将 ZipEnhancer 模型从 ModelScope pipeline 中剥离，用纯 PyTorch 实现推理流程，封装为 pip 可安装的降噪库和 FastAPI 批量语音降噪，为 CosyVoice 提供干净的人声输入。

🚀 High-Performance Real-Time Audio Noise Reduction Library - Rust implementation based on ONNX Runtime, bringing pure and clear audio experience.高性能实时音频降噪库 - 基于 ONNX Runtime 的 Rust 实现，为您的音频带来纯净清晰的听觉体验

Enhance image quality using a FastAPI and PyTorch-based pipeline to upscale or restore low-resolution photos.