background light. This background light should correspond exactly to the white point of the monitor, i.e. 6,500 K (D65) and on average 1/10 of the maximum luminance of the monitor; at 100 Nits this would correspond to 10 Nits. For HDR, the ambient luminance is again reduced to 5 Nits at D65 white point.

Colour Managed Workflows

Many colourists are familiar with adapting  movie grading to TV grading. In comparison to an adjustment between HDR and SDR this is comparatively simple, because only the gamut and the gamma are changed.

With HDR10 we not only change the gamma adjustment, but also use a completely different curve. As a result, the lights are very different from each other. It is therefore not possible to convert a film coloured for SDR into a high-quality HDR master with an adjustment. Through SDR grading, the material has already been transferred from the logarithmic recording curve to a Gamma 2.4-compatible image through technical and creative grading. All differences in brightness of the lights in the scenery recorded by the camera are thus brought together. If one were now to convert to HDR10, the brightest lights would be at 100 Nits and one would only have a technical conversion without added value - an SDR image in an HDR container.

You could try to stretch the brightest lights a bit brighter, but you won’t  have a good colour in these areas, and access to the natural highlight graduation is no longer possible.
An advantage of a delivery of SDR in an HDR container in comparison to the SDR delivery would be that the HDR metadata would be sent to the TV terminals and these would then communicate the picture settings, so that the content would probably look more correct than would have been the case,  but if you want to create an HDR master from an SDR master that also has a visual added value, you either have to colour it completely new, or use Colour Managed Workflows.

Basically, these workflows work in a similar way. The camera material has to be characterized by the recording gamut and the recording curve and then transformed into a working colour space with the right transformation. Alternatively, this can also be done using colour grading. By choosing the working colour spaces, you can influence how the colour tools work, and choose one that best fits your grading style.

Then there is a target transform into a target colour space. The target colour spaces should be able to cover as many colour ranges as possible and for HDR10 many maximum brightness levels. The officially available system, which is probably the best on the market, is the Colour Management System (CMS) in Baselight from Filmlight. 

It automatically diagnoses the camera material and converts it into the working colour space. If a conversion or setting is not optimal, the system warns the user. With the DRT (Display Rendering Transform) Family Truelight CAM you can realize different target transformations, e.g. for HDR10 1.000 Nits P3 D65, HDR10 600 Nits P3 D65 and SDR Gamma 2,4 100 Nits Rec709.

However, the possible trim passes are very limited in comparison to a free adjustment with all grading tools, and depending on the look, it can be very difficult to generate an SDR variant corresponding to the HDR master.
At Cine Chromatix we use a self-developed system with Assimilate Scratch, which is very similar to the Filmlight system. In Resolve there is the Resolve Colour Management (RCM). The camera formats are converted into the ACES working colour space using IDT (Input Device Transform) and are then given a technical transform that is supposed to reproduce the image as realistically as possible (RRT - Reference Rendering Transform). This is followed by ODT (Output Device Transform), which can be used to cover various target colour spaces. In HDR, there are not yet all the necessary transformations. However, work is underway to enable parameterized ODTs for all uses (colour spaces, maximum luminance, black levels, viewing room brightness).

Dolby Vision Workflow is a special form of Colour Managed Workflow. You either need a hardware box for this, or you can work with a software solution. In contrast to HDR10, this box, which is called CMU (Content Management Unit), performs the trim passes from  4,000 Nits to 1,000 Nits as optimally as possible. For individual adjustments the user now has special functions within the CMU. These are stored as metadata shot by shot and can thus be called up by the TV terminals.

A visual advantage of Dolby Vision, in addition to the greater bit depth (12 bits instead of 10 bits), is that the consumer TVs are certified by Dolby and are set so that they visually reproduce the result of the mastering monitor in the best possible way. With HDR10, the display itself is responsible for mapping content.

This is done from manufacturer to manufacturer with different approaches. Sony maintains the colour representation,  this may result in a slight white clipping. LG, on the other hand, avoids white clipping. However, the disadvantage is that the picture can be darker than the creative wanted it to be.

Grading techniques for HDR

If you're grading your first project for HDR, you should plan a little more time. The colour space is completely different from SDR, and it takes some time to find your way around.

What's new about HDR is the brilliance of the lights. This does not only mean direct light sources in the picture, but above all also the effect of light on the skin of actors. Edge lights have a much stronger effect in HDR. Even a guide light has much more power on one half of the face. This makes the gradation feel different. One sees the structures of the skin more, impurities in skin colour are more obvious. Changes in light (such as the impact of cloudy conditions) have a stronger effect.

When you start grading, you should build a Layer-Tree / Node-Tree that already has the appropriate tools ready to control the lights. If you wouldn't see the lights a little weakened during matching, then you would always expect your eyes to have especially bright lights, which can easily be as high as 1,000 Nits. This puts a strain on the eyes when working for several hours, especially with the lower room brightness.

By weakening the lights you can slow down the enormous power of HDR a bit and leave the picture impression feeling a  bit more natural. The potential to give more power to the lights remains and this can be used. Shots that do not contain special lights should not necessarily be coloured much brighter than diffuse white (100 Nits).

Related Articles: