Загрузка видео...
Не удалось загрузить видео
Efficient Streaming Audio-Visual Target Speaker Extraction for Real-World Acoustic Scenes paper:
16,262 просмотров • 20 дней назад •via X (Twitter)
Комментарии: 2

Hao19 дней назад
For video calls I'd check lip sync first. Studio's README lists about 0.7-0.8s of designed delay. Meeting video isn't delayed with the audio. The paper's 51ms is GPU inference.

Marvin20 дней назад
RealSSA 把近场、远场和人数变化放进来,这个测试设置更贴近会议。还想看目标说话人转头或嘴部被遮挡时,提取效果会掉多少。


