Get sound as a binary, not output it to speakers

Hi. I use speech dispatcher as a tts server middleware. I mean I give it text, and it creates sound. Its main part is interfaces: one interface to the client and another one to modules. But as I understand, currently there is no interface for getting sound by a client. I mean I give it text, but the sound created should go not to ALSA/PulseAudio, but to one of codecs installed on a host (i.e flac, but a dummy "codec" returning input should be also present) and the result should be returned to the client that has requested the synthesis. It is useful if we wanna use the sound in other way than directly output it to speakers, i.e. we may send it into a browser (it's my use case), or we may record it into a file.