Most traditional approaches to voice isolation are essentially a combination of EQ and filtering, it's interesting to see an AI-based approach to the problem.