Universal Music Group has signed a groundbreaking agreement with AI start-up SoundLabs to create AI-trained vocal models of the label’s artists. This innovative partnership aims to develop “official ultra-high fidelity vocal models” using the artists’ own voice data for training.
The first product of this partnership is a software named MicDrop, which can mimic an artist’s voice through advanced machine learning techniques. This technology provides artists with numerous creative possibilities, such as singing in foreign languages, recording while ill, or even continuing to produce music posthumously. Similar to synthesizer voices, MicDrop’s voice model opens up a range of creative applications, though its potential for live performance use remains unclear.
In a press release, UMG simply said that this technology “empowers artists and producers to explore bleeding-edge vocal transformations,” including various innovative processes like voice-to-voice, voice-to-instrument, speech-to-singing, and language transposition. These previously impossible vocal transformations present a new frontier for musical creativity.
UMG and SoundLabs have ensured that the custom vocal models created through this partnership will be available exclusively for UMG artists’ use and not accessible to the general public. This exclusivity underscores the importance of maintaining artistic integrity and control over the new technology.
”the technology empowers artists and producers to explore bleeding-edge vocal transformations, including voice-to-voice, voice-to-instrument, speech-to-singing, language transposition, and a myriad of previously impossible vocal transformations.”
“Together, UMG and SoundLabs are collaborating to allow UMG artists to create custom vocal models that will be available for their exclusive creative use cases and not available to the general public.”
SoundLabs founder BT was also very enthusiastic about the project, highlighting the ethical and consensual training of AI