Shooting action-cams with a "red" filter in flat profile, and manually white-balancing in post-production

[bghazzal]

Yes, the pictures I'd posted are also from their product page, but I don't understand how it's possible unless they moved the lens further away from the supersuit port.
It's snug on the original AOI, no room for a piece of plastic as far as i can see, as the back of the lens sits almost touching the supersuit port.

Moving it forward seems risky (increasing the risk of vignetting and corner deformation), so not sure how this works. This thread mentioned vignetting with the mount, but then none if cropped to 16/9, so not sure...
 

 

Edited June 23 by bghazzal

[Davide DB]

 

Yes, AFAIR the lens is moved away a bit and there's small aperture on the side.

On this video is clearly visible on the instructions and the mount itself.

I wonder if the original AOI mount works with the full resolutions/ aspect ratio of the latest gopros: 5K and 8:7 while the Backscatter mod doesn't work as per several feedbacks.

 

 

 

 

[DreiFish]

9 hours ago, bghazzal said:

 

Yes, that’s exactly it - clearly the same issue, and as you pinpointed it, it's clearly a weaker filtering of the greenish cast in the ambient spectrum...

This is actually also an issue I had why I tried the Keldan Spectrum on the GoPro – it was way too weak, even in Palau’s generally blue water.  And to add to this, the Keldan Spectrum gels are also slightly weaker than the SF2, as they’re sold as SF -1.5... 
 

Interestingly, they worked fine when manually white balancing the Lumix to ambient light at depth, the UR-Pro performing the worst of the 3 (Keldan SF-1.5 gel > Original Magic > Ur-Pro CY), and it’s exactly the opposite ranking on the GoPro (Ur-Pro CY > Original Magic  > Keldan SF-1.5 gel ), which clearly needs a more powerful filter.
 

All this highlights the importance of having a good filter to work with, as I was mentioning above – these 3 filters are well-designed specialist products, and yet they behave quite differently, which becomes even clearer when working on the footage in post...
 

While the Magic is useable, I’m also a little worried about the red/magenta/orange boost I saw a little deeper – I’m guessing it’s also an effect of the white balance being thrown further off by blues/greens, leading to overcompensation.
This is not good, as needs more work in post and this further degrades image quality…
 

Thanks for the links to gels, it would be great to try to reverse-engineer the UR-PRO SWCY formula – they really had something going with it, it's really a shame to lose such a tool.

 I have no idea how manufacturers work on designing custom filters – do you think the UR-PRO SWCY was actually a proprietary design (like the Keldans), or based on some pre-existent tone (which could then be source / purchased?).
 

Unfortunately, the Rosco/Lee gels are near impossible to source in Indonesia – there is a distributor in Singapore, but the sample I ordered didn’t make it over the first time, and with custom duties it’s a bit of a nightmare.
I think it’ll have to wait a few months, when I’ll be in a place where gels are not accessible...

cheers

 

 

 

I'm really not sure how the various manufacturers have developed their particular filter. For some, it may have been a combination of trial and error using various existing gels. Of course, you could also do it by measuring the target underwater ambient spectrum you want to 'normalize' to daylight -- so, if you have a spectrometer in an underwater housing, you could take it to say, 10m, measure the spectrum of the light there (of course it would vary form location to location and conditions) and then reverse engineer which wavelengths you need to filter out to neutralize the light and work out what sort of filter you need using math/physics 🙂

[Robin.snapshots]

12 hours ago, Davide DB said:

I wonder if the original AOI mount works with the full resolutions/ aspect ratio of the latest gopros: 5K and 8:7 while the Backscatter mod doesn't work as per several feedbacks.

 

The AOI mount also vignettes in 8:7 but considerably less than the Backscatter mount.

 

cropped to 16:9 (or 9:16 for that matter) both mounting options work fine. So if filters are desired, I would definitely opt for the Backscatter mount.

