RickAI
Back to all resources
ToolsTool-agnostic

Dessa and Moshi: The Most Human AI Voices

The links from my VOICE Reel: join the waitlist for Block's Dessa, and try Moshi, the free open-source voice AI you can run on your own computer today.

Your links

What Dessa is

Dessa is the AI voice from Block, Jack Dorsey's company. Most voice AI turns what you say into text, writes an answer, then reads it back, which is why it sounds flat.

Dessa skips the text. It was trained on recordings of people talking, so it learns to speak by listening, the way a baby does. It laughs, reacts and gets excited with you.

What Moshi is

Moshi is a free, open-source voice AI from Kyutai. It listens and talks at the same time, so a conversation flows like a real one and you can interrupt it.

It isn't as human as Dessa yet, but it's the closest thing you can run yourself today.

How to run Moshi yourself

  1. Try the browser demo first, so you know what it sounds like.
  2. On a Mac with Apple Silicon, install the MLX version and start it:
pip install -U moshi_mlx
python -m moshi_mlx.local_web
  1. On a PC, the PyTorch version needs an NVIDIA GPU with about 24 GB of memory. Follow the setup steps in the Moshi README.

RickAI newsletter

Useful AI, twice a week.

Practical prompts, products, and workflows for building faster, finding customers, and turning AI skills into income.

By joining, you agree to receive Rick's AI notes every Wednesday and Sunday. Unsubscribe anytime. Your data is never sold. Privacy policy