---
title: "Speak, Read and Prompt — paper by Kharitonov et al. (2023)"
description: "The authors build on AudioLM and split synthesis into two Transformer seq2seq stages. 92.4% accuracy. 42 references, cited by 41 papers here."
authors: ["Eugene Kharitonov", "Damien Vincent", "Zalán Borsos", "Raphaël Marinier", "Sertan Girgin", "Olivier Pietquin", "Matt Sharifi", "Marco Tagliasacchi", "Neil Zeghidour"]
year: 2023
venue: "Transactions of the Association for Computational Linguistics"
source: "https://aclanthology.org/2023.tacl-1.95/"
doi: "10.1162/tacl_a_00618"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/kharitonov-2023-speak-read-prompt-high-fidelity/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/kharitonov-2023-speak-read-prompt-high-fidelity/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2023](https://md.research.elamur.ai/free/en/papers/2023/page/5/) / Speak, Read and Prompt: High-Fidelity Text-to-Speech with Minimal Supervision

# Speak, Read and Prompt: High-Fidelity Text-to-Speech with Minimal Supervision

**Authors:** [Eugene Kharitonov](https://md.research.elamur.ai/free/en/authors/eugene-kharitonov/), [Damien Vincent](https://md.research.elamur.ai/free/en/authors/damien-vincent/), [Zalán Borsos](https://md.research.elamur.ai/free/en/authors/zalan-borsos/), [Raphaël Marinier](https://md.research.elamur.ai/free/en/authors/raphael-marinier/), Sertan Girgin, [Olivier Pietquin](https://md.research.elamur.ai/free/en/authors/olivier-pietquin/), [Matt Sharifi](https://md.research.elamur.ai/free/en/authors/matt-sharifi/), [Marco Tagliasacchi](https://md.research.elamur.ai/free/en/authors/marco-tagliasacchi/) and [Neil Zeghidour](https://md.research.elamur.ai/free/en/authors/neil-zeghidour/) · **Year:** 2023 · **Venue:** [Transactions of the Association for Computational Linguistics](https://md.research.elamur.ai/free/en/venues/tacl/) ([TACL 2023](https://md.research.elamur.ai/free/en/venues/tacl/2023/)) · **Publisher:** [The MIT Press](https://md.research.elamur.ai/free/en/publishers/the-mit-press/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2023.tacl-1.95/) · [PDF](https://aclanthology.org/2023.tacl-1.95.pdf) · [DOI 10.1162/tacl\_a\_00618](https://doi.org/10.1162/tacl_a_00618) · [OpenAlex](https://openalex.org/W4390075359)

**Citations:** 102 citations (OpenAlex, 2026-10-14) · [41](https://md.research.elamur.ai/free/en/papers/kharitonov-2023-speak-read-prompt-high-fidelity/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
SPEAR-TTS addresses the fact that text-to-speech systems usually need hundreds of hours of transcribed audio, which leaves most of the world's languages without high-quality TTS. The authors build on AudioLM and split synthesis into two Transformer seq2seq stages. The first stage, "reading", converts text into semantic tokens derived from a w2v-BERT model. The second stage, "speaking", turns those semantic tokens into acoustic tokens from a SoundStream codec, which are then decoded to a waveform. Because the speaking stage never sees text, it can learn from untranscribed audio alone. The reading stage, in turn, uses BART-style denoising pretraining and backtranslation, so it needs only a small parallel set. Voice control works by prompting the model with a short audio clip, with no speaker labels or speaker embeddings involved. On English, 15 minutes of single-speaker parallel data was enough to reach intelligibility close to strong supervised systems and naturalness on par with real recordings. This makes the approach promising for under-resourced languages, Ukrainian included, though the paper only tests English.