Edited June 23 by Robin.snapshots

[bghazzal]

10 hours ago, DreiFish said:

 

I'm really not sure how the various manufacturers have developed their particular filter. For some, it may have been a combination of trial and error using various existing gels. Of course, you could also do it by measuring the target underwater ambient spectrum you want to 'normalize' to daylight -- so, if you have a spectrometer in an underwater housing, you could take it to say, 10m, measure the spectrum of the light there (of course it would vary form location to location and conditions) and then reverse engineer which wavelengths you need to filter out to neutralize the light and work out what sort of filter you need using math/physics 🙂

 

Hmmm, logisitics are a tad tricky, my spectrometer is still at the cleaners, all that 😁 😉

Looking at your tests for the ambient filters, I'm wondering if would it make sense to try shooting a still of a colour checker on land, first without, then with the UR-Pro and original Magic filters on.

With high CRI strobe and WB set to the strobe's, it should be close to neutral lighting on the raw file, right?
 

Then maybe the individual colour reading differences, especially blues and greens squares, could be measured in software (I'm really bad at image software, but I guess photoshop or lightroom can give such readouts?)

Wondering if this could be a quick and dirty way to get a rough curve highlighting the main differences between the two filters, and then seeing what best matches Rosco / Lee graphs in terms of required filtering range.

 

 

 

Edited June 24 by bghazzal

[bghazzal]

2 hours ago, Robin.snapshots said:

The AOI mount also vignettes in 8:7 but considerably less than the Backscatter mount.

 

cropped to 16:9 (or 9:16 for that matter) both mounting options work fine. So if filters are desired, I would definitely opt for the Backscatter mount.

Thanks - Backscatter's mount isn't sold in these parts, only the standard AOI, and importing isn't an option - but it could possibly be an option elsewhere.
 

The tricky bit, given the proprietary design, would then be to find a flat acrylic UR-PRO CY somewhere, and one large enough to work with.
Most of the ones still on sale are glass, and the acrylic ones I have are slightly convex (SRP's blurfix mount, with 55mm threads directly on the side of the filter)

Once the large acrylic UR-Pro filter is found, would then need to cut/machined it to fit in the Backscatter mount...

 

Other options I could see, that would require no machining and standard UR-Pro filters would be buying the AOI 55mm thread adapter.

 

- The mount would then be:

 

GoPro Supersuit > 55mm filter adapter > 55mm UR-Pro Filter > AOI 55mm thread adapter > AOI UWL-03 lens
 

But with the extra weight of the lens it probably wouldn't hold on my 55mm filter adapter, and if I ever upgrade the GoPro, the filter adapter is no longer available for more recent models (one of the reasons I didn't upgrade, as previously discussed)

 

A more stable option would be to get the more sturdy and universal  Inon 67mm filter mount for the supersuit -

 

There are no 67mm UR-Pro filters available, so the mount would then be:

 

GoPro Supersuit > Inon 67mm filter mount > 67mm to 55mm stepdown ring > 55mm UR-Pro filter > AOI 55mm thread adapter > AOI UWL-03 lens

But clunky as well, and what of vignetting with all those adapters...

 

Not easy, not easy...

 

 

Edited June 24 by bghazzal

[DreiFish]

12 hours ago, bghazzal said:

 

Hmmm, logisitics are a tad tricky, my spectrometer is still at the cleaners, all that 😁 😉

Looking at your tests for the ambient filters, I'm wondering if would it make sense to try shooting a still of a colour checker on land, first without, then with the UR-Pro and original Magic filters on.

With high CRI strobe and WB set to the strobe's, it should be close to neutral lighting on the raw file, right?
 

Then maybe the individual colour reading differences, especially blues and greens squares, could be measured in software (I'm really bad at image software, but I guess photoshop or lightroom can give such readouts?)

Wondering if this could be a quick and dirty way to get a rough curve highlighting the main differences between the two filters, and then seeing what best matches Rosco / Lee graphs in terms of required filtering range.

 

 

 

 

If you shoot the test chart illuminated primarily by a strobe with WB set to the strobe's 'native' temperature (which, actually, probably isn't the advertised temperature from the manufacturers), and with a red filter on the lens, what you're going to get is not a 'neutral' picture -- instead, you'll see exactly what filtration effect the on-lens filter has on the light spectrum produced by the on-lens filter.

