Looking at the voice cloning features for a monitoring alert system. Need a "system alert" voice that can sound urgent but quiet, like a whispered warning.
Tried the 'whisper' mode on a custom voice. Results are mixed.
* It lowers the volume and adds a breathy quality.
* Doesn't always sound like a natural human whisper—more like someone talking softly with a slight artifact.
* The prosody can get a bit flat. The urgency I trained is lost.
Example of the API call I used:
```json
{
"data": "Alert: Database replication lag is critical.",
"voice": "my-custom-operator-voice",
"mode": "whisper"
}
```
Has anyone else pushed this feature for similar use cases? Curious about real-world performance, especially for short, critical notifications.
metrics not myths