# SLICE: Speech Enhancement via Layer-wise Injection of Conditioning Embeddings

arXiv: https://arxiv.org/abs/2603.05302 · HTML: https://mlsanity.com/paper/2603.05302/

## Key facts

- No citing papers in the map yet (data through 19 Oct 2026).
- Placed in Speech enhancement and noise suppression, one of 11 directions in Speech and audio language models.
- Builds on 23 of its 32 references that are in the map.
- First posted on arXiv 5 Mar 2026.

## Placed in

- Speech and audio language models / [Speech enhancement and noise suppression](https://mlsanity.com/speech/speech-enhancement-and-noise-suppression/)

## Builds on (most cited first, 23 of 23)

- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing](https://mlsanity.com/paper/2110.13900/) (2021)
- [SDR - half-baked or well done?](https://mlsanity.com/paper/1811.02508/) (2018)
- [UTMOS: UTokyo-SaruLab System for VoiceMOS Challenge 2022](https://mlsanity.com/paper/2204.02152/) (2022)
- [BEATs: Audio Pre-Training with Acoustic Tokenizers](https://mlsanity.com/paper/2212.09058/) (2022)
- [Speech Enhancement and Dereverberation with Diffusion-based Generative Models](https://mlsanity.com/paper/2208.05830/) (2022)
- [Conditional Diffusion Probabilistic Model for Speech Enhancement](https://mlsanity.com/paper/2202.05256/) (2022)
- [Pyroomacoustics: A Python package for audio room simulations and array processing algorithms](https://mlsanity.com/paper/1710.04196/) (2017)
- [MetricGAN+: An Improved Version of MetricGAN for Speech Enhancement](https://mlsanity.com/paper/2104.03538/) (2021)
- [MP-SENet: A Speech Enhancement Model with Parallel Denoising of Magnitude and Phase Spectra](https://mlsanity.com/paper/2305.13686/) (2023)
- [HiFi-GAN: High-Fidelity Denoising and Dereverberation Based on Speech Deep Features in Adversarial Networks](https://mlsanity.com/paper/2006.05694/) (2020)
- [VoiceFixer: A Unified Framework for High-Fidelity Speech Restoration](https://mlsanity.com/paper/2204.05841/) (2022)
- [Voices Obscured in Complex Environmental Settings (VOICES) corpus](https://mlsanity.com/paper/1804.05053/) (2018)
- [Interspeech 2025 URGENT Speech Enhancement Challenge](https://mlsanity.com/paper/2505.23212/) (2025)
- [Noise-aware Speech Enhancement using Diffusion Probabilistic Model](https://mlsanity.com/paper/2307.08029/) (2023)
- [Universal Speech Enhancement with Regression and Generative Mamba](https://mlsanity.com/paper/2505.21198/) (2025)
- [Lessons Learned from the URGENT 2024 Speech Enhancement Challenge](https://mlsanity.com/paper/2506.01611/) (2025)
- [Few-step Adversarial Schr"odinger Bridge for Generative Speech Enhancement](https://mlsanity.com/paper/2506.01460/) (2025)
- [Diffusion Buffer: Online Diffusion-based Speech Enhancement with Sub-Second Latency](https://mlsanity.com/paper/2506.02908/) (2025)
- [TS-URGENet: A Three-stage Universal Robust and Generalizable Speech Enhancement Network](https://mlsanity.com/paper/2505.18533/) (2025)
- [Non-intrusive Speech Quality Assessment with Diffusion Models Trained on Clean Speech](https://mlsanity.com/paper/2410.17834/) (2024)
- [NADiffuSE: Noise-aware Diffusion-based Model for Speech Enhancement](https://mlsanity.com/paper/2309.01212/) (2023)

## Abstract (arXiv)

Real-world speech is often corrupted by multiple degradations simultaneously, including additive noise, reverberation, and nonlinear distortion. Diffusion-based enhancement methods perform well on single degradations but struggle with compound corruptions. Prior noise-aware approaches inject conditioning at the input layer only, which can degrade performance below that of an unconditioned model. To address this, we propose injecting degradation conditioning, derived from a pretrained encoder with multi-task heads for noise type, reverberation, and distortion, into the timestep embedding so that it propagates through all residual blocks without architectural changes. In controlled experiments where only the injection method varies, input-level conditioning performs worse than no encoder at all on compound degradations, while layer-wise injection achieves the best results. The method also generalizes to diverse real-world recordings.