search

MycroftAI / mimic3

A fast local neural text to speech engine for Mycroft
GNU Affero General Public License v3.0
1.08k stars 103 forks source link

SSML doesn't work as expected #55

Open Slyke opened 10 months ago

Slyke commented 10 months ago

Describe the bug SSML doesn't work as expected

To Reproduce Steps to reproduce the behavior:

  1. Click 'Enabled SSML'
  2. Paste the following: 
    <speak>
<prosody rate="10%">
The weather today is rainy and clouds.
With a current temperature of 7 degrees, with it feeling like 7 degrees.
The highest temperature will be 9 and a lowest of 5.
Have a good day.
</prosody>
</speak>

    or

    <speak>
<prosody rate="10%">
<break time="1000ms"/>
The weather today is rainy and clouds.
With a current temperature of 7 degrees, with it feeling like 7 degrees.
The highest temperature will be 9 and a lowest of 5.
Have a good day.
</prosody>
</speak>

Expected behavior Everything should be at 10% speed.

Log files There are no logs

image

Environment (please complete the following information):

Additional context Seems to happen with any voice. It's like the prosody parameter is ignored for the first sentence. If I tell it to use a different voice from the begging, the first sentence will be in the default voice before it switches. It does this when running on Home Assistant, or when running in Docker on my main computer.