@inproceedings{wu-etal-2025-towards,
title = "Towards Codec-{LM} Co-design for Neural Codec Language Models",
author = "Wu, Shih-Lun and
Lahoti, Aakash and
Desai, Arjun D and
Goel, Karan and
Donahue, Chris and
Gu, Albert",
editor = "Ebrahimi, Abteen and
Haider, Samar and
Liu, Emmy and
Haider, Sammar and
Leonor Pacheco, Maria and
Wein, Shira",
booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop)",
month = apr,
year = "2025",
address = "Albuquerque, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.naacl-srw.6/",
doi = "10.18653/v1/2025.naacl-srw.6",
pages = "55--65",
ISBN = "979-8-89176-192-6",
abstract = "Neural codec language models (or codec LMs) are emerging as a powerful framework for audio generation tasks like text-to-speech (TTS). These models leverage advancements in language modeling and residual vector quantization (RVQ)-based audio codecs, which compress audios into discrete codes for LMs to process. Despite the close interdependence of codecs and LMs in these systems, research on codecs and LMs has largely remained siloed. In this work, we propose three techniques for better codec-LM co-design: (i) a frame-wise codec encoder that improves both LM log-likelihood and end-to-end TTS metrics, (ii) LM codebook level dropout, a method to efficiently navigate a portion of the codec-LM design space by training a single LM, and (iii) increased codec frame duration, which we show can accelerate inference while maintaining end-to-end performance. Our experiments demonstrate that combining all three co-design techniques results in doubled inference speed, and improvements in intelligibility, audio quality, and speaker control in TTS relative to a siloed baseline."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wu-etal-2025-towards">
<titleInfo>
<title>Towards Codec-LM Co-design for Neural Codec Language Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shih-Lun</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aakash</namePart>
<namePart type="family">Lahoti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arjun</namePart>
<namePart type="given">D</namePart>
<namePart type="family">Desai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Karan</namePart>
<namePart type="family">Goel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chris</namePart>
<namePart type="family">Donahue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Albert</namePart>
<namePart type="family">Gu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-04</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Abteen</namePart>
<namePart type="family">Ebrahimi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Samar</namePart>
<namePart type="family">Haider</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Emmy</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sammar</namePart>
<namePart type="family">Haider</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Leonor Pacheco</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shira</namePart>
<namePart type="family">Wein</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-192-6</identifier>
</relatedItem>
<abstract>Neural codec language models (or codec LMs) are emerging as a powerful framework for audio generation tasks like text-to-speech (TTS). These models leverage advancements in language modeling and residual vector quantization (RVQ)-based audio codecs, which compress audios into discrete codes for LMs to process. Despite the close interdependence of codecs and LMs in these systems, research on codecs and LMs has largely remained siloed. In this work, we propose three techniques for better codec-LM co-design: (i) a frame-wise codec encoder that improves both LM log-likelihood and end-to-end TTS metrics, (ii) LM codebook level dropout, a method to efficiently navigate a portion of the codec-LM design space by training a single LM, and (iii) increased codec frame duration, which we show can accelerate inference while maintaining end-to-end performance. Our experiments demonstrate that combining all three co-design techniques results in doubled inference speed, and improvements in intelligibility, audio quality, and speaker control in TTS relative to a siloed baseline.</abstract>
<identifier type="citekey">wu-etal-2025-towards</identifier>
<identifier type="doi">10.18653/v1/2025.naacl-srw.6</identifier>
<location>
<url>https://aclanthology.org/2025.naacl-srw.6/</url>
</location>
<part>
<date>2025-04</date>
<extent unit="page">
<start>55</start>
<end>65</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Towards Codec-LM Co-design for Neural Codec Language Models
%A Wu, Shih-Lun
%A Lahoti, Aakash
%A Desai, Arjun D.
%A Goel, Karan
%A Donahue, Chris
%A Gu, Albert
%Y Ebrahimi, Abteen
%Y Haider, Samar
%Y Liu, Emmy
%Y Haider, Sammar
%Y Leonor Pacheco, Maria
%Y Wein, Shira
%S Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop)
%D 2025
%8 April
%I Association for Computational Linguistics
%C Albuquerque, USA
%@ 979-8-89176-192-6
%F wu-etal-2025-towards
%X Neural codec language models (or codec LMs) are emerging as a powerful framework for audio generation tasks like text-to-speech (TTS). These models leverage advancements in language modeling and residual vector quantization (RVQ)-based audio codecs, which compress audios into discrete codes for LMs to process. Despite the close interdependence of codecs and LMs in these systems, research on codecs and LMs has largely remained siloed. In this work, we propose three techniques for better codec-LM co-design: (i) a frame-wise codec encoder that improves both LM log-likelihood and end-to-end TTS metrics, (ii) LM codebook level dropout, a method to efficiently navigate a portion of the codec-LM design space by training a single LM, and (iii) increased codec frame duration, which we show can accelerate inference while maintaining end-to-end performance. Our experiments demonstrate that combining all three co-design techniques results in doubled inference speed, and improvements in intelligibility, audio quality, and speaker control in TTS relative to a siloed baseline.
%R 10.18653/v1/2025.naacl-srw.6
%U https://aclanthology.org/2025.naacl-srw.6/
%U https://doi.org/10.18653/v1/2025.naacl-srw.6
%P 55-65
Markdown (Informal)
[Towards Codec-LM Co-design for Neural Codec Language Models](https://aclanthology.org/2025.naacl-srw.6/) (Wu et al., NAACL 2025)
ACL
- Shih-Lun Wu, Aakash Lahoti, Arjun D Desai, Karan Goel, Chris Donahue, and Albert Gu. 2025. Towards Codec-LM Co-design for Neural Codec Language Models. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 4: Student Research Workshop), pages 55–65, Albuquerque, USA. Association for Computational Linguistics.