AI Voice Best Practices? · Assistable.ai

AI Voice Best Practices?

Hey everyone, I have a question about voice AI best practices.

I’ve been experimenting a lot with voice AI, but I feel like some of you have absolutely mastered it. Your call recordings sound insanely human and incredibly fast. Meanwhile, when I try to replicate the same thing, I can never seem to reach that level.

So I wanted to ask the community:

What exact best practices are you using to make your AI calls sound both extremely fast and extremely human?

Specifically, I’d love to hear about:

Which AI model you’re using — GPT-4.0, GPT-5, or something else
Whether voice quality mostly comes from the voice clone, the model, or the prompting
Any prompting structures, tricks, or settings that noticeably improve speed or realism
Any small details you’ve found that make a huge difference in conversational flow

There isn’t any real documentation that explains what the “gold standard” setup looks like, and I feel like we could all benefit from sharing what has actually worked in the real world. I’ve tried talking to a few people one-on-one, but it feels like we don’t yet have a centralized place to compare notes.

If you’ve dialed in a setup that works incredibly well, I’d really appreciate anything you’re willing to share.

Thanks in advance! I think we can all learn a lot from each other on this.

2 comments