When a sound wave propagates from a source through a medium and is reflected on surfaces before reaching microphones, the measured signals consist of a mixture of the direct path signal with delayed and attenuated copies of itself. This phenomenon is commonly referred to as "echoes", or "reverberation", and is generally considered as a nuisance in audio signal processing. After a gentle introduction to relevant concepts in acoustics and signal processing, this seminar will present recent works showing how acoustic echoes can be blindly estimated from audio recordings, using either non-linear inverse techniques or machine learning. We will then show how the knowledge of such echoes can in fact help some audio signal processing tasks such as the separation, enhancement or localisation of sound sources.