The NPN has + on its collector and less positive on its emitter.

The PNP Has + on its emitter and zero on its collector.

Therefore both have correct polarity voltages.

The voltage at the junction of R3 and R4 is half the supply voltage.

If R2 is replaced by a wire link, and R1 selected to give half the supply volts at TR1 and TR2 bases then they would both be biased in class B, and both would be non conducting.

Look at the waveforms on the bases of the output pair.

During the first half cycle the signal is going less positive.
This is reversing biasing the NPN transistor and increasing forward bias on the PNP transistor.
This means that the NPN remains cut off and the PNP conducts during this first half cycle.

In the next half cycle the signal on the bases is now increasing in a more positive direction.
During this half cycle the NPN conducts while the PNP is cutoff.

So the NPN is off and the PNP is on during the first half cycle, and current flows through the loudspeaker as C1 charges.

The NPN is on and the PNP off during the second half cycle, and current flows the other way through the loudspeaker, as C1 discharges.

R2 is added to give a small forward bias to both transistors thereby avoiding crossover distortion. (see the page on push-pull power amplifiers).