Cross-lingual speaker adaptation for HMM-based speech synthesis and speech-to-speech translation

Demonstrations

Utterances
No.1
No.2
Input English waveforms
Correct English texts
One sultry morning last June, Allen Murray, then four months into his job as chairman of Mobil Corporation, strode into a headquarters conference room.
Discarding his jacket and lighting a cigar, he waded into an hour's verbal fencing with a visiting banker.
Automatically recognized English texts
ONCE ULTRA MORNING LAST JUNE ALAN MURRAY THAN FOUR MONTHS INTO HIS JOB AS CHAIRMAN OF MOBIL CORPORATION STARTED TO A HEADQUARTERS CONFERENCE ROOM
DISCARDING HIS JACKET IN LIGHT IN A CIGAR HE WAITED INTO AN HOUR'S VERBAL FINANCING WITH A VISITING BANKER
Texts translated into Japanese
彼の仕事モービルコーポレーション会長は昨年6月に一度ウルトラ朝アランマレーは4ヶ月本部会議室を開始
彼は銀行を訪れるとの1時間の口頭ファイナンスに待った葉巻の光の中で彼の上着を破棄
Japanese average voice (non-adapted synthetic speech)
Japanese synthetic speech (adapted synthetic speech)



These voices have been built through collaboration and cooperation with many researchers of the EMIME project
Reference: e.g.
K. Oura, J. Yamagishi, S. King, M. Wester, K. Tokuda
``Unsupervised English-to-Japanese speaker adaptation for HMM-based speech synthesis,''
ASJ 2009