How does your "Crisper Speech Model" compared to parakeet-0.8b or similar? I've had great experience with that model via handy.computer.