I haven't checked the details but what you might find is to get more power R3 needs to be reduce. What often happens is the resistor on the base (R3 = 1k) cannot supply enough base current.
Bootstrapping can help a bit - add C5 and split your R3 into two resistors, here R8 and R9.
You still need R8, R9 to be low enough to supply supply the base current.

A Darlington output stage lets you use higher value of R3 but then you lose 0.7V swing, which can be significant, on a 9V supply. However then you will need to add more diodes to bias the output stage.
Also make sure the DC bias point at the emitters is mid-supply to maximize swing.
The biggest problem with the amp is no emitter resistors on the output transistor emitters. If the output stage heats up it could get thermal run away. Thermally connecting (touching) the diodes onto the power transistors helps.
At the end of the day. It's a small amp but you need to *design* the amp, and that's not a simple task for a newbie.
This perhaps summarized a lot of the points. The feedback bias with the 470k helps DC biasing of the output stage. You design is hard to maintain the DC voltage at the emitters. You might need a resistor in series with the input to tame the gain. Change the value to suite the lower supply voltage.

[FWIW, you don't need such large driver transistors for the Darlingtons.]