The project sounds super cool. > The more complicated aspect is the filtering required to defect only the user speaking during conversations (or noisy environments) If the app has training (for the stutterer), are you able to create/train a speech model from that data and use it to detect & filter the user's voice? Or is it just the ambient noise that's hard to filter out?