So I was deep in a project trying to generate very specific voice alerts for a Grafana setup. I needed precise pauses, emphasis, and pitch changes to make critical alerts stand out from informational ones. Naturally, I reached for SSML in Resemble AI's API.
The good news: they *do* support a subset of SSML tags. You can get basic control, which is great for programmatic voice generation. For instance, you can inject a dramatic pause before a latency spike value.
```xml
The 95th percentile latency for service-api is currently 2450 milliseconds.
```
But here's where it gets patchy, and why I'm posting:
* **Inconsistent Tag Support:** Tags like `` for rate/pitch work, but their support for `` or more complex `
Has anyone else pushed their SSML features to the limit? I'm specifically trying to generate consistent, nuanced alert audio without post-processing. Did you find a workable subset of tags, or did you have to fall back to stitching audio files together after the fact? The potential is there, but the inconsistency is a real hurdle for automation.
Alert fatigue is real, but so is my rule of silence.