## Key points
- **Training data:** the parallel data is 15 minutes sampled from LJSpeech. Audio-only data comes from LibriTTS (551 hours) for pretraining and backtranslation. The speaking stage still needs several thousand hours of untranscribed audio.
- **Intelligibility:** CER is 1.92% on LibriSpeech test-clean. This is comparable to an adapted FastSpeech2-LR trained on 24 hours of LJSpeech, and SPEAR-TTS beats a single-voice baseline in naturalness while using 96× less parallel data.
- **Zero-shot voice cloning:** a 3-second prompt preserves the target speaker's voice with 92.4% accuracy according to a speaker classifier. Speaker similarity is close to VALL-E and clearly above YourTTS.
- **Audio quality (MOS):** SPEAR-TTS scores 4.96 versus 4.92 for ground-truth speech. In zero-shot voice transfer it scores 4.75 versus 3.35 for VALL-E, which was trained on about 60k hours of LibriLight.
- **Safety and availability:** a classifier detects the synthetic speech with 82.5% accuracy. The authors did not release checkpoints or training code because of misuse concerns, and applicability to non-English or tonal languages remains untested.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Google](https://md.research.elamur.ai/free/en/companies/google/) — Company, Mountain View, United States · 353 works · 364,806 citations
- [Google (Switzerland)](https://md.research.elamur.ai/free/en/companies/google-switzerland/) — Company, Zurich, Switzerland · 10 works · 27,513 citations

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works · [🔊 Signal Processing](https://md.research.elamur.ai/free/en/shelves/computer-science/signal-processing/) · 213 works

🧭 Topics (1): [🎙️ Speech & spoken language](https://md.research.elamur.ai/free/en/topics/speech/) · 233 works

