Yes, constant. For the most important feature it would require two 1k point FFT's and two 1k point IFFT's (possibly reducible to 512 pt) every 22 useconds and

Yes, constant. For the most important feature it would require two 1k point FFT's and two 1k point IFFT's (possibly reducible to 512 pt) every 22 useconds and for an extended feature, 4k. I don't think Javascript is up to that which is why I mentioned somehow making calls to the native coded FFTW routines. If that can be done from Javscript there should be no problem. It's actually filtering of the browser's output audio samples that remains a total mystery to me. I'm not sure that even can be done which is why I brought it up.