# Pyannote

Run **Pyannote** optimized for **Qualcomm SnapDragon device's NPU** with [nexaSDK](https://sdk.nexa.ai).

## Quickstart

1. **Install NexaSDK** and create a free account at [sdk.nexa.ai](https://sdk.nexa.ai)
2. **Activate your device** with your access token:

   ```bash
   nexa config set license '<access_token>'
   ```
3. Run the model on Qualcomm NPU in one line:

   ```bash
   nexa infer NexaAI/Pyannote-NPU
   ```
   
- Input: Enter input audio path,
- Output: Returns speech diarization results, or report error if any required input cannot be found


## Model Description
**pyannote-audio (Community Version)** is an open-source **speech diarization** model designed for accurate speaker segmentation and labeling in audio streams.  
Developed by the **Pyannote community**, it combines **audio processing**, **speaker embedding**, and **clustering** into a unified framework, enabling robust speech segmentation on local machines without cloud dependency.

## Features
- 🔊 **End-to-End Diarization Pipeline** — Automatically detects and labels who spoke when in an audio file.  
- ⚡ **Lightweight & Efficient** — Optimized for real-time or batch processing on consumer hardware and GPUs.  
- 🧠 **Speaker Embedding & Clustering** — Extracts rich speaker representations and groups them for identity separation.  
- 🔧 **Customizable & Modular** — Easily integrates with PyTorch pipelines or modified components for research and prototyping.  
- 🌍 **Community-Driven & Transparent** — Fully open and maintained by an active community of speech researchers and developers.

## Use Cases
- **Meeting Transcription**: Segment conversations by speaker for clearer transcripts.  
- **Broadcast and Podcast Analysis**: Attribute voices and structure long-form audio content.  
- **Call Center Analytics**: Separate agent and customer segments for interaction insights.  
- **Research**: Test diarization algorithms or contribute new speaker models.  
- **Voice Dataset Preparation**: Preprocess large audio datasets for training ASR or emotion recognition systems.

## Inputs and Outputs
**Input**
- Audio file or stream  

**Output**
- Speaker-labeled time segments  


## License
This repo is licensed under the **Creative Commons Attribution–NonCommercial 4.0 (CC BY-NC 4.0)** license, which allows use, sharing, and modification only for non-commercial purposes with proper attribution.  
All NPU-related models, runtimes, and code in this project are protected under this non-commercial license and cannot be used in any commercial or revenue-generating applications.  
Commercial licensing or enterprise usage requires a separate agreement.  
For inquiries, please contact `dev@nexa.ai`.