Omnilert's Voice Calling endpoint includes the ability to convert text-to-speech, changing your written content into an audio message in voice calls.
Generally, the system can and will be able to properly convert any full sentences in English.
So, we advise writing out messages fully if using text-to-speech.
The following can have unpredictable results when processed by a text-to-speech engine:
- Proper nouns: Most common names of places and people should sound OK, but there are no guarantees. Always test out proper nouns with a small-scale test message.
- Non-English language script. The text-to-speech engine is English language only. It will attempt to pronounce any text as if it were English.
- Web addresses: URLs tend to sound very odd when pronounced by a computer.
- Phone numbers: Phone numbers can often be interpreted as "math" by the computer. So your number sounds like an equation!
- Acronyms and abbreviations: The system will attempt to pronounce any word it sees, but always test out acronyms, abbreviations, and initialisms.
- Emoji / "text speak" and technical jargon: Generally speaking, the text-to-speech engine can't handle non-dictionary words.
- Dates and time in shorthand: Dates and times should be written out phonetically.
Let's take a simple "Campus opens tomorrow" message, but write it three ways to be converted to speech:
- Bad: "Campus open tmw, 2/7 @ 10:00am."
- Better: "Campus will open tomorrow, February 7th at 10 A.M."
- Best: "Campus will open tomorrow, Wednesday, February Seventh, at Ten o'clock."
In some cases, using text-to-speech may be at-odds with SMS or other endpoints that would use shorthand, abbreviations, URLs, etc. Different media may just have different needs.
In such cases, you may want to either send the voice message separately (using a more "text-to-speech friendly" message) or use Omnilert's voice call recording capability to pre-record an outgoing voice message.