Речевые Технологии #14 Speech-to-Speech LLMs
Автор: Georgy Gospodinov
Загружено: 2025-12-19
Просмотров: 51
Speech-to-Speech LLM assistants that listen and respond in voice, emphasizing low latency, turn-taking, and natural interruption handling
Architectures: cascaded pipelines (VAD→ASR→LLM→TTS) vs end-to-end approaches (chain-of-modality, parallel generation, Thinker–Talker, full-duplex)
Data & evaluation: synthetic data generation and multi-metric evaluation for S2S dialogue quality
Доступные форматы для скачивания:
Скачать видео mp4
-
Информация по загрузке: