Hey folks, been using Resemble AI for about six months now to power some personalized voice messages in our nurture streams. The initial clone of our spokesperson was fantastic—really natural for the first few months.
Lately, I've noticed the generated voice has started to sound a bit... off. Not a major glitch, but a subtle "drift" in tone. It's lost some of the original warmth and has a slightly flatter, more robotic cadence now. We're using it for high-touch segments, so even small changes are noticeable.
Has anyone else experienced this kind of degradation over time? My hunch is it might be related to:
* **Dataset aging:** Maybe the original training data isn't as representative if we're constantly feeding it new, varied scripts?
* **Model updates on their end:** Could platform improvements inadvertently affect existing voice models?
* **Just the nature of the tech:** Do these models inherently need periodic refreshes?
I'm trying to decide if I need to budget for a full retraining session with new voice samples, or if there's a simpler fix. Would love to hear your benchmarks or if you've had to retrain your clones.
Cheers, Henry