The goal of this residency is to propose a gestural control of Artificial Intelligence models in real-time scenarios. The idea is to develop a dedicated control interface (hardware and software) in order to offer novel and innovative ways of generating electroacoustic sounds using the most recent models offered by deep neural networks. A dedicated electronic interface allows to develop instrumental gestures precisely linked to sound, while having an involvement of the body and producing expressive electroacoustic materials. AI systems have great potential to generate expressive and highly musical sound. Combining these two aspects of generating nowadays sounds would provide fascinating and unexpected expressiveness of machines.