The project
This customer-service software vendor wanted to turn written application responses into audio. The engine needed to connect to its tools and produce speech suitable for a conversation.
Quality depends on immediately audible details, such as a cut-off sentence or a mispronounced word. These needed correcting while keeping response time acceptable.
We developed the speech synthesis engine and its integration interface. Main initial defects were corrected and tests made reproducible so versions could be compared using the same examples.
The engine is built and integration has been prepared. Pronunciation adjustments and testing within the applications remain to be completed before launch.
