Google MusicFX: Transform Text Into Unique Soundscapes with AI

Google has launched MusicFX, an online service that generates music based on text queries. Furthermore, the product utilizes Google’s MusicLM model, and each audio file contains an inaudible watermark created via DeepMind’s SynthID, enabling the traceability of the file’s origin.

MusicFX opens new horizons for musicians, producers, and enthusiasts to experiment with music creation. However, it sets boundaries to safeguard the voices and styles of original artists, as it will not generate music from queries mentioning specific artists or containing vocals.

The service is part of AI Test Kitchen, Google’s platform designed to provide early public access to its AI products and gather user feedback. MusicFX is available to users in the USA, Kenya, New Zealand, and Australia.

Copyright protection is achieved by introducing distortions to the spectrogram of generated audio, imperceptible to the human ear but extractable using DeepMind’s SynthID verification tool.

Users can delete their generated content within a session, yet after its conclusion, all data is anonymized and will be stored for up to 18 months. Google affirms that the data collected during interactions with MusicFX are not linked to user Google accounts. This data will be utilized for analytics and model enhancement.

The product is available at this link.

Notify of

Inline Feedbacks
View all comments