@inproceedings{shikhar-etal-2025-llmvox,
title = "{LLMV}o{X}: Autoregressive Streaming Text-to-Speech Model for Any {LLM}",
author = "Shikhar, Sambal and
Kurpath, Mohammed Irfan and
Mullappilly, Sahal Shaji and
Lahoud, Jean and
Khan, Fahad Shahbaz and
Anwer, Rao Muhammad and
Khan, Salman and
Cholakkal, Hisham",
editor = "Che, Wanxiang and
Nabende, Joyce and
Shutova, Ekaterina and
Pilehvar, Mohammad Taher",
booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
month = jul,
year = "2025",
address = "Vienna, Austria",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.findings-acl.1051/",
doi = "10.18653/v1/2025.findings-acl.1051",
pages = "20481--20493",
ISBN = "979-8-89176-256-5",
abstract = "Recent advancements in speech-to-speech dialogue systems leverage LLMs for multimodal interactions, yet they remain hindered by fine-tuning requirements, high computational overhead, and text-speech misalignment. Existing speech-enabled LLMs often degrade conversational quality by modifying the LLM, thereby compromising its linguistic capabilities. In contrast, we propose LLMVoX, a lightweight 30M-parameter, LLM-agnostic, autoregressive streaming TTS system that generates high-quality speech with low latency, while fully preserving the capabilities of the base LLM. Our approach achieves a significantly lower Word Error Rate compared to speech-enabled LLMs, while operating at comparable latency. By decoupling speech synthesis from LLM processing via a multi-queue token streaming system, LLMVoX enables seamless, infinite-length dialogues. Its plug-and-play design also facilitates extension to various tasks with different backbones. Furthermore, LLMVoX generalizes to new languages with minimal dataset adaptation, attaining a low Character Error Rate on an Arabic speech task. Evaluations demonstrate that LLMVoX matches or surpasses existing speech-enabled LLMs in both speech quality and latency, while maintaining the original linguistic strengths of the LLM. Additionally, we have integrated LLMVoX with a Vision-Language Model to create an omni-model with speech, text, and vision capabilities, without requiring additional multimodal training."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="shikhar-etal-2025-llmvox">
<titleInfo>
<title>LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM</title>
</titleInfo>
<name type="personal">
<namePart type="given">Sambal</namePart>
<namePart type="family">Shikhar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammed</namePart>
<namePart type="given">Irfan</namePart>
<namePart type="family">Kurpath</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sahal</namePart>
<namePart type="given">Shaji</namePart>
<namePart type="family">Mullappilly</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jean</namePart>
<namePart type="family">Lahoud</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Fahad</namePart>
<namePart type="given">Shahbaz</namePart>
<namePart type="family">Khan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rao</namePart>
<namePart type="given">Muhammad</namePart>
<namePart type="family">Anwer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Salman</namePart>
<namePart type="family">Khan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hisham</namePart>
<namePart type="family">Cholakkal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2025</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wanxiang</namePart>
<namePart type="family">Che</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Joyce</namePart>
<namePart type="family">Nabende</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Shutova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammad</namePart>
<namePart type="given">Taher</namePart>
<namePart type="family">Pilehvar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-256-5</identifier>
</relatedItem>
<abstract>Recent advancements in speech-to-speech dialogue systems leverage LLMs for multimodal interactions, yet they remain hindered by fine-tuning requirements, high computational overhead, and text-speech misalignment. Existing speech-enabled LLMs often degrade conversational quality by modifying the LLM, thereby compromising its linguistic capabilities. In contrast, we propose LLMVoX, a lightweight 30M-parameter, LLM-agnostic, autoregressive streaming TTS system that generates high-quality speech with low latency, while fully preserving the capabilities of the base LLM. Our approach achieves a significantly lower Word Error Rate compared to speech-enabled LLMs, while operating at comparable latency. By decoupling speech synthesis from LLM processing via a multi-queue token streaming system, LLMVoX enables seamless, infinite-length dialogues. Its plug-and-play design also facilitates extension to various tasks with different backbones. Furthermore, LLMVoX generalizes to new languages with minimal dataset adaptation, attaining a low Character Error Rate on an Arabic speech task. Evaluations demonstrate that LLMVoX matches or surpasses existing speech-enabled LLMs in both speech quality and latency, while maintaining the original linguistic strengths of the LLM. Additionally, we have integrated LLMVoX with a Vision-Language Model to create an omni-model with speech, text, and vision capabilities, without requiring additional multimodal training.</abstract>
<identifier type="citekey">shikhar-etal-2025-llmvox</identifier>
<identifier type="doi">10.18653/v1/2025.findings-acl.1051</identifier>
<location>
<url>https://aclanthology.org/2025.findings-acl.1051/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>20481</start>
<end>20493</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM
%A Shikhar, Sambal
%A Kurpath, Mohammed Irfan
%A Mullappilly, Sahal Shaji
%A Lahoud, Jean
%A Khan, Fahad Shahbaz
%A Anwer, Rao Muhammad
%A Khan, Salman
%A Cholakkal, Hisham
%Y Che, Wanxiang
%Y Nabende, Joyce
%Y Shutova, Ekaterina
%Y Pilehvar, Mohammad Taher
%S Findings of the Association for Computational Linguistics: ACL 2025
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria
%@ 979-8-89176-256-5
%F shikhar-etal-2025-llmvox
%X Recent advancements in speech-to-speech dialogue systems leverage LLMs for multimodal interactions, yet they remain hindered by fine-tuning requirements, high computational overhead, and text-speech misalignment. Existing speech-enabled LLMs often degrade conversational quality by modifying the LLM, thereby compromising its linguistic capabilities. In contrast, we propose LLMVoX, a lightweight 30M-parameter, LLM-agnostic, autoregressive streaming TTS system that generates high-quality speech with low latency, while fully preserving the capabilities of the base LLM. Our approach achieves a significantly lower Word Error Rate compared to speech-enabled LLMs, while operating at comparable latency. By decoupling speech synthesis from LLM processing via a multi-queue token streaming system, LLMVoX enables seamless, infinite-length dialogues. Its plug-and-play design also facilitates extension to various tasks with different backbones. Furthermore, LLMVoX generalizes to new languages with minimal dataset adaptation, attaining a low Character Error Rate on an Arabic speech task. Evaluations demonstrate that LLMVoX matches or surpasses existing speech-enabled LLMs in both speech quality and latency, while maintaining the original linguistic strengths of the LLM. Additionally, we have integrated LLMVoX with a Vision-Language Model to create an omni-model with speech, text, and vision capabilities, without requiring additional multimodal training.
%R 10.18653/v1/2025.findings-acl.1051
%U https://aclanthology.org/2025.findings-acl.1051/
%U https://doi.org/10.18653/v1/2025.findings-acl.1051
%P 20481-20493
Markdown (Informal)
[LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM](https://aclanthology.org/2025.findings-acl.1051/) (Shikhar et al., Findings 2025)
ACL
- Sambal Shikhar, Mohammed Irfan Kurpath, Sahal Shaji Mullappilly, Jean Lahoud, Fahad Shahbaz Khan, Rao Muhammad Anwer, Salman Khan, and Hisham Cholakkal. 2025. LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM. In Findings of the Association for Computational Linguistics: ACL 2025, pages 20481–20493, Vienna, Austria. Association for Computational Linguistics.