We've gone live with a model update and some new tools for tailoring your speech!
This update includes native coverage of natural sounding number sequences, and word spelling, and pausing too!
Model Update
Building on our popular Mist for IVR model, we've refreshed and expanded our offerings with vastly improved list intonation as well as new and varied voices!
Below are some example new voices:
Custom Pauses
Rime's TTS is unique among next-gen TTS offerings in allowing users to customize their output. Out-of-the-box, Rime's voices will speak fluently and correctly, but since strings of words can be pronounced differents for special effect, we allow users to add custom pauses and custom pronunciations. Let's see how.
To insert a pause within a sentence, simply inset the length of your desired pause in milliseconds inside angle brackets, for example: <750>.
To hear the difference, compare the following:
Here's a sample API request for this:
{
"text": "wait, <750> are you actually serious.",
"speaker": "cove",
"modelId": "mist",
"pauseBetweenBrackets": true
}
Custom Pronunciation
We can add words to our dictionary with in about 24 hours, but if you want to stipulate a pronunciation yourself, you can input a specific pronunciation within curly brackets. For example here's a made up word 'gorbulets', which can spoken by inputting {g1orby0ul2Ets}.
Take a listen:
Here's a sample API request for this:
{
"text": "actually, {g1orby0ul2Ets} is a word i just made up.",
"speaker": "peak",
"modelId": "mist",
"phonemizeBetweenBrackets": true
}
For more details, see our documentation!