The MusicLM tool is designed to produce accurate and realistic music based on written descriptions. It utilizes a hierarchical sequence-to-sequence modeling approach to create music that maintains its quality and coherence for extended periods of time. This tool has the capability to incorporate both text and melody- enabling it to convert whistled or hummed tunes into the desired musical style described in the text. Moreover- it can generate music based on various inputs such as painting descriptions- instruments- genres- musician experience levels- places- and epochs. Additionally- the tool can generate multiple variations of the same written prompt and semantic tokens.