The structure consists of a main melody, usually short, repeated many times, with variation and elaboration at the performer's discretion. There is a second, contrasting melody, usually with higher notes, which is introduced once before returning to the main melody. The tarana may include a Persian couplet, and may use syllables from sitar or tabla such as "dar-dar" or "dir-dir"; singers might recite full compositions (e.g. tihais, gats, tukdas) within the body of the tarana.

In the words of Thakur Jaidev Singh, an influential commentator on Indian music:

was entirely an invention of Khusrau. Tarana is a Persian word meaning a song. Tillana is a corrupt form of this word. True, Khusrau had before him the example of Nirgit songs using ?u?k-ak?aras (meaningless words) and p??-ak?aras (mnemonic syllables of the mridang). Such songs were in vogue at least from the time of Bharat. But generally speaking, the Nirgit used hard consonants. Khusrau introduced two innovations in this form of vocal music. Firstly, he introduced mostly Persian words with soft consonants. Secondly, he so arranged these words that they bore some sense. He also introduced a few Hindi words to complete the sense…. It was only Khusrau’s genius that could arrange these words in such a way to yield some meaning. Composers after him could not succeed in doing so, and the tarana became as meaningless as the ancient Nirgit.^[3]

The thillana from Carnatic music is based on the tarana, according to Balasaraswati,^[4] and is widely used in dance performances.^{[citation needed]}