Such circuits may be usable for motor control, but I really wouldn't recommend one for audio applications. A typical PWM will behave most linearly when it's near 50% duty cycle; linearity falls off badly at the extremes. In a class D amplifier, a zero-volt input signal is represented by a 50%-duty-cycle output. As a consequence the zero-crossing point, which is where the ear is most sensitive to distortion, is where the circuit behaves the most cleanly. If you try to use an H-bridge simply for direction switching and a separate transistor to modulate the output amplitude, it will be very difficult to achieve any sort of smooth behavior near the crossover point. This will be especially true if one is driving a reactive load where voltage and current are not in phase. The behavior of a class D amplifier when voltage and current are out of phase will be relatively clean and consistent when the voltage crosses from positive to negative. By contrast, in a circuit such as you describe, the transition from positive to negative voltage will trigger a sudden change in how the circuit handles the currents that are flowing in the load at the time of the change. With a practical load, such currents are likely to exist and to be significant. An abrupt change in them will almost certainly generate audible crossover distortion.
Your first concern in selecting a gate driver is to find one that can drive enough current to switch your selected MOSFETs fast enough for your application. As a rough estimation, you can divide the total gate charge of your MOSFET by the current the driver can sink/supply.
$$ t_{on}=\frac{Q_g}{I_g}$$
Using the worst-case values for IRF1405 and the slower of your two gate drivers, IRS4253:
$$ \require{cancel} \begin{align}
t_{on} &= \frac{260 \cdot 10^{-9} C}{180 \cdot 10^{-3}A} \\
&= \frac{260 \cdot 10^{-6} \cancel{C}}{180 \cdot \cancel{C}/s} \\
&= 1.44 \mu s
\end{align} $$
Off is faster, because this driver (which is typical) can sink more current than it can source:
$$ \require{cancel} \begin{align}
t_{on} &= \frac{260 \cdot 10^{-9} C}{260 \cdot 10^{-3}A} \\
&= 1 \mu s
\end{align} $$
If your switching frequency is 10kHz, each switch period is \$1/10000 = 100\mu s\$ and you will spend \$ (1.44\mu s + 1\mu s) / 100\mu s = 2.44\%\$ of that time switching. Probably acceptable, but you should calculate your switching losses and check.
Also, keep in mind this calculation is an approximation. The current specified in the gate driver datasheet is current into a short circuit, but a MOSFET gate isn't that. Unlike a short circuit, the gate voltage rises as it is charged, which will reduce the current the driver can provide. Also, your layout may introduce more inductance and resistance than there was in the test circuit the manufacturer used, further reducing current. Consequently, your actual switching losses may be higher than this calculation suggests.
In selecting the bootstrap capacitor, you want to make sure it's significantly bigger than the gate capacitance it will be charging, so that the bootstrap voltage doesn't sag appreciably when you switch. It also needs to supply whatever leakage current there is as long as you keep the high-side switched on. You can calculate these leakage currents, or just make the bootstrap capacitor way bigger to be safe. 100 times bigger than the gate capacitance should be good, so at least \$26\mu F\$. Bigger doesn't hurt much, so round up to a standard value or whatever you already have in the BOM or stock.
Since this capacitor is the power supply for the high-side gate current, you also want it to be very low impedance. It wouldn't hurt to parallel your big capacitor with some smaller \$100nF\$ decoupling capacitors very near the gate driver(s).
Selecting a bootstrap diode isn't terribly difficult. It needs to be able to withstand the reverse voltage when the H-bridge is switched high. Also keep in mind that you will lose the diode's voltage drop from the gate voltage. A Schottky diode might be nice for this reason, but depending on your circuit, you may not find one that can take the reverse voltage. A simple 1N4148 can take reverse voltage up to \$100V\$.
The reverse recovery time of the diode can also be relevant if your are switching very fast; 1N4148 has a reverse recovery time of \$4ns\$, so you will have to have the H-bridge switched low for significantly longer than that for the bootstrap capacitor to have time to recharge between cycles.
Best Answer
In my opinion you would need complementary transistors only for more linear/analog applications like in a (audio) power amplifier.
In an H-bridge you're using the MOSFETs as switches, they're either on or off. I would prefer to use FETs with the same Rdson (on resistance) though although as long as Rdson is low enough that might not even matter.
I would also pay attention to the threshold voltage but also here different Vt for NMOS and PMOS would not matter so much if you drive them properly with a large Vgs (which you should to make Rdson low).
So I would just try to find a suitable PMOS that can do the job and not worry about it not being complementary to the NMOS.