I developed a 120V 60Hz sinewave inverter prototype switching around 50khz a few years ago, with a full bridge of SKP10N60 triplets, powered from rectified 230V 50Hz. All hard switched. It was my first attempt at high power class D. Efficiency was around 87% at 1800W. Not amazing at all.
Why? These IGBT exhibit high tail losses at low currents, idle losses of the coverter were around 50W. Gapped ferrite output inductors were contributing to idle losses substantially, the gap was long and the flux leaking was producing high losses in the windings. There was not much room for bigger output inductors allowing a lower switching frequency.
The next inverter prototype was half bridge, with SGH80N60UF paralleled with SPW35N60CFD (IGBT turns off, then 400ns delay, then MOSFET turns off). It had a PFC front end producing 420V on a center tapped capacitor bank. All was soft switched with magnetic snubbers, and I developed a PIC-assisted low power mode with a modulation scheme allowing discontinuous current in output inductor. The PFC also had a low power burst mode. I got <10W idle losses and 94% efficiency at 1800W that way.
Then I started with audio class D and MOSFET full bridges, keeping supply voltage below 200V. I'm getting high-power efficiency figures around 90%-93%-95% from mains input to 2-4-8 ohm speaker that way.
The biggest problem with IGBT is the compromise between Vce-sat and turn-off delay. The delay is not a problem in SMPS but it hurts THD performance in audio class D (not practical for self oscillating), except for LF amplifiers.
There are new amazing IGBT like IGP50N60T ("TrenchStop and FieldStop"), that exhibit an extremely short tail like ultrafast IGBT and low Vce-sat like standard speed IGBT, but there is still a substantial turn-off delay (300ns). I have recently used that one in a PFC, but still paralleled with a MOSFET that is turned off slighlyt later, resulting in around 400ns overall turn-off delay.
There are also newer ultrafast IGBT (see STGP30NC60W), but these still exhibit not-so-low Vce-sat, similar to a MOSFET+Schottky+HyperFastDiode combo (except maybe at very high currents) and are not intended for paralleling (negative temp coeff. in Vce-sat). There is still an advantage in part count and less stuff needs to be mounted to the heatsink.