After ~10 messages, every AI voice sounds the same

If you’ve tried building anything with AI voices + characters, you’ve probably seen this:

They start strong… then slowly become the same person.

Different “voices”, same personality underneath.

That felt like the wrong abstraction.

So instead of picking a voice, we tried defining identity first:

– how the character thinks

– how they respond

– how they speak over time

Voice just renders that.

We built a small API around this idea:

one persona → consistent voice behavior across turns

Not sure if this is the right direction yet, but it’s been interesting to test.

Is anyone else here working on character-driven apps?

4 views