TotalVoice
checking
workers
queue
memory

Text

Plain text or SSML. Accepted: <break>, <prosody rate>, <emphasis>, <sub alias>, <say-as>, <p>, <s>.
First sound
Total
Audio
Parts
The tender gives synthesis 400 ms for the first sound, inside a total chain of 3,500 ms. "First sound" measures exactly that: from the request leaving to there being audio to play. Stop is barge-in: whatever was still to be generated is discarded.

Recordings

Every render is kept here so voices and settings can be compared.
Nothing rendered yet.

How it is split

What the engine will synthesize, in order. The first part is deliberately cut shorter: it is what puts a voice on the line in time.
Type something to see it.

Holding phrases

The ones that cover the silence while a source system is queried.