The development of real-time audio applications for general purpose OS like Linux or MS Windows is a well-known and not trivial problem. Several real-time audio software with libraries for multi-purposes programming languages like C, C++ or Java have been developed. Although these solutions provide portability through multiple OS, they often introduce problems related to dependencies and versions alignment over the time. In the present article, the process to build a framework for real-time audio relying on the programming language Python is described. We selected Python because it offers a fast learning rate coupled with several modules that can be used inside real-time
audio projects. The advantages of this programming language but also the limitations in terms of execution time imposed by the interpreted code are both discussed. In particular, a procedure of code refactoring to obtain the real-time execution is presented. This procedure involves the inclusion of compiled code generated by Cython and C language, only where strictly needed. In the last part, some case studies related to the implementation of Virtual Analog models are presented introducing different amounts of compiled code and measuring the execution time.
