mu-hashmi

mu-hashmi / personaplex-mlx

Public

PersonaPlex on Apple Silicon: an MLX port of NVIDIA’s full-duplex speech-to-speech model with realtime local/web modes and offline WAV inference.

apple-silicon full-duplex huggingface mlx moshi
29
3
100% credibility
Found Feb 19, 2026 at 21 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

A user-friendly package to run NVIDIA PersonaPlex voice AI conversations locally on Apple Silicon Macs in real-time web, terminal, or offline audio modes.

How It Works

1
📰 Discover PersonaPlex

You hear about a cool voice AI tool that lets you have natural conversations right on your Apple Mac.

2
🔑 Unlock the voices

You visit the model page, accept the license, and link your account to download the AI voices and brains.

3
💻 Set up on your Mac

You download and prepare the tool, which is made just for newer Apple computers.

4
🌐 Launch the web chat

You start the friendly web version with one simple command, and it opens in your browser ready to talk.

5
🗣️ Pick voice and style

You choose a natural-sounding voice like NATF2 and give it a personality prompt, such as a wise teacher.

6
💬 Start talking live

You speak into your microphone, and the AI listens, thinks, and speaks back in real-time – grab headphones to avoid echoes!

7
Try other ways
⌨️
Terminal mode

Chat right in the command window for quick text-to-voice fun.

🎵
Offline mode

Transform your audio clips into new voiced versions with custom prompts.

🎉 Voice AI magic at home

You now have a smooth, private voice companion running entirely on your Mac, chatting naturally anytime.

Sign up to see the full architecture

6 more

Sign Up Free

Star Growth

See how this repo grew from 21 to 29 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is personaplex-mlx?

PersonaPlex-MLX is a Python port of NVIDIA's PersonaPlex full-duplex speech-to-speech model optimized for Apple Silicon using MLX. It delivers realtime local terminal chats, web-based inference at localhost:8998, and offline WAV-to-WAV processing via simple CLI commands like `personaplex-local` or `personaplex-offline`. Developers get instant voice conversations with customizable voices and text prompts, all running locally without cloud dependencies.

Why is it gaining traction?

This MLX port unlocks NVIDIA-grade full-duplex inference—simultaneous listening and speaking—on Apple Silicon Macs, sidestepping GPU shortages. Realtime local/web modes plus offline WAV support make prototyping speech apps frictionless, with quantization (4/8-bit) for speed. The HF model integration and voice bundles handle downloads automatically, letting users focus on building.

Who should use this?

Speech AI researchers testing full-duplex models on Macs, indie devs prototyping voice assistants or chatbots, and Apple Silicon users needing local speech-to-speech inference for apps like virtual tutors. Ideal for anyone evaluating realtime voice UX without NVIDIA hardware.

Verdict

Worth trying for Apple devs eyeing efficient speech-to-speech on Silicon—installs cleanly via pip and runs out-of-box. But with 17 stars and 1.0% credibility, it's early-stage; expect basic docs and no echo cancellation, so pair headphones and watch for updates.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.