In speech communication, background noise and acoustic echo are challenging obstacles to overcome. This is especially the case for applications where a loudspeaker is close to the microphone and the acoustic environment can change quickly, e.g., when using smartphones in hands-free mode. We propose a flexible speech communication demonstrator which not only allows easy implementation of various models (including our newest published models), but also realizes a full-duplex connection with on-device processing for arbitrary devices. Demo participants are encouraged to try our models on their own smartphones through our web interface. Presented models include a classical subband NLMS echo canceller with our own data-driven CGGN16 postfilter as well as our end-to-end DiffVQE2-S model. With the latter, to our knowledge, we are the first to showcase diffusion-based acoustic echo and noise control running in real time on a smartphone. Our framework will be publicly released.