---
format: "aidr-story-markdown/v1"
id: "014e881b43d71b51234e4f3875e4bf035fc3ed610975f581c356801dded7f2c6"
canonical_url: "https://aidr.today/014e881b?lang=en"
title: "Efficient Streaming Audio-Visual Target Speaker Extraction for Real-World Acoustic Scenes"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-29T00:00:00.000Z"
category: "Research"
topics: ["inference"]
source_urls: ["https://huggingface.co/papers/2610.32749"]
summary: "Researchers introduced RealSSA, a benchmark for realistic audio-visual target speaker extraction, and Falcon, a causal method using multi-scale time-frequency separation trained with an encoder-space speaker contrastive loss."
---

# Efficient Streaming Audio\-Visual Target Speaker Extraction for Real\-World Acoustic Scenes

> [Open the canonical story](<https://aidr.today/014e881b?lang=en>)

**Published:** 2026-10-29T00:00:00.000Z
**Category:** Research
**Topics:** inference

## Summary

Researchers introduced RealSSA, a benchmark for realistic audio\-visual target speaker extraction, and Falcon, a causal method using multi\-scale time\-frequency separation trained with an encoder\-space speaker contrastive loss\.

## Sources

- [Story source](<https://huggingface.co/papers/2610.32749>)