 

You can then apply white balance in post to a white or grey square on the color checker and the resulting adjustments will give you an idea of what the filter is doing. So this approach is a valid one for comparing one filter to another and to see  which parts of the spectrum the filter attenuates.

[bghazzal]

19 hours ago, DreiFish said:

 

If you shoot the test chart illuminated primarily by a strobe with WB set to the strobe's 'native' temperature (which, actually, probably isn't the advertised temperature from the manufacturers), and with a red filter on the lens, what you're going to get is not a 'neutral' picture -- instead, you'll see exactly what filtration effect the on-lens filter has on the light spectrum produced by the on-lens filter.

 

You can then apply white balance in post to a white or grey square on the color checker and the resulting adjustments will give you an idea of what the filter is doing. So this approach is a valid one for comparing one filter to another and to see  which parts of the spectrum the filter attenuates.

Thanks - I'll try this with GoPro raw stills - I don't have a colour checker or a strobe but do have a calibrated Whibal grey card, and the Backscatter MW4300 video light is supposed to be calibrated to 6000K in wide mode (for pairing) - not sure if the is truly accurate (I do find it quite warm in post...), but it's the best I can do here, since I also don't have a light-meter.

 

I mentioned colour checker more for the colours themselves - my idea was that if it's possible to read colour values in lightroom, then it checking each colour square without then with the two filters, especially the greens and blues, could give an idea how the on-lens filter affects specific ranges in the spectrum.

I'll give it a go.

EDIT - I ran the test on the GoPro and LX10 - test details are here
Working from the LX10 raw files, I get the following data for WB in lightroom, with different filters:

 

No filter, raw, camera WB set to 6000K to match light supposed to be 6000K

- white balance "as shot" 5300K, +19 Magenta

- corrected white balance 5000K, +28 Magenta

UR-PRO acrylic
corrected WB 2350K, -14 Magenta

UR-PRO glass 55mm
corrected WB 2250K, -8 Magenta

UR-PRO (?) glass 2
corrected WB2300K, -23 Magenta

Keldan SF -1.5
corrected WB 3050K, -27 Magenta

original Magic filter
corrected WB2950K, +24 Magenta

Howshot
corrected WB2200K, -32 Magenta

However, I'm wondering if there is a way to use LR to measure the impact of the filter on greens/blues? I didn't have a full colour checker but do have a few colour squares in my test picture.

Once I'm here:

Is there any way to measure the effect of the filter is having on each of the colour squares?
Data seems identical when I move the dropper on both images, so I must be doing something wrong.

The idea would be to pinpoint broad differences in the way each filter affects a colour range, to highlight differences.

 

Thanks!

b

 

 

Edited June 25 by bghazzal

[bghazzal]

I ran the test raw stills again today, testing just the UR-Pro acrylic and the Original Magic filters.
White balance was set to 6000K then to NATIVE (flat).

 

Results are pretty much identical on both setting, and as follows:

 

NO FILTER, WB set to NATIVE

As shot WB 5500K Magenta +13 

Corrected WB 4900K Magenta +11

 

(for reference NO FILTER, WB set to 6000K, supposed calibrated light output gives:

As shot WB 5200K Magenta +6 Corrected WB 4900K Magenta +9 - filtered readings are identical)

 

UR-PRO acrylic, WB set to NATIVE

Corrected WB 3100K Magenta -44

 

Original Magic, WB set to NATIVE

Corrected WB 3500K Magenta +20


As observed on the LX10, a major difference is that the rebalancing with UR-Pro gives a negative Magenta reading, whereas rebalancing with the Magic gives a positive Magenta reading.

If i'm calculating correctly, based on GoPro raw files readings:

- the UR-Pro SW Cyan filter warms the light by 2400K (5500K to 3100K), and increases the magenta by +31 units (+13 to -44)
 

- the original Magic filter warms the light by 2000K (5500K to 3500K so 400K cooler/less warm than the UR-Pro), and decreases the Magenta by -31 units (+13 to +20)

There is a 400K warming effect difference between the two filters, which is not much, however for the original Magic to match the UR-Pro's handling of magenta, you would need to boost magenta by +76 (+14 > +62)...

