Techniques are presented for dynamically controlling an audio crossfade from a first audio track to a second audio track. The first audio track may be decoded and buffered to a first buffer. A user may select a second audio track at some point during playback of the first audio track. The second audio track may be decoded, and a processor may access the decoded second audio track and the buffered first audio track and mix the two audio tracks. The mixed audio track may be inserted in the same first buffer near or at the read pointer of the first buffer, such that latency between the user selection and the crossfade into the selected audio track may be minimized. The insertion of the mixed audio track may depend on various factors, including the type of audio file and/or the decoder type used to decode the second audio track.