OK, that’s not true.
HDR doesn’t suck, tone-mapping sucks!
OK, that’s not exactly true either. Tone-mapping usually sucks. It’s rarely true that a technique always sucks outright. There’s a time and a place for everything, just like bagpipes have their place in music, tone-mapping has its place in realistic natural landscape photography. A very very small place.
HDR is a general concept: a means of achieving a greater dynamic range in your image than what your camera sensor can capture with a single exposure. Tone-mapping is one way to achieve this. Manual exposure blending is another. Today HDR and tone-mapping have become synonymous, but HDR can be accomplished with a wide variety of techniques. Tone-mapping just happens to be the worst one.
There are times when tone-mapping makes sense such as for surrealistic or pseudo-realistic landscapes, architecture/interior shots, or any shot where the traditional tone-mapping defects (halos, ghosting, softness, over-saturation) are actually desired. Those are typically not attributes one wants for realistic natural landscapes. My issue with tone-mapping for realistic natural landscapes isn’t just with the current implementations (e.g., Photomatix), which will likely improve over time and get rid of those defects, it’s the concept itself.
With tone-mapping you’re taking multiple pixel values and averaging them. This is almost never what you want. Taking an example with two exposures, you usually want pixel values from one or the other, not an average. What that average ends up being is dependent on whether you’re using a local or global-contrast tone-mapping algorithm, and whatever knobs and levers your tone-mapping program provides, but ultimately you’re deferring that choice to the algorithm. The only place it sometimes makes sense to have an average is where the two exposures meet.
Worse yet are those who do this when the dynamic range fits entirely within a single exposure. They become so used to cranking images through the tone-mapping grinder that they don’t realize when it’s time to turn off the machine. If your post-processing technique has no variance and could be replaced with an automated script, it either means all of your photos are the same or you’re not playing enough of a role in deciding how they should look after they come out of the camera. Neither is very artistic, is it? Instead of spending 5 minutes waiting for that HDR merge to finish, spend 5 minutes thinking exactly how you want each part of the final image to look and learn how to get it there.
Now some can use tone-mapping in a subtle way such that it’s hard to tell they used it at all. This is fine – and better than the way most people use it – but what’s the point? You can almost always achieve the same effects without tone-mapping and take less time to get there.
All of this is why I prefer to manually blend exposures. Well, I actually prefer not to have to blend exposures (and usually I don’t have to). In the cases when I do need to blend exposures, the overwhelming majority of the time I can get all the dynamic range I need with just two exposures. Using manual selections allows me to decide what exposure is used on each area of the image. It’s also no more difficult (after a time) and in fact can be done faster than tone-mapping which often takes forever even on a fast computer. Exposure blending can also be used for more than increasing dynamic range too (for instance, increasing depth of field by blending exposures with different focus points; blending two exposures with different ISOs, one to stop motion, the other to reduce noise; blending the same composition taken at two points in time, and many other ways).
So I prefer HDR if done by manual exposure blending, and almost never if done by tone-mapping. The tone-mapping algorithms will get better (and will likely use a combination of heuristics and pixel-choosing rather than pixel-averaging and fix the obvious defects of current implementations), but I haven’t seen anything yet that produces better results than manual exposure blends. I used to use Photomatix and tone-mapping, but found I could achieve better results without it, and haven’t used it in the last few years at all.
And while we’re on the subject, the notion that there has to be “detail” everywhere in a photo, and that all shadows are evil, needs to go away. It’s OK if part of the image falls off the left of the histogram, and, I know it’s heresy, but it’s also OK if some of the channels are blown out to the right of the histogram too, depending on the situation.
It’s also OK to use tone-mapping for natural realistic landscapes. Sometimes. But make sure you’re doing it for conscious artistic reasons and not just out of habit.