## Cite

Cite as: Eugene Kharitonov, Damien Vincent, Zalán Borsos, Raphaël Marinier, Sertan Girgin, Olivier Pietquin, Matt Sharifi, Marco Tagliasacchi and Neil Zeghidour. 2023. Speak, Read and Prompt: High-Fidelity Text-to-Speech with Minimal Supervision. Transactions of the Association for Computational Linguistics.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@article{kharitonov2023speak,
  title = {{Speak, Read and Prompt: High-Fidelity Text-to-Speech with Minimal Supervision}},
  author = {Eugene Kharitonov, Damien Vincent, Zalán Borsos, Raphaël Marinier, Sertan Girgin, Olivier Pietquin, Matt Sharifi, Marco Tagliasacchi and Neil Zeghidour},
  journal = {Transactions of the Association for Computational Linguistics},
  year = {2023},
  doi = {10.1162/tacl\_a\_00618},
  url = {https://doi.org/10.1162/tacl_a_00618}
}
```

- [← previous MiniLLM: On-Policy Distillation of Large Language Models](https://md.research.elamur.ai/free/en/papers/gu-2023-minillm-policy-distillation-large-language/)
- [next → In-Context Learning Creates Task Vectors](https://md.research.elamur.ai/free/en/papers/hendel-2023-context-learning-creates-task-vectors/)
- [↑ back to papers of 2023](https://md.research.elamur.ai/free/en/papers/2023/page/5/)

## References

42 references, 13 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Andrea Agostinelli, Timo I. Denk, Zalán Borsos, Jesse Engel, Mauro Verzetti, Antoine Caillon, Qingqing Huang, Aren Jansen, Adam Roberts, Marco Tagliasacchi, Matt Sharifi, Neil Zeghidour, and Christian Frank. 2023. Music-LM: Generating music from text. arXiv preprint arXiv:2301.11325. — [arXiv:2301.11325](https://arxiv.org/abs/2301.11325) [find](https://scholar.google.com/scholar?q=Andrea+Agostinelli%2C+Timo+I%2E+Denk%2C+Zala%CC%81n+Borsos%2C+Jesse+Engel%2C+Mauro+Verzetti%2C+Antoine+Caillon%2C+Qingqing+Huang%2C+Aren+Jansen%2C+Adam+Roberts%2C+Marco+Tagliasacchi%2C+Matt+Sharifi%2C+Neil+Zeghidour%2C+and+Christia)
2. Junyi Ao, Rui Wang, Long Zhou, Chengyi Wang, Shuo Ren, Yu Wu, Shujie Liu, Tom Ko, Qing Li, Yu Zhang, Zhihua Wei, Yao Qian, Jinyu Li, and Furu Wei. 2022. SpeechT5: Unified-modal encoder-decoder pre-training for spoken language processing. In TACL. https://doi.org/10.18653/v1/2022 .acl-long.393 [↗](https://doi.org/10.18653/v1/2022) [our card](https://md.research.elamur.ai/free/en/papers/ao-2022-speecht5-unified-modal-encoder-decoder/)
3. James Betker. 2022. TorToiSe text-to-speech. [find](https://scholar.google.com/scholar?q=James+Betker%2E+2022%2E+TorToiSe+text%2Dto%2Dspeech%2E)
4. Zalán Borsos, Raphaël Marinier, Damien Vincent, Eugene Kharitonov, Olivier Pietquin, Matt Sharifi, Dominik Roblek, Olivier Teboul, David Grangier, Marco Tagliasacchi, and Neil Zeghidour. 2023. AudioLM: A language modeling approach to audio generation. IEEE/ACM Transactions on Audio, Speech, and La… [our card](https://md.research.elamur.ai/free/en/papers/borsos-2022-audiolm-language-modeling-approach-audio/)
5. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D. Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Edresson Casanova, Julian Weber, Christopher D.… [find](https://scholar.google.com/scholar?q=Tom+Brown%2C+Benjamin+Mann%2C+Nick+Ryder%2C+Melanie+Subbiah%2C+Jared+D%2E+Kaplan%2C+Prafulla+Dhariwal%2C+Arvind+Neelakantan%2C+Pranav+Shyam%2C+Girish+Sastry%2C+Amanda+Askell%2C+Sandhini+Agarwal%2C+Ariel+Herbert%2DVoss%2C+Gretche)
6. Sanyuan Chen, Chengyi Wang, Zhengyang Chen, Yu Wu, Shujie Liu, Zhuo Chen, Jinyu Li, Naoyuki Kanda, Takuya Yoshioka, Xiong Xiao, Jian Wu, Long Zhou, Shuo Ren, Yanmin Qian, Yao Qian, Jian Wu, Michael Zeng, Xiangzhan Yu, and Furu Wei. 2022. WavLM: Large-scale self-supervised pre-training for full stack… [find](https://scholar.google.com/scholar?q=Sanyuan+Chen%2C+Chengyi+Wang%2C+Zhengyang+Chen%2C+Yu+Wu%2C+Shujie+Liu%2C+Zhuo+Chen%2C+Jinyu+Li%2C+Naoyuki+Kanda%2C+Takuya+Yoshioka%2C+Xiong+Xiao%2C+Jian+Wu%2C+Long+Zhou%2C+Shuo+Ren%2C+Yanmin+Qian%2C+Yao+Qian%2C+Jian+Wu%2C+Michael+Ze)
