real time use of mdx-net models for vocal separation?

[HeChengHui]

@Anjok07, thank you for your work and I am impressed with the mdx-net models for vocal separation. I would like to use it in real time (audio streaming in from a mic, separate vocals using python, stream the output to earpiece/speaker). From my research, it seems like mdx models are not made to do real time? Is that an architecture limitation or an io limitation? Wonder if you can shed some light on this possibility.