Shooting in low light

[Ted Parkes]

Elsie .. You are right and I completely agree. But as a cinematographer it is our job to know the limitations and strengths of that technology. Pick the right tools for the right project for the right reasons. This is an exciting time with all the amazing tools available to us. But this is lost without doing the home work, continuously learning and evolving that craft. There are no short cuts. Reduser is a great form and I hope the advice and council the less experienced get from here inspires them to learn, test and experiment so they gain the necessary knowledg and not just what camera setting might work.

 

[Elsie N]

Elsie .. You are right and I completely agree. But as a cinematographer it is our job to know the limitations and strengths of that technology. Pick the right tools for the right project for the right reasons. This is an exciting time with all the amazing tools available to us. But this is lost without doing the home work, continuously learning and evolving that craft. There are no short cuts. Reduser is a great form and I hope the advice and council the less experienced get from here inspires them to learn, test and experiment so they gain the necessary knowledg and not just what camera setting might work.

All good and valid points.

And I think you might agree that the less experienced need to learn to think in terms that when it is impossible (surprises, financial restraints, etc.) to work within the latitudes of the tried and true ways that they should have the ability to work around said obstacles and know the currently available myriad ways to "skin the cat". '-)

Not trying to be argumentative here Ted... just offering a perspective from someone who doesn't have the training or experience of someone as yourself, and therefore has a different approach to tackling some of the obstacles that a truly Indie movie maker employs.

 

[Michael Frymus]

Questions about expose for 320ISO but shoot at 800ISO.

Lets say you have a good exposure at 320 @T/4.0
Now if you want to shoot it as 800ISO, do I pop in an ND filter to keep my aperture @T/4.0?

How exactly is this done?

 

[Jake Wilganowski]

If u expose for 320, at an f4, then that's your exposure. You can then bump up the iso to brighten up the monitor for example.

320 will show you a closer histogram to raw

 

[Noel R.]

Questions about expose for 320ISO but shoot at 800ISO.

Lets say you have a good exposure at 320 @T/4.0
Now if you want to shoot it as 800ISO, do I pop in an ND filter to keep my aperture @T/4.0?

How exactly is this done?

I think you are overthinking it. One of the issues with lowlight is that your monitor looks cleaner in lowlight situations because for one most people will use the same brightness level on the LCD that they use during the day. This can trick you into thinking your image is brighter than it actually is. Two, you are looking at an uncompressed image on your LCD, once it goes to compressed RAW it might have a little more noise. Some people recommend you shoot 320 so that you have more latitude to play with in post in case you under expose which many do because they only pay attention to what they see in the monitor.

 

[Timur Civan]

If you are interested in the blue filter and its effects:

http://timurcivan.com/2013/06/extracting-the-best-color-from-your-red-camera-an-informal-experiment/

 

[Timur Civan]

+1 on lost craft. There is a difference between "low light" and "no light". Its about lighting. Test, shoot, experiment, learn the limitations of the camera sensor and how to expose for every situation you might encounter. Its about ratios and you deciding what you want to see and what detail is important to the scene.

then there is
'low key" lighting.....

 

[Mondo Ghulam]

Seems I've read somewhere here on Reduser that it is best to do noise reduction (with Neat Video, that is) last... probably because other things you've done that may affect noise will have made less of it, or perhaps because it makes for a faster render. I've followed that advice but after thinking about it, I'm not so sure that's the best way to go. Wish someone from Neat Video would chime in because I'm sure they've tested for this.

Thanks Elsie,

Despite an extensive search here, I somehow missed this thread, that may be of some intrerest to you too:

http://www.reduser.net/forum/showth...ould-Noise-Reduction-come-in-the-post-process

After you posted I decided to test this. I kicked off a bunch of scaled renders from RCX 5K -> 2K using different 3 different scaling algorithms. I wanted to see if any of them improved noise performance on the way down, but aside from the expected sharpening / smoothing of details, there was no material difference in the noise that I could detect. I'm viewing on a calibrated Eizo monitor and reasonably confident of that. I had my debayer set to full and output to ProRes 4444 (later I'll be going to 32 bit EXR).

