Audio Enhancer (Standalone)

A modern, standalone desktop application to enhance audio quality using AI (AudioSR). Built with C# / Avalonia UI (.NET 9.0) and Python.

🚀 Features

• High-Quality AI Enhancement: Uses the audiosr library to upscale and restore audio.
• Portable Python: Runs completely offline with an embedded Python environment (no installation required).
• GPU Acceleration: Utilizes NVIDIA GPUs (CUDA 13.0) via PyTorch Nightly for faster processing (optimized for RTX 5000 series).
• Smart Preprocessing: Applies an 8000Hz Low-Pass Filter to prevent AI artifacts (spectrogram holes).
• Professional Normalization: Optional EBU R128 normalization to target broadcast standards (-14 LUFS).
• Chunk-Based Processing: Splits large files into 30s chunks to manage VRAM usage, and is configurable to prevent crashes and ensure responsive UI.
• Modern UI: Dark-themed, drag-and-drop interface built with Avalonia.
• Memory Management: Aggressive garbage collection preventing VRAM leaks on long files.

📦 Installation / Quick Start

1. Download the App Build the solution or use the pre-compiled executable.

2. Python Environment (Portable) The app runs offline. It tries to download the required 3GB Python environment automatically on first start. If you prefer to download it manually (e.g. for faster download managers):

• Download Link: pythonlib_cuda13.zip (Pixeldrain)
• Instructions: Extract the pythonlib folder next to AudioEnhancer.UI.exe.

🎵 Results Comparison

Original Spectrum	Enhanced Spectrum

�🛠 Prerequisites

To run this application, you need:

1. Windows OS (Linux/macOS pending verification).
2. .NET 9.0 Runtime (or SDK to build).
3. FFmpeg: Installed and accessible (or path configured in Settings).
4. Python & Conda Environment:
   • You need a Conda environment (default name: vasr-cuda13) with audiosr installed.
   • Installation command: pip install audiosr
   • Must support CUDA for decent performance.

⚙️ Configuration & Performance

The app allows tuning for your specific hardware via the Settings menu.

Processing Chunk Size (Seconds)

This setting controls how many seconds of audio are processed at once.

• 30.0s (Default): Optimized for High-End GPUs (e.g., RTX 5070 Ti, 4070, 3080). Offers the best speed/stability balance.
• 10.0s - 15.0s: Recommended for Mid-Range GPUs or Laptops (e.g., RTX 3060 Laptop).
• 5.0s: Safe Mode for GPUs with low VRAM (< 8GB).
• 60.0s: Experimental. Only for GPUs with massive VRAM (24GB+). May cause instability.

📦 Installation & Usage

1. Clone the repository:

   git clone https://github.com/widlers/Audio-Enhancer.git
   cd Audio-Enhancer

2. Build/Run:

   dotnet run --project AudioEnhancer.UI/AudioEnhancer.UI.csproj

3. First Run Setup:

   • Click on Settings (Gear Icon).
   • Ensure "FFmpeg Path" points to ffmpeg.exe.
   • Ensure "Python/Conda Path" points to your Conda python.exe (e.g., .../envs/vasr-cuda13/python.exe).
   • Set your "Conda Environment Name" (default: vasr-cuda13).

4. Enhance:

   • Drag & Drop an audio file (MP3, WAV, FLAC).
   • Click Start Enhancing.
   • Wait for the magic! ✨

🔧 Troubleshooting

• "GPU Memory Full" / Crash at X%:
  • Go to Settings and reduce the "Processing Chunk Size" (e.g., from 30.0 to 10.0).
• Python not found:
  • Make sure you selected the python.exe inside your Conda environment folder, NOT the base python.
• UI Freezes:
  • Should not happen anymore. If it does, check the logs for massive spam and report an issue.

📝 License

MIT