I'm trying to drive a speaker from an ESP8266 for a project, and found an article suggesting (ab)using I2S as a 1-bit Pulse Density Modulation DAC. Apparently that's pretty noisy, so the article gives the following circuit diagram for the low pass filter:
The first RC does indeed act as a low pass filter.
The second capacitor (the one on the right), combined with the load impedance presented by whatever is connected to the audio out, acts as a high pass filter.
Together, these two filters act as a band pass filter.
It's simply an AF (audio frequency) coupling capacitor, removing any DC that is on the signal.