Personal tools
You are here: Home → Downloads → TTS average voice models
Archived Site

This is an old static copy of the original Plone based site. Some functions will not work.

 

TTS average voice models

English, Finnish and Mandarin average voice models for HMM-based speech synthesis

Link Mandarin SPEECON Average Voice Model
The Mandarin average voice models were created from the SPEECON databases (The Mandarin SPEECON speech databases are proprietary databases of Nokia Research Center, P.O. Box 100, FIN-33720 Tampere, Finland) and available from ELRA. The release of these models are subject to the ELRA conditions and STRAIGHT conditions, that is, the use of these models is restricted to only research use.
Link English SI284 (WSJ0 + WSJ1) Average Voice Model
The SI84, SI284, and WSJCAM average voice models were created from LDC WSJ0, WSJ1 and WSJCAM0 speech databases (The LDC speech database is proprietary database of Linguistic Data Consortium, 3600 Market Street, Suite 810, Philadelphia, PA, 19104-2653, USA.) below. The release of these models are subject to the LDC conditions and STRAIGHT conditions, that is, the use of these models is restricted to only research use.
Link English SI84 (WSJ0) Average Voice Model
The SI84, SI284, and WSJCAM average voice models were created from LDC WSJ0, WSJ1 and WSJCAM0 speech databases (The LDC speech database is proprietary database of Linguistic Data Consortium, 3600 Market Street, Suite 810, Philadelphia, PA, 19104-2653, USA.) below. The release of these models are subject to the LDC conditions and STRAIGHT conditions, that is, the use of these models is restricted to only research use.
Link English WSJCAM0 Average Voice Model
The SI84, SI284, and WSJCAM average voice models were created from LDC WSJ0, WSJ1 and WSJCAM0 speech databases (The LDC speech database is proprietary database of Linguistic Data Consortium, 3600 Market Street, Suite 810, Philadelphia, PA, 19104-2653, USA.) below. The release of these models are subject to the LDC conditions and STRAIGHT conditions, that is, the use of these models is restricted to only research use.
Link Finnish SPEECON Average Voice Model
The Finnish average voice models were created from the SPEECON databases (The Finnish SPEECON speech databases are proprietary databases of Nokia Research Center, P.O. Box 100, FIN-33720 Tampere, Finland) and available from ELRA. The release of these models are subject to the ELRA conditions and STRAIGHT conditions, that is, the use of these models is restricted to only research use.
Link Japanese
Japanese JNAS Average Voice Model - coming soon
Document Actions