The objective of the Sugar Voice Project is to develop deep learning models for detecting voice changes in patients with diabetes. A non-invasive, fully remote, free, real-time screening tool may prove to have great potential to detect the disease early in an effort to prevent progression to complications. Subsequently, we will assess the clinical performance of these machine learning models using real-world clinical cases. We hypothesize that audio data can assist us in screening patients for diabetes. This affordable, non-invasive test can easily be conducted in the comfort of your own home, making it accessible to a wide range of individuals. Its potential to significantly enhance the identification of diabetic patients cannot be overstated.