[Trac] [AllemaniACs] #66: Improve speech synthesis (speed, fancyness, usability)

AllemaniACs trac-notifications at lists.kbsg.rwth-aachen.de
Tue Aug 5 17:39:29 CEST 2008


#66: Improve speech synthesis (speed, fancyness, usability)
-------------------------+--------------------------------------------------
 Reporter:  stf          |       Owner:  stf                               
     Type:  enhancement  |      Status:  new                               
 Priority:  minor        |   Milestone:  GO2009                            
Component:  RCSoftX      |    Keywords:  speech synthesis, HRI, TTS, athome
-------------------------+--------------------------------------------------
 Our current speech synthesis system uses the Festival TTS software with
 the kal-diphone voice. The system exhibits an almost unbearable latency,
 does NOT offer feedback on its status yet and the voice an synthesis
 itself are not very natural.

 The goal is to achieve more natural speech synthesis with extended
 feedback (at least some status-info such as "busy/processing", "speaking",
 "idle/ready") and preferably also integrated sound-playback (wav/mp3).
 This incorporates:
  * evaluation of alternatives to Festival
  * evaluation of other voices that sound more natural
  * evaluate and employ ways to influence the synthesis (per word/sentence
 intonation, voice-mode [singing,bored,tense,...])
  * speedup festival-engine usage by instantiating a festival-server once
 and then control this with remote commands instead of spawning a new
 festival-instance for every utterance (STF: already implemented and
 working on some systems, but due to a bug in the current festival package
 in Fedora not committed yet)
  * set up a collection of parameterizable sentences that allow for
 automatic variability (e.g. I am (going|moving|on my way) to %s).

-- 
Ticket URL: <https://lechuck.informatik.rwth-aachen.de:2342/trac/ticket/66>
AllemaniACs <http://robocup.rwth-aachen.de>
AllemaniACs Trac


More information about the fawkes-trac mailing list