@inproceedings{hussain-etal-2025-koel,
title = "Koel-{TTS}: Enhancing {LLM} based Speech Generation with Preference Alignment and Classifier Free Guidance",
author = "Hussain, Shehzeen Samarah and
Neekhara, Paarth and
Yang, Xuesong and
Casanova, Edresson and
Ghosh, Subhankar and
Fejgin, Roy and
Desta, Mikyas T. and
Valle, Rafael and
Li, Jason",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.emnlp-main.1076/",
pages = "21230--21245",
ISBN = "979-8-89176-332-6",
abstract = "Autoregressive speech token generation models produce speech with remarkable variety and naturalness but often suffer from hallucinations and undesired vocalizations that do not conform to conditioning inputs. To address these challenges, we introduce Koel-TTS, an encoder-decoder transformer model for multilingual TTS that improves contextual adherence of speech generation LLMs through preference alignment and classifier-free guidance (CFG). For preference alignment, we design a reward system that ranks model outputs using automatic metrics derived from speech recognition and speaker verification models, encouraging generations that better match the input text and speaker identity. CFG further allows fine-grained control over the influence of conditioning inputs during inference by interpolating conditional and unconditional logits. Notably, applying CFG to a preference-aligned model yields additional gains in transcription accuracy and speaker similarity, demonstrating the complementary benefits of both techniques. Koel-TTS achieves state-of-the-art results in zero-shot TTS, outperforming prior LLM-based models on intelligibility, speaker similarity, and naturalness, despite being trained on significantly less data."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="hussain-etal-2025-koel">
<titleInfo>
<title>Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shehzeen</namePart>
<namePart type="given">Samarah</namePart>
<namePart type="family">Hussain</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paarth</namePart>
<namePart type="family">Neekhara</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xuesong</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Edresson</namePart>
<namePart type="family">Casanova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Subhankar</namePart>
<namePart type="family">Ghosh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roy</namePart>
<namePart type="family">Fejgin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mikyas</namePart>
<namePart type="given">T</namePart>
<namePart type="family">Desta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rafael</namePart>
<namePart type="family">Valle</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jason</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christos</namePart>
<namePart type="family">Christodoulopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolyn</namePart>
<namePart type="family">Rose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Violet</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-332-6</identifier>
</relatedItem>
<abstract>Autoregressive speech token generation models produce speech with remarkable variety and naturalness but often suffer from hallucinations and undesired vocalizations that do not conform to conditioning inputs. To address these challenges, we introduce Koel-TTS, an encoder-decoder transformer model for multilingual TTS that improves contextual adherence of speech generation LLMs through preference alignment and classifier-free guidance (CFG). For preference alignment, we design a reward system that ranks model outputs using automatic metrics derived from speech recognition and speaker verification models, encouraging generations that better match the input text and speaker identity. CFG further allows fine-grained control over the influence of conditioning inputs during inference by interpolating conditional and unconditional logits. Notably, applying CFG to a preference-aligned model yields additional gains in transcription accuracy and speaker similarity, demonstrating the complementary benefits of both techniques. Koel-TTS achieves state-of-the-art results in zero-shot TTS, outperforming prior LLM-based models on intelligibility, speaker similarity, and naturalness, despite being trained on significantly less data.</abstract>
<identifier type="citekey">hussain-etal-2025-koel</identifier>
<location>
<url>https://aclanthology.org/2025.emnlp-main.1076/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>21230</start>
<end>21245</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance
%A Hussain, Shehzeen Samarah
%A Neekhara, Paarth
%A Yang, Xuesong
%A Casanova, Edresson
%A Ghosh, Subhankar
%A Fejgin, Roy
%A Desta, Mikyas T.
%A Valle, Rafael
%A Li, Jason
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-332-6
%F hussain-etal-2025-koel
%X Autoregressive speech token generation models produce speech with remarkable variety and naturalness but often suffer from hallucinations and undesired vocalizations that do not conform to conditioning inputs. To address these challenges, we introduce Koel-TTS, an encoder-decoder transformer model for multilingual TTS that improves contextual adherence of speech generation LLMs through preference alignment and classifier-free guidance (CFG). For preference alignment, we design a reward system that ranks model outputs using automatic metrics derived from speech recognition and speaker verification models, encouraging generations that better match the input text and speaker identity. CFG further allows fine-grained control over the influence of conditioning inputs during inference by interpolating conditional and unconditional logits. Notably, applying CFG to a preference-aligned model yields additional gains in transcription accuracy and speaker similarity, demonstrating the complementary benefits of both techniques. Koel-TTS achieves state-of-the-art results in zero-shot TTS, outperforming prior LLM-based models on intelligibility, speaker similarity, and naturalness, despite being trained on significantly less data.
%U https://aclanthology.org/2025.emnlp-main.1076/
%P 21230-21245
Markdown (Informal)
[Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance](https://aclanthology.org/2025.emnlp-main.1076/) (Hussain et al., EMNLP 2025)
ACL
- Shehzeen Samarah Hussain, Paarth Neekhara, Xuesong Yang, Edresson Casanova, Subhankar Ghosh, Roy Fejgin, Mikyas T. Desta, Rafael Valle, and Jason Li. 2025. Koel-TTS: Enhancing LLM based Speech Generation with Preference Alignment and Classifier Free Guidance. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 21230–21245, Suzhou, China. Association for Computational Linguistics.