A recent Twitter conversation about the pros and cons of graduated neutral density filters (often just called GNDs) made me think about my own approach towards handling high-contrast scenes. I’m not using GNDs anymore, rarely did so in the past, and I prefer exposure blending on the computer, or even just using “software GNDs” in a single exposure. I find it far more flexible* and prefer to have “pure data” without the (unalterable) effects of the filter in my original/initial exposures.
But that’s just me and in the end, the discussion came down to minute details and fringe cases and the conclusion, as it is so very often, is that there’s no right or wrong of course. That’s legit, and perfectly fine: whatever works for you will be great if it yields results that you’re happy with. This could be the end of this article. Except…
Well, let’s go back a little bit to briefly talk about when “to GND, or not to GND”.
There are hard and soft edge GNDs – ie. their transition from transparent to darkening can be abrupt, or more gradual. There are also so-called reverse-GNDs, where the center is abruptly dark along a hard edge, and they get lighter towards the top of the filter. This is to help with those situations where the horizon is very bright (for example, around or just after sunset).
Whatever GND you’re using – you can’t create a good-looking transition with a GND if your sky (or more generally speaking the portion of the frame that you’d like to darken) can not be separated from the transparent area of the filter by anything than a straight line. Mountains, hillsides, forests… any typical landscape photo that is not a seascape with a flat horizon is usually not very suitable for using a GND.
And here is a case against graduated neutral density filters despite the fact that there’s a straight horizon line:
This image is blended from two exposures: a 3-second exposure for the sky (upper third of the frame), and a 10-second exposure for the foreground. It didn’t need luminosity masking because the horizon is a straight line and a hard transition anyway. I didn’t even bother to make a selection and simply used a brush to manually mask the sky, and that was it. Total time required: 40 seconds perhaps? :)
Both exposures were also made at ISO 200, one stop above the base ISO of my camera (100) – and here’s why: the clouds were moving relatively fast. At three seconds, they are well-defined and only a small amount of motion blur appears at the edges of the frame. Had I used ISO 100, I would have needed a six second exposure time. More of the clouds would have lost their definition.
Now had I used a graduated neutral density filter (a two-stop in this case) to bring down the sky, I would have ended up with a ~12-second exposure for the entire frame – and with a partially streaked, partially solid look of the clouds. Something that I find aesthetically rather undesirable, most of the time.
And had I used ISO 100, the exposure time with a GND would have been ~25 seconds – even more streaking, but not enough to produce entirely blurred clouds yet either. I might have added a 3-stop neutral density filter then, to turn this into a really long exposure, with an entirely different look. However, the resulting exposure time (well over 2 minutes) wouldn’t have been possible, at all: the tide was coming in and there was no way to keep the tripod steady for that long a time at this camera position.
The point is, as before: there’s no right or wrong. You can’t say you “prefer to get it right in camera (with a GND)” when there’s no actual way to do it. Blending the image above from two exposures on the computer was the only way to achieve the look I desired, and work in these conditions. Don’t exclude a technique because you reject it based on some self-made ideology. It may turn out to be a restriction instead of a helper. ;)
Allow me to add on thing though. The consensus of the Twitter conversation was that, even if everything works out for using a GND, most images do require some additional attention on the computer, to create a seamless, and most of all logical transition between foreground and sky. A three-stop GND would have been too much for the photo above (and will be for many similar situations). With today’s camera sensors and their dynamic range, a two-stop GND is probably a much better choice – you can recover additional highlight and shadow detail from the file later, on the computer.
But if I need to work on that stuff on the computer anyway, it begs the question: why fiddle with the filters? ;) To have less work on the computer, perhaps – but then again, blending an image from two separate exposures, even if you use luminosity masking, doesn’t exactly add much to the total time it takes to develop a photo on the computer.
*) for example, a “software GND” allows you to not just adjust exposure, but also white balance, contrast, whites/blacks, texture, clarity or anything else that your software of choice allows to be used as a local adjustment.