Research map / Image generation / Autoregressive and latent generation
Separate and Diffuse: Using a Pretrained Diffusion Model for Better Source Separation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Uses a pretrained diffusion model to perform source separation on speech mixtures, arguing that generative models can exceed the deterministic upper bound on separation.
Problem
Speech separation (the cocktail party problem): isolating a single speech signal from a mixture; prior work derived an upper bound on separation performance that applies only to deterministic models.
Approach
Generative models may break the deterministic separation upper bound; the paper uses a pretrained diffusion model to separate sources, rather than a deterministic separator.
Surprising findings
- Generative models challenge the deterministic upper bound on speech separation performance.
Closest papers by meaning
- DiffAR: Denoising Diffusion Autoregressive Model for Raw Speech Waveform Generation (ICLR 2024 · similarity 0.88)
- ZeroSep: Separate Anything in Audio with Zero Training (NeurIPS 2025 · similarity 0.88)
- Cauchy Diffusion: A Heavy-tailed Denoising Diffusion Probabilistic Model for Speech Synthesis (AAAI 2025 · similarity 0.87)
- Predictive–Generative Drift Decomposition for Speech Enhancement and Separation (NeurIPS 2026 · similarity 0.87)
- ArrayDPS: Unsupervised Blind Speech Separation with a Diffusion Prior (ICML 2025 · similarity 0.87)
- Multi-Source Diffusion Models for Simultaneous Music Generation and Separation (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.