KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kuroki, So, Kubo, Yotaro, Akiba, Takuya, Tang, Yujin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911714865315840
author Kuroki, So
Kubo, Yotaro
Akiba, Takuya
Tang, Yujin
author_facet Kuroki, So
Kubo, Yotaro
Akiba, Takuya
Tang, Yujin
contents Real-time speech-to-speech (S2S) models excel at generating natural, low-latency conversational responses but often lack deep knowledge and semantic understanding. Conversely, cascaded systems combining automatic speech recognition, a text-based Large Language Model (LLM), and text-to-speech synthesis offer superior knowledge representation at the cost of high latency, which disrupts the flow of natural interaction. This paper introduces a novel hybrid architecture that bridges the gap between these two paradigms. Our framework processes user speech through an S2S transformer for immediate responsiveness while concurrently relaying the query to a powerful back-end LLM. The LLM's text-based response is then injected in real time to guide the S2S model's speech generation, effectively infusing its output with rich knowledge without the full latency penalty of a cascaded system. We evaluated our method using a speech-synthesized variant of the MT-Bench benchmark that consists of multi-turn question-answering sessions. The results demonstrate that our system substantially outperforms a baseline S2S model in response correctness, approaching that of a cascaded system, while maintaining a latency on par with the baseline.
format Preprint
id arxiv_https___arxiv_org_abs_2510_02327
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI
Kuroki, So
Kubo, Yotaro
Akiba, Takuya
Tang, Yujin
Computation and Language
Artificial Intelligence
Audio and Speech Processing
Real-time speech-to-speech (S2S) models excel at generating natural, low-latency conversational responses but often lack deep knowledge and semantic understanding. Conversely, cascaded systems combining automatic speech recognition, a text-based Large Language Model (LLM), and text-to-speech synthesis offer superior knowledge representation at the cost of high latency, which disrupts the flow of natural interaction. This paper introduces a novel hybrid architecture that bridges the gap between these two paradigms. Our framework processes user speech through an S2S transformer for immediate responsiveness while concurrently relaying the query to a powerful back-end LLM. The LLM's text-based response is then injected in real time to guide the S2S model's speech generation, effectively infusing its output with rich knowledge without the full latency penalty of a cascaded system. We evaluated our method using a speech-synthesized variant of the MT-Bench benchmark that consists of multi-turn question-answering sessions. The results demonstrate that our system substantially outperforms a baseline S2S model in response correctness, approaching that of a cascaded system, while maintaining a latency on par with the baseline.
title KAME: Tandem Architecture for Enhancing Knowledge in Real-Time Speech-to-Speech Conversational AI
topic Computation and Language
Artificial Intelligence
Audio and Speech Processing
url https://arxiv.org/abs/2510.02327