personaplex

# PersonaPlex: Voice and Role Control for Full Duplex Conversational Speech Models

[![Weights](https://img.shields.io/badge/🤗-Weights-yellow)](https://huggingface.co/nvidia/personaplex-7b-v1)
[![Paper](https://img.shields.io/badge/📄-Paper-blue)](https://arxiv.org/abs/2602.06053)
[![Demo](https://img.shields.io/badge/🎮-Demo-green)](https://research.nvidia.com/labs/adlr/personaplex/)
[![Discord](https://img.shields.io/badge/Discord-Join-purple?logo=discord)](https://discord.gg/5jAXrrbwRb)

PersonaPlex is a real-time, full-duplex speech-to-speech conversational model that enables persona control through text-based role prompts and audio-based voice conditioning. Trained on a combination of synthetic and real conversations, it produces natural, low-latency spoken interactions with a consistent persona. PersonaPlex is based on the [Moshi](https://arxiv.org/abs/2410.00037) architecture and weights.

**NEW:** the system prompt (voice + text) now stays in the model's context for the whole conversation instead of dropping out of context. The prompt also acts as the attention sink. This enabled endless stable conversations while preserving voice and role. However, conversation memory is still limited to a sliding window of the last ~2.5 minutes ((2048-system_prompt_frames)/12.5). Also the system prompt is loaded in one prefill instead of stepping though token-by-token. Both are on by default; see [Endless Conversations](#endless-conversations).

<p align="center">
  <img src="assets/architecture_diagram.png" alt="PersonaPlex Model Architecture">
  <br>
  <em>PersonaPlex Architecture</em>
</p>

## Usage

### Prerequisites

Install the [Opus audio codec](https://github.com/xiph/opus) development library:
```bash
# Ubuntu/Debian
sudo apt install libopus-dev

# Fedora/RHEL
sudo dnf install opus-devel
```

### Installation

Download this repository and install with:
```bash
pip install moshi/.
```

Extra step for Blackwell based GPUs as suggested in (See https://github.com/NVIDIA/personaplex/issues/2):
```bash
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu130
```


### Accept Model License
Log in to your Huggingface account and accept the PersonaPlex model license [here](https://huggingface.co/nvidia/personaplex-7b-v1). <br>
Then set up your Huggingface authentication:
```bash
export HF_TOKEN=<YOUR_HUGGINGFACE_TOKEN>
```

### Launch Server

Launch server for live interaction (temporary SSL certs for https):
```bash
SSL_DIR=$(mktemp -d); python -m moshi.server --ssl "$SSL_DIR"
```

**CPU Offload:** If your GPU has insufficient memory, use the `--cpu-offload` flag to offload model layers to CPU. This requires the `accelerate` package (`pip install accelerate`):
```bash
SSL_DIR=$(mktemp -d); python -m moshi.server --ssl "$SSL_DIR" --cpu-offload
```

Access the Web UI from a browser at `localhost:8998` if running locally, otherwise look for the access link printed by the script:
```
Access the Web UI directly at https://11