I then did a Grain Reduction in AE on the 2K footage and with moderate settings was able to produce a very clean image that was tolerant of good levels/exposure adjustments. I don't have Neat Video for AE right now but I've used it in FCP and expect you'd get an even better result.

I'm going to run the same test again, only with 5K footage in AE, applying grain reduction there and output to 2K.

From an asset management perspective, the latter makes more sense to me as you could just go 5K all the way. Having read Axel's last post in the thread I linked to above, I think I will have to go that route for the VFX shots at least.

 

[Elsie N]

Thanks Elsie,

Despite an extensive search here, I somehow missed this thread, that may be of some intrerest to you too:

http://www.reduser.net/forum/showth...ould-Noise-Reduction-come-in-the-post-process

After you posted I decided to test this. I kicked off a bunch of scaled renders from RCX 5K -> 2K using different 3 different scaling algorithms. I wanted to see if any of them improved noise performance on the way down, but aside from the expected sharpening / smoothing of details, there was no material difference in the noise that I could detect. I'm viewing on a calibrated Eizo monitor and reasonably confident of that. I had my debayer set to full and output to ProRes 4444 (later I'll be going to 32 bit EXR).

I then did a Grain Reduction in AE on the 2K footage and with moderate settings was able to produce a very clean image that was tolerant of good levels/exposure adjustments. I don't have Neat Video for AE right now but I've used it in FCP and expect you'd get an even better result.

I'm going to run the same test again, only with 5K footage in AE, applying grain reduction there and output to 2K.

From an asset management perspective, the latter makes more sense to me as you could just go 5K all the way. Having read Axel's last post in the thread I linked to above, I think I will have to go that route for the VFX shots at least.

Yes, that's the thread I was referring to. Thanks for linking to it, and thanks for posting your test results. Very helpful.

 

[Mondo Ghulam]

Yes, that's the thread I was referring to. Thanks for linking to it, and thanks for posting your test results. Very helpful.

Hi Elsie,

I managed to conclude the other half of the test, to perform NR at 5K R3D and then scale down. I tried the same settings as before and modified settings to see if I could more closely match the result at 2K. Again all in AE using the built-in grain reduction tool at 16bit.

Using the same and modified settings, performing noise reduction on the 5K footage before scaling, produced a grittier / slightly noisier image at 2K, than what I see when scaling first and then performing grain reduction at 2K.

I think in the first instance (same settings), this was due to the settings just being less useful at 5K.

In the second instance (modified settings), I had to use much higher values in the grain reduction tool in order to achieve something closer to what I got in the first tests. At 2K, the NR values are more reasonable. I think the modifier itself adversely affects the image, due to the more extreme values. In the absence of a more sophisticated noise reduction tool, I'd say that this is what is happening here.

It's possible that NR prior to the bit depth change (16 bit RAW to 12 Bit ProRes) is part of the problem, but that's beyond the realm of my understanding.

As you would expect, performing NR at 5K is very slow to work with, to render, review etc. I'm working without a rocket, so you can imagine...

I'm investigating this because we have to create a separate branch of assets for artists working on VFX. To complete the project on time, we have to break out images to them way ahead of the conform & grading.

It would seem my somewhat unscientific tests bear out the advice you'd been following previously.

-M

 

[Michael Frymus]

Does anyone know if the Blue filter would be needed on dslrs when shooting video - such as the 5D iii?

 

[Matt W.]

Does anyone know if the Blue filter would be needed on dslrs when shooting video - such as the 5D iii?

Generally no, and the Mark III does perform well in low light. Turn off HTP and set it to 2500 ISO at any white balance... should be pretty clean for what it is. Usable at 5000 and 10000 ISO, but really getting ugly there.