 

At a quick glance in lightroom, setting the original Magic's wb to 5900K and boosting the magenta to +62 gives a very similar result

 

 

Modified original Magic, with WB set to 5900 and Magenta boosted to +76

 

I posted more illustrations in this thread
 

Results are now visually quite close for such rough measurements.
But while the results are visually similar, the histogram still shows significant differences in colour distribution.
To fine-tune this, it would be interesting to get readings on main individual colour channels (squares) as well, see how both filters affect them, but I'm not sure how to approach this.

 

cheers

b

Edited June 26 by bghazzal

[bghazzal]

As a follow-up, I ran some tests today on the GoPro with a combination of two stacked gels, one original Magic Filter gel and one Keldan Spectrum SF-15.

 

Tests were done on an overcast morning, in murky greenish-blue water, in tropical ambient light only (east Bali).

As in the previous tests, GoPro7 was on flat profile (WB native and GoPro colours), auto ISO locked at min 100 and max 1,600 (which is what I usually use for wide angle tropical ambient light shooting with the GoPro, max 800 if possible), 4K, 60fps, medium fov.

 

Note: AOI wide lens was not used as the person handling the camera for these tests - who happens to be my wife - couldn't be bothered with the extra weight 😅

 

Quick summary: not usuable / a practical solution at depth.

- good in the shallows (the Keldan gel evens out the Magic Filter), but too much light loss at depth (combined exposure loss should be Keldan -1.5 Ev and Magic -1.6 Ev so a total of roughly -3 ev - too much to handle for the GoPro at depth in today's murky conditions, footage was grainy, meaning the camera probably pumped up to 1,600 ISO around 15m)

 

The other issue is the Magic Filter's orange/magenta spike at depth, already noticed on the previous test. The Keldan gel doesn't help with this, and simply reduces the colours data making it to the sensor.

 

Here are some quick screenshots:

 

- 8m depth, on a flat plateau
worked well given conditions -  white balanced and quickly graded in FCPX (looked good just after setting the WB, plenty to work with...)

 

 

 

 

However, deeper and with less light on the slope, things got ugly...

18m depth, on the slope
 

 

While the footage isn't pure garbage, it's quite desaturated (which looks ok-ish) but the issue is really that's there's really not much left to work with.

The marked orange/magenta dayglo boost of the Magic, already noticed on the previous tests, is still there, waiting to be revealed....
All it takes is boosting the highlights+midtone saturation to the max, to see what colour info we really have to work with:

 

 

Ouch...

 

The other issue being the grain - ok, conditions were really not great and got worse during the dive, low light and particules, but it's still very grainy - other footage is worse in this respect.

This is something to keep in mind when working on stacked filters - on my aging GP7 at least, -3Ev is too much for the camera to be practical, unless staying above 10m in tropical light.

 

I guess a loss of -1.5 to -2 Ev is the sweet spot.

 

 

Edited July 1 by bghazzal

[Jochen]

I have been following this discussion with great interest, and learned a lot, thank you.

 

I recently bought a Gopro and I'm planning to film in flat profile and colour grade the video with Davinci Resolve in post.

 

Does it make sense to use a grey card and  start each recording with videoing the grey card at the beginning so that I have a "white reference" to be used in post?

 

[bghazzal]

39 minutes ago, Jochen said:

I have been following this discussion with great interest, and learned a lot, thank you.

 

I recently bought a Gopro and I'm planning to film in flat profile and colour grade the video with Davinci Resolve in post.

 

Does it make sense to use a grey card and  start each recording with videoing the grey card at the beginning so that I have a "white reference" to be used in post?

 

Thanks Jochen, glad you find the discussion interesting!
 

For the grey card, I would say yes, if you can be bothered to do it.
It will be more precise, and having the same exact same reference (especially a good one like a wb card) on different clips make life easier in post.
 

The only issue is that white balance is also relative to the distance to a subject, and you will be holding it at arm's length on a cam with a relatively wide lens, which might be a little close, but it still usually works quite well.

And even if it doesn't, it's only one possible reference point.
Since you'll be picking the WB setting point in post anyway, so if you find it doesn't work as well on a shot, you can always choose another one.
 

