Just spent another 20 minutes trying to dial in a voice for a new project. I swear, WellSaid's "Voice Lab" feels like piloting a 1970s mixing board where all the labels have faded off.
Take the **"Stability"** slider. Moving it up... makes the voice more stable? Or does it make the *performance* more stable? The result sounds less emotional, so is it stabilizing the tone? But then what's "Expressiveness" for? And "Pacing" is a nightmare—faster pacing sometimes makes it sound *less* urgent because it messes with the cadence. I end up doing this weird dance:
* Adjust Stability up a notch.
* Crank Expressiveness down.
* Listen. It sounds dead.
* Reverse it. Now it sounds unhinged.
* Land on some utterly random middle ground and call it a day.
It's not just me, right? Coming from other platforms, I expect sliders like "Warmth," "Clarity," "Speed," "Pitch Variance"—things that map to an actual auditory result. Here, the parameters feel like they're named after internal engineering concepts, not user-facing controls.
The worst part is the inconsistency. A setting that works for "Ava" makes "Henry" sound like he's having a meltdown. So you're not just learning one set of interactions, you're relearning it for each voice. Fantastic use of my billable hours.
Would kill for a simple "Advanced" toggle that just gave you plain English descriptions. Or even better, some preset "profiles" (Conversational, Authoritative, Empathetic) that set a sane baseline to tweak from. Until then, it's guesswork disguised as customization.
been there, migrated that