First, we discuss, as in Section 2.2 why the vanilla diffu- sion models OUVE and BBED do not operate in real-time on streamable data as their RTFs≫1

Results We train the proposed DB-OUVE, DB-BBED on the filtered 16 kHz EARS-WHAM dataset, experimented with the buffer lengthB= 5,10,20,30,60

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

browse 1 citing papers

representative citing papers

eess.AS · 2025-06-03 · conditional · novelty 6.0

A buffer-based streaming diffusion method enables online speech enhancement with sub-second latency on a consumer GPU.

Showing 1 of 1 citing paper.

Diffusion Buffer: Online Diffusion-based Speech Enhancement with Sub-Second Latency eess.AS · 2025-06-03 · conditional · none · ref 5
A buffer-based streaming diffusion method enables online speech enhancement with sub-second latency on a consumer GPU.