jeaneudesAyilo/enudiffuse

This repository provides the checkpoints used in the paper:

Diffusion-based Frameworks for Unsupervised Speech Enhancement.

Please, read also the main github repository for a complete understanding.

About the checkpoints:

• separate_wsjqut_speech_modeling.ckpt: Audio-only diffusion model trained separately on WSJ0 speech dataset (~5.2 M parameters).

• separate_wsjqut_noise_modeling.ckpt: Audio-only diffusion model trained separately on QUT noise dataset (~5.2 M parameters).

• joint_wsjqut_speech_noise_modeling.ckpt: Audio-only diffusion model trained jointly on WSJ0 speech and QUT noise datasets (~5.9 M parameters).

• separate_vbdmd_speech_modeling.ckpt: Audio-only diffusion model trained separately on VoiceBank speech dataset (~5.2 M parameters).

• separate_vbdmd_noise_modeling.ckpt: Audio-only diffusion model trained separately on DEMAND noise dataset (~5.2 M parameters).

• joint_vbdmd_speech_noise_modeling.ckpt: Audio-only diffusion model trained jointly on VoiceBank speech and QUT noise datasets (~5.9 M parameters)