🎭 PersonaPlex 7B Demo

Real-time speech-to-speech with nvidia/personaplex-7b-v1

⚠️ Demo Limitation

This static demo uses the fal.ai Inference Provider via Hugging Face's API. The PersonaPlex model requires complex streaming inference (Moshi library) that isn't fully supported by the standard Inference API.

For the full experience with persona/voice control and streaming, run locally or use the Gradio Space (requires HF PRO for ZeroGPU).

Persona

Voice

Temperature: 0.7

Max Duration (s): 10

🎤 Your Input

Click "Start Recording" to speak

🔊 PersonaPlex Response

Response will appear here

About PersonaPlex

Model: NVIDIA's PersonaPlex 7B - a full-duplex speech-to-speech conversational model
Architecture: Based on Moshi (Mimi codec + Transformer)
Capabilities: Real-time voice conversation with persona and voice control
Input: 24kHz audio + text persona prompt + voice embedding
Output: 24kHz streaming audio response

Note: This demo uses the fal.ai Inference Provider via HF API. Accept the license if prompted.