Say returns both phonemes and speech...
Some experiments with multipart messages resulted in the addition
of the both fields to the return type selector in the form.
The forms now allow you to ask the server to return both text (phonemes)
and audio (speech).
Please note this is still very experimental, the current 'implementation'
is some sort of a kluge:
it only works if you use the Mosaic or Netscape for X browser
(or any other browser that offers some sort of remote control),
and have perl installed on your system, and do the following to make it work:
- add a few mailcap entries to eg.
your ~/.mailcap file, and
- make sure that the metamail program is in your PATH, and
- put the showhtml script in a directory
that is in your PATH.
(please don't forget to update the first line of it, to reflect the
location of perl on your system, nor to make it executable)
the script has been updated to also support Netscape
If the above doesn't make sense to you, then please simply use the usual
text or phonemes return types in the
form interface.
How does it work?
The idea is that the server returns a MIME multipart/parallel message that
contains both an html page (which contains the phonemes),
and an audio file (which contains the sound bytes).
Metamail then 'splits' the multipart document, and uses showhtml to pass the
html page back to Mosaic or Netscape, and its usual mechanism to play the sound bytes.
The audio bytes are encoded as base64.
Please Note:
I could not convince Mosaic nor Netscape to grok the MIME multipart/parallel documents
(it offers to save them (as binary file) to disk, instead of passing them to
metamail - does someone have a solution?),
therefore the form also offers to sent the output either as MIME message/rfc822 documents,
which Mosaic does pass to metamail
(although it creates an xterm window to display metamail output),
(but Netscape offers to save them to disk),
or as MIME application/x-metamail documents,
which Mosaic does pass to metamail if your mailcap file contains a
corresponding entry.
Feedback!
This is all quite experimental, and quite likely not the best way to do this -
it is simply the best I could get working right now (for me, at least),
so all feedback is welcome!
Information provided by the Say... Maintainer as part of the Say... text to speech interface