UniSE: Unified Speech Enhancement

Upload noisy or mixed speech audio and get enhanced output. Supports Speech Enhancement (denoising), Target Speaker Extraction (extract a target speaker using reference audio), and Speech Separation (separate mixed speakers into individual streams).

📄 Paper | 🤗 Model | 💻 Code

Task Mode

Speech Enhancement (SE) Target Speaker Extraction (TSE) Speech Separation (SS)

Input Audio (noisy/mixed)

Reference Enrollment Audio (required for TSE, ignored for SE/SS)

Enhanced Output

Examples