I didn't carry a card for most of the GoPro footage I've shot, but found myself using mostly the same references in post for tropical ambient light, mostly:

Other diver's tanks (works really well, especially aluminium tanks like AL80s), diver's bubbles, also the classical sand (sand can be a bit of a hit and miss depending on conditions and sand type), or bits of white rocks in the sand. Certain light skin tones / fin colours (white Gull fins...) can be good as well.

If there is a reef, there's often a piece of bleached coral somewhere (tips) or other whitish elements in the substrate.

 

But filming a proper WB card will help with consistancy - just film it a couple of second at the start like you said, holding the card as far as you can reach and in the direction of your shot and ideally with the sun in your back, like you would for manual white balancing.

 

Looking forward to seeing the results. DaVinci is great for grading - if you put sufficient time learning to use it.

I'd started working on it but unfortunately my current laptop isn't powerful enough to get fluid playback, even running from an external SSD and using proxies, whereas FCPX works fine natively that way. There was the option to edit in FCPX, export to DaVinci to grade, but with no fluid playback on DaVinci it's a nightmare.

Planning on jumping in again when I have access to more processing power.

 

cheers

b

Edited July 19 by bghazzal

[Jochen]

thank you for the detailed answers! I do have a grey card so I will give it try in a couple of weeks in Komodo. 

 

For most dives I probably will have video lights with me as well. So I am thinking to use a flat colour profile without lights (and maybe a red filter) and auto white balance when using the lights.

 

Will post some results when I am back.

 

again thank you for your advise.

[Davide DB]

Plenty of posts about red filters 😜

[canislupus]

I was wondering if anyone has tried GoPro labs and the beta firmware for using “Log”, in the GoPro labs page says:

LOGB=logbase,offset - Super experimental, alter the log encoding for more dynamic range, or for a closer match with other camera’s log curves. i.e design your own flat profile. Ideal for use with 10-bit, and the existing flat color setting.

P.S. I’m posting here but for not creating a new post, I couldn’t find information about that.

[bghazzal]

15 hours ago, canislupus said:

I was wondering if anyone has tried GoPro labs and the beta firmware for using “Log”, in the GoPro labs page says:

LOGB=logbase,offset - Super experimental, alter the log encoding for more dynamic range, or for a closer match with other camera’s log curves. i.e design your own flat profile. Ideal for use with 10-bit, and the existing flat color setting.

No clue as I only have an older model, but I'm quite curious as to what this will change on such a camera (log is a little tricky to use on bigger cams as well from what I gathered).

While the results might expand grading options, I doubt it will be a "filter-killer" however
The gold-standard for manual UW white-balance is the Canon algo, which gives lovely results, but is most likely lovely because it shifts certain hues (greens to yellows) which gives a pleasing end result, rather than going for a truly accurate WB, and in this it is akin to using a filter.
Canon also does this on certain camera models, such as this currently viral old G7something model (the name of which escapes me) it seems to be something of a constant, generating aesthetically pleasing results ("Canon colours")

In the shallows, no problem, but below a certain depth (usually 10m but this is heavily conditions/location dependent) ambient light underwater image making is diving with a strong colour filter on the light source, and there's only so much camera manual balance can adapt to and compensate for before being thrown off-balance. Water filtration of sunlight, the effect on the colour spectrum, is more complex than a simple cooling of light temperature.

That said, anything that can record more flat data will help work on the footage colour balance in post, so I'd be curious to see the results of flat log footage shot on a GoPro!

Edited September 9 by bghazzal

[canislupus]

9 hours ago, bghazzal said:

No clue as I only have an older model, but I'm quite curious as to what this will change on such a camera (log is a little tricky to use on bigger cams as well from what I gathered).

While the results might expand grading options, I doubt it will be a "filter-killer" however
The gold-standard for manual UW white-balance is the Canon algo, which gives lovely results, but is most likely lovely because it shifts certain hues (greens to yellows) which gives a pleasing end result, rather than going for a truly accurate WB, and in this it is akin to using a filter.
Canon also does this on certain camera models, such as this currently viral old G7something model (the name of which escapes me) it seems to be something of a constant, generating aesthetically pleasing results ("Canon colours")

In the shallows, no problem, but below a certain depth (usually 10m but this is heavily conditions/location dependent) ambient light underwater image making is diving with a strong colour filter on the light source, and there's only so much camera manual balance can adapt to and compensate for before being thrown off-balance. Water filtration of sunlight, the effect on the colour spectrum, is more complex than a simple cooling of light temperature.

