IssueTracker

When using the phoneme tag in SSML, AU_Wavenet_C performance better than US_Wavenet_C.

[SSML TEXT]
For Client Services MONEY MARKET Fund, On 06/23/2022, 0.6160 shares were bought for $50.00 For CLIENT SERVICES <phoneme alphabet="x-sampa" ph='"spE.SQl.aIzd'>SPECIALIZED</phoneme> PORTFOLIOS, On 06/23/2022, $5324.12 of long term earnings was reinvested as 118.5770 shares. For CLIENT SERVICES <phoneme alphabet="x-sampa" ph='"spE.SQl.aIzd'>SPECIALIZED</phoneme> PORTFOLIOS, On 06/23/2022, $2073.60 of dividend earnings was reinvested as 46.6610 shares. For CLIENT SERVICES <phoneme alphabet="x-sampa" ph='"spE.SQl.aIzd'>SPECIALIZED</phoneme> PORTFOLIOS, On 06/23/2022, $5604.33 of short term earnings was reinvested as 124.8180 shares. For CLIENT SERVICES <phoneme alphabet="x-sampa" ph='"spE.SQl.aIzd'>SPECIALIZED</phoneme> PORTFOLIOS, On 06/23/2022, $749.62 of long term earnings was reinvested as 16.6950 shares. Would you like to repeat or continue?


Reproduce steps:
1. Create request_en_US_Wavenet_C.json and request_en_AU_Wavenet_C.json with the SSML above.

2. Create 'request_en_US_Wavenet_C_v1.wav' and 'request_en_AU_Wavenet_C_v1.wav' by the curl command below.
curl -X POST \
-H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
-H "Content-Type: application/json; charset=utf-8" \
-d @request_en_US_Wavenet_C.json \
"https://texttospeech.googleapis.com/v1/text:synthesize" > request_en_US_Wavenet_C_v1.txt

base64 request_en_US_Wavenet_C_v1.txt -d > request_en_US_Wavenet_C_v1.wav

3. The AU_Wavenet_C performance better than US_Wavenet_C on the second Date of "06/23/2022" that has an slight gap between "20" and "22".