| |
Recorded Prompt Sets
In our SVOX recording studio in Zurich, equipped with the latest technology, we have recorded more than twenty thousand prompts in 18 different languages. These recording sessions are always supervised by a sound engineer and at least one native language expert.
Our TTS voice talents are available to record prompt sets for customer projects, so that the voice stays the same when prompts and TTS inclusions are combined by the engine (using a technique called phrase splicing).
If an exclusively branded custom voice is created, the recordings of prompts can be recorded during the same session.
Recordings
We strive to make the recordings of your prompts using the same speakers who originally contributed the material for the unit selection corpus or your customized voices. We recommend this procedure because it guarantees the audio characteristics of recorded prompts and TTS voice will be identical.
Audio Processing
Then, before shipping a prompt set, we apply the same audio filters to the recorded prompts and the unit selection corpus to make the speech as consistent as possible. Likewise, if the target system is designed for a specific car line, we will conduct listening tests in the vehicle and adjust the final audio filtering to match the car's audio system.
Tokenizing
If memory consumption on the target platform is an issue – and it usually is – we can tokenize the recorded prompt set into smaller splices, which can be reused in different contexts. This method of tokenizing is helpful at substantially reducing the required memory needed to store recordings in navigation systems.
|
|