tts (text-to-speech) is not such a big problem
there are some web services that can do it pretty good
like: Google translate (has English pronunciation now), Ispeech, loquendo engine ...

speech-to-text is my problem
i can't exactly tell why aren't there any online services that can do this, except because of potential slow upload of audio files to server
obviously i have to implement it on the client, which is in this case N900 and Maemo