Horn speakers are a vast many things; horn-hybrids (like they are, most of them), all-horns, large, very large, not so large, bonkers expensive, not so expensive, cheap, well-implemented, not so well implemented, astonishing sounding, good sounding, blah sounding, outright horrid sounding, passively configured, actively ditto, etc. Claiming it comes down to good vs. LOUD (i.e.: essentially low vs high efficiency, which is to say mostly direct radiation vs. horn variations) is hideously reductive and fails to acknowledge that low efficiency speakers can sound like crap (as in dull, lifeless, incoherent, strained, malnourished, etc.) as well as very, very good.
However speaking of the more fully flexed potential of either "camp," to my ears well-implemented and -designed horn speakers, be they hybrids or all-horns, that actually adhere to physics and lets size have its say (and they needn't be bonkers expensive to be so if one goes about it wisely and with an open mind) are in another league compared to the rather vast array of low efficiency speakers I've heard - not least when the horn variations are actively configured with the main speakers high-passed and further horn subs augmented. The presentation is simply more life-sized, uninhibited, effortless and dynamically real by comparison.
Ideally I'd go for an actively configured, large horn-based system. Where many if not most are comprising with horns, and thus fails to experience their fuller potential, is in regards to size and overall implementation. Granted, most can't go with a full-wave horn implementation in the lower octaves for obvious practical reasons, but even 1/4-wave iterations here can be truncated to negatively impact sensitivity, bandwidth and response smoothness. A midrange horn too small doesn't control directivity in its lower range and thus effectively stops acting like a horn here, the same with a lower mids to upper/central bass horn (and where the negative sonic effects with folded horns can be quite audible due to throat constrictions and too high air velocity in an effort to extend bandwidth), and in the subs region even a non-truncated 1/4-wave Front Loaded Horn will be quite large, certainly if the vicinity of 20Hz is to be reached, but as well with a 25-30Hz lower knee. If such sizes can nonetheless be willed in a home environment with quality, non-truncated FLH designs, it's bass reproduction that simply levels any direct radiating counterparts. A close-to-equal option (yet advantageous in the lower end) is using tapped horns instead which shaves off some size, but with a tune below ~30-35Hz it's at the expense of sensitivity and upper bandwidth. The least desirable sub bass horn variation in my mind, though sonically still very capable, is the truncated FLH.
Most main speaker horn users tend to go with smaller, direct radiating (and low eff.) subs for augmentation, but I find that to be a compromise too severe. They just don't blend well, not even as DBA's. Using horn-hybrid main speakers isn't necessarily ideal either, but here dispersion pattern matching at the crossover to the horn above can often be achieved more successfully. Using dual 15" woofer/mids vertically aligned is a good match to a large format horn above, and when you high-pass the woofers fittingly high (~80Hz on up) it frees up the woofers for an even cleaner reproduction in their band, while boosting headroom significantly. A star-quad array of 15" woofers is likely an even better option with the right (and big!) horn combination, while more readily matching the sensitivity of a properly sized midbass horn variant.
With regard to overall implementation I find active configuration to be ideal with horns for precise delay adjustments and notch placements, as well as steep crossover slopes to keep the horns within their "safe" frequency bands. It also provides for much better amp-driver control with the direct amp-to-driver coupling and the passive crossover out of the way between the amp and drivers. There's also the advantage of driver segment independency with separate amps for each driver section; whatever is required of the amp driving the woofers won't impact the rest of the frequency band at all. Some swear by 1st order slopes passively configured, and while it can no doubt provide excellent results as well, not least with lower wattage tube amps, I find the combined options with an active setup + similar SS amps top to bottom (for the best coherency) to be the preferred scenario. I'm sure others may disagree, but I'd be interested to learn whether that stance is supported by actual experience with an active setup as described.