ARRI Tech Talk: ALEXA 35 - Image Quality (Subtitles in EN, ES, IT, PT, 한국어, 日本語)

Рет қаралды 9,799

ARRIChannel

Күн бұрын

Пікірлер: 20

@themustang181 2 жыл бұрын

I haven't been this excited about an S35 camera in a long time. Way to go, Arri. Keeping the lead by a long shot.

@sa.t.2507 2 жыл бұрын

Well I better get starting with those wedding videos so that I can afford a top plate for this camera in a year

@TechEnduro 2 жыл бұрын

Can I get further information about "you must use these LUTs as the bases for your grade"? For example I'm a DIT and on set and often we use custom LUTs developed by colourists or the DP and myself. We often don't start working on theses LUTs from for example rec709. Is this not recommends anymore to make a custom LUT from Logc4 direct? If so why or why not?

@ARRIChannel 2 жыл бұрын

Hi Noahr, we'll try to get you a more technical response soon, but the LogC4 tonal curve is a rather "tricky" curve to manually correct with L/G/G controls. Our experience with nearly 6 months of field testing and 11 Encounters films was that those who tried to manually correct LogC4 were unhappy with the results. The LUT pack that we will release with the camera carries a lot of the work of our image science department, and we think you'll lose some of the magic of the camera if you ignore it. As the camera's colour pipeline now separates the "creative intent" as a log-to-log ALF4 file from the display colour space transform which is handled in the monitoring pathway, using our LUT as an output node in your own colour work will preserve the same level of flexibility.

@kedbear79 2 жыл бұрын

@@mikaelsalander256 An 'accurate' color transform doesn't mean it's an aesthetically pleasing one. In any case, if Arri are using LUTS, those knowledgeable in color science will still be able to create their own custom transforms, bypassing the Arri ones. I presume Arri have basically had to create more non linear math for the new LUTS in order to deal with the high sat high luma colors, whereas the old ones were just matrix + tone curve, the downside being they blew out high luma high sat colours and smashed them into edge of gamut creating the color clipping etc. Non linear / more complex math was the only way to solve the problem that the previous matrix+tone curve introduced.

@gabj9017 2 жыл бұрын

@@ARRIChannel Just confirming - so completely understand that to retain the additional latitude in your Camera Log container you need to move the linear bias and have a 'harder' level of compression in terms of values per-stop. However, will you deliver a CST/IDT/Transform to Log-C3 (after acquisition) in which (in some NLE's) we have the latitude to retain information past the 12-bit 'cap' but the curve still works 'normally' in relation to what we've worked on before.

@gabj9017 2 жыл бұрын

@@mikaelsalander256 I think you may be 'overthinking' this. To retain additional latitude utilising toe (which doesn't make lens with the linear bias) and shoulder is unnecessary. At the moment ARRI's curve in a 12-bit container utilises 512 values per stop above 1024 (the linear bias). With this all they need to do is add additional stops of latitude to this container is either to move the linear bias slightly (which I believe they haven't) and assign less values per-stop which therefore means middle grey reads lower than before therefore will respond very different to previous curves. So! The fix for this is to transform the new 'LogC' curve to the old LogC curve in a space (NLE) that has the latitude to retain new information past the 12-bit cap of the acquisition container. Hope this is helpful and accurate *(however who knows at this stage!).

@kedbear79 2 жыл бұрын

@@mikaelsalander256 No when i mention non linear that isn't in reference to gamma / log encoding, it's in reference to how the color volume is shaped by the LUT, which is what is effecting the 'better' results with high luma / high sat colors such as car brake lights etc. Previously K1S1 was a tone curve and matrix, the new LUTS must be using more complex math. In any case, the idea of having to use a manufacturer LUT is not an attractive one. Re the 'creative intent' LUT, much like how ACES works, it's like an RRT, separate from the ODT. So it's just a LUT that is used in log space and doesn't include the transform to display space, which instead comes later. The harsh gamut clipping you reference isn't due to the gamma / log encoding, it was due to the color matrix. With Alev3 this can be fixed by inverting the AWGRGB back to camera native RGB, then creating a custom color transform. This solves the magenta and cyan / green clipping, but it takes a person knowledgeable and skilled with color science to create the custom transform.

@simonrabeder1599 2 жыл бұрын

When will the logC4 IDT be out?

@ARRIChannel 2 жыл бұрын

Hey Simon, you should find what you're looking for in the ALEXA 35 workflow section: www.arri.com/en/learn-help/learn-help-camera-system/alexa-35-workflows/log-c If you need anything else, please email digitalworkflow@arri.de

@JoshWilliams123 2 жыл бұрын

I'm looking at some LogC4 right now in 18% grade and it looks darker than LogC3! I'm totally freaking out! Is this normal?

@ARRIChannel 2 жыл бұрын

Hey Josh - totally normal! LogC4 puts 18% grey around 29% on a waveform monitor, where as it will sit at 38% in LogC3

@bushgreen260 2 жыл бұрын

*Why did you reduce the pixel pitch from 8 to 6? This will effect quality, especially in low light. This is a 25% reduction in light, which is massive.*

@ARRIChannel 2 жыл бұрын

Hi bushgreen, yes, while the pixel pitch has decreased to allow us to have more resolution in a smaller sensor, many other parts of the sensor design and imaging pathway have been improved in the 12 years since the ALEV3 was first announced. We think the ALEXA 35 is about two stops better in low light performance compared to the ALEXA Mini, and also has at least 2.5 extra stops of dynamic range.