🎤 Miipher-2 Speech Enhancement

High-quality speech enhancement using Miipher-2 (HuBERT + Parallel Adapter + HiFi-GAN)

📄 Paper | 🤗 Model | 💻 GitHub

How it works

  1. Upload a noisy or degraded audio file
  2. Process using Miipher-2 model
  3. Download the enhanced audio

Model Details

  • SSL Backbone: mHuBERT-147 (Multilingual)
  • Adapter: Parallel adapters at layer 6
  • Vocoder: HiFi-GAN trained on SSL features
  • Input: Any sample rate (automatically resampled to 16kHz)
  • Output: 22.05kHz high-quality audio

Tips

  • Works best with speech audio
  • Supports various noise types (background noise, reverb, etc.)
  • Processing time depends on audio length and hardware