That said, anything that can record more flat data will help work on the footage colour balance in post, so I'd be curious to see the results of flat log footage shot on a GoPro!

Hi bghazzal,

I was not thinking about log a as an alternative to filters but as a “add-on”.

I will try to give it a try, for what I understand you can always go back to previous firmware in case that “not official” Gopro Lab” doesn’t work.

I like the idea also that with that firmware Gopro can read QR codes underwater and change settings alone.

Thanks for the info.

[Davide DB]

I don't want to sound like a pessimist, but I'll tell you what my impression was in an admittedly very limited use case.
Basically I had 2 GP11s, one GP8 and two GP5s, shooting the same scene in 50 cm depth. I started enthusiastically with gopro labs fw, 10 bit, high bitrate and log profile. My buddy was very skeptical and left his GP11 pretty much standard: 10 bit color Natural. Both had fixed WB 5000K.
We wasted several times doing color correction and grading of the gopro files with all features enabled only to arrive, after much effort, at the same quality as the standard one! 

I basically removed all the extra settings. By the way the GP with the high bitrate also consumed more battery while the most important thing for me was just the battery life.
In the end we left the gopro labs fw only because it was convenient to configure the various gopros on the fly with a ready-made qrcode with all the settings and exact time.
Again, maybe in another scenario the result would have been different (getting a creative look) but I remain very skeptical. As much fun as it can be to do the experiments, in the end no matter how hard you try, the sensor is what it is. The real mojo of these cameras is in the marketing.

[canislupus]

9 hours ago, Davide DB said:

I don't want to sound like a pessimist, but I'll tell you what my impression was in an admittedly very limited use case.
Basically I had 2 GP11s, one GP8 and two GP5s, shooting the same scene in 50 cm depth. I started enthusiastically with gopro labs fw, 10 bit, high bitrate and log profile. My buddy was very skeptical and left his GP11 pretty much standard: 10 bit color Natural. Both had fixed WB 5000K.
We wasted several times doing color correction and grading of the gopro files with all features enabled only to arrive, after much effort, at the same quality as the standard one! 

I basically removed all the extra settings. By the way the GP with the high bitrate also consumed more battery while the most important thing for me was just the battery life.
In the end we left the gopro labs fw only because it was convenient to configure the various gopros on the fly with a ready-made qrcode with all the settings and exact time.
Again, maybe in another scenario the result would have been different (getting a creative look) but I remain very skeptical. As much fun as it can be to do the experiments, in the end no matter how hard you try, the sensor is what it is. The real mojo of these cameras is in the marketing.

Hi Davide,

Lol, that was a reality check!

Good to know, before I even tried, the battery life reduction itself it’s big a limitation.

Thank you very much for the info

[Davide DB]

3 hours ago, canislupus said:

Hi Davide,

Lol, that was a reality check!

Good to know, before I even tried, the battery life reduction itself it’s big a limitation.

Thank you very much for the info

As @bghazzal's tests have shown, certainly in a real dive having at least a “flat” starting image helps a lot with color correction but in my case it was practically like shooting out of the water (50 cm in fresh water).

Under these conditions, a GoPro already gives its best with classic settings. There was almost no color absorption, and the color differences between the various GoPro, shooting the same scene, were due to the different orientation relative to the sun. So it was enough to set the WB to 5000K (sunlight) and everything was more or less fine.

Another nasty blow was to realize that filming the same static scene at 1080@50p with a GP5, GP8 and GP11, color aside (easily adjustable in post), the quality of the shot was virtually the same. There was not this abyss that we are led to believe by the hype. In the most extreme cases, a pass of Topaz and the shots were identical. Sad but true.

I have seen several videos with mind blowing creative looks obtained from the Log files but these examples only show that it is possible to work with it but none demonstrate that it cannot be done from the standard Protune settings.