Voice-Swap, an ethically trained AI voice company, and BMAT Music Innovators, a company that indexes music usage and ownership data using machine learning, have partnered to launch a new technical certification for voice and music AI models. It is designed to verify that audio content used to train voice models does not violate copyright or publicity rights.
While a former sound executive at Stability AI, Ed Newton-Rex, also recently launched a certification for AI models using licensed datasets through its non-profit Fairly Trained earlier this year, the new certification from BMAT and Voice-Swap takes a distinct approach. Fairly Trained is based on voluntary disclosures about the materials used in AI training using a written form, while the BMAT and Voice-Swap certification is a more technology-based certification.
Companies wishing to gain Voice-Swap certification and BMAT must allow the two to analyze a particular model's training data and compare it to the 180 million copyrights that BMAT has listed in its music index.
CRO of BMAT, Alex Loskos, says of the partnership: “At BMAT, we are driven to be pioneers at the intersection of music and technology. The certification of training datasets is an important step forward – the first of many ground-breaking initiatives we have planned. We are thrilled to have found such a like-minded partner in Voice-Swap, who has helped us realize this vision and who shares our passion for transparency in music.”
“We believe artificial intelligence should be beneficial for artists,” says Voice-Swap CEO Dan “DJ Fresh” Stein. “This next-generation audio certification could be the first step in solving the music industry's AI problems.”
from our partners at https://www.billboard.com/business/tech/voice-swap-bmat-launch-artificial-intelligence-technical-certification-1235688477/