Neural instrument synthesizers have made it possible to clone the timbre of a solo instrument, letting producers reuse a desired sound in new music. But in practice, the instrument a producer wants to clone is rarely heard alone; it is usually embedded in a fully produced, multi-instrument track. A new preprint on arXiv, titled "Point-and-Play: Instrument Timbre Cloning from Music Mixtures," directly addresses this gap.

The paper's title suggests a user-friendly workflow: point at an instrument within a mixture and play it, with the system cloning that instrument's timbre from the mixed audio. The abstract notes that existing methods rely on solo recordings, which limits their usefulness in real-world production. By focusing on mixtures, the work aims to make timbre cloning more practical for music producers.

Because the provided abstract is cut off mid-sentence, the article does not include details on the proposed architecture, training data, or evaluation results. The source is a single arXiv preprint, so no independent verification or comparison with other approaches is possible here. Readers interested in the full method will need to consult the complete paper.