7. Yu-An Chung, Yu Zhang, Wei Han, Chung-Cheng Chiu, James Qin, Ruoming Pang, and Yonghui Wu. 2021. W2V-Bert: Combining contrastive learning and masked language modeling for self-supervised speech pre-training. In ASRU. https://doi.org/10.1109/ASRU51503.2021.9688253 [↗](https://doi.org/10.1109/ASRU51503.2021.9688253)
8. Héctor Delgado, Nicholas Evans, Tomi Kinnunen, Kong Aik Lee, Xuechen Liu, Andreas Nautsch, Jose Patino, Md Sahidullah, Massimiliano Todisco, Xin Wang, Junichi Yamagishi. 2021. [find](https://scholar.google.com/scholar?q=He%CC%81ctor+Delgado%2C+Nicholas+Evans%2C+Tomi+Kinnunen%2C+Kong+Aik+Lee%2C+Xuechen+Liu%2C+Andreas+Nautsch%2C+Jose+Patino%2C+Md+Sahidullah%2C+Massimiliano+Todisco%2C+Xin+Wang%2C+Junichi+Yamagishi%2E+2021%2E)
9. Asvspoof 2021: Automatic speaker verification spoofing and countermeasures challenge evaluation plan. arXiv preprint arXiv:2109.00535. — [arXiv:2109.00535](https://arxiv.org/abs/2109.00535) [find](https://scholar.google.com/scholar?q=Asvspoof+2021%3A+Automatic+speaker+verification+spoofing+and+countermeasures+challenge+evaluation+plan%2E+arXiv+preprint+arXiv%3A2109%2E00535%2E+%E2%80%94+%5BarXiv%3A2109%2E00535%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2109%2E00535%29)
10. Sergey Edunov, Myle Ott, Michael Auli, and David Grangier. 2018. Understanding back-translation at scale. In EMNLP. https://doi.org/10.18653/v1/D18-1045 [↗](https://doi.org/10.18653/v1/D18-1045)
11. Sergey Ioffe and Christian Szegedy. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. In ICML. [find](https://scholar.google.com/scholar?q=Sergey+Ioffe+and+Christian+Szegedy%2E+2015%2E+Batch+normalization%3A+Accelerating+deep+network+training+by+reducing+internal+covariate+shift%2E+In+ICML%2E)
12. Keith Ito and Linda Johnson. 2017. The LJ speech dataset. [https://keithito.com/LJ-Speech-Dataset/](https://keithito.com/LJ-Speech-Dataset/). [↗](https://keithito.com/LJ-Speech-Dataset/)
13. Norm Jouppi, George Kurian, Sheng Li, Peter Ma, Rahul Nagarajan, Lifeng Nai, Nishant Patil, Suvinay Subramanian, Andy Swing, Brian Towles, Clifford Young, Xiang Zhou, Zongwei Zhou, and David A. Patterson. 2023. Tpu v4: An optically reconfigurable supercomputer for machine learning with hardware supp… [find](https://scholar.google.com/scholar?q=Norm+Jouppi%2C+George+Kurian%2C+Sheng+Li%2C+Peter+Ma%2C+Rahul+Nagarajan%2C+Lifeng+Nai%2C+Nishant+Patil%2C+Suvinay+Subramanian%2C+Andy+Swing%2C+Brian+Towles%2C+Clifford+Young%2C+Xiang+Zhou%2C+Zongwei+Zhou%2C+and+David+A%2E+Patter)
14. Jacob Kahn, Morgane Rivière, Weiyi Zheng, Evgeny Kharitonov, Qiantong Xu, Pierre-Emmanuel Mazaré, Julien Karadayi, Vitaliy Liptchinsky, Ronan Collobert, Christian Fuegen, Tatiana Likhomanenko, Gabriel Synnaeve, Armand Joulin, Abdelrahman Mohamed, and Emmanuel Dupoux. 2020. Libri-Light: A benchmark… [find](https://scholar.google.com/scholar?q=Jacob+Kahn%2C+Morgane+Rivie%CC%80re%2C+Weiyi+Zheng%2C+Evgeny+Kharitonov%2C+Qiantong+Xu%2C+Pierre%2DEmmanuel+Mazare%CC%81%2C+Julien+Karadayi%2C+Vitaliy+Liptchinsky%2C+Ronan+Collobert%2C+Christian+Fuegen%2C+Tatiana+Likhomanenko%2C+Gabri)
15. Eugene Kharitonov, Ann Lee, Adam Polyak, Yossi Adi, Jade Copet, Kushal Lakhotia, Tu Anh Nguyen, Morgane Riviere, Abdelrahman Mohamed, Emmanuel Dupoux, and Wei-Ning Hsu. 2022. Text-free prosody-aware generative spoken language modeling. In ACL. https://doi.org/10.18653/v1/2022.acl-long.593 [↗](https://doi.org/10.18653/v1/2022.acl-long.593) [our card](https://md.research.elamur.ai/free/en/papers/kharitonov-2022-text-free-prosody-aware-generative/)
16. Heeseung Kim, Sungwon Kim, and Sungroh Yoon. 2021. Guided-TTS: Text-to-speech with untranscribed speech. arXiv preprint arXiv: 2111.11755. — [arXiv:2111.11755](https://arxiv.org/abs/2111.11755) [find](https://scholar.google.com/scholar?q=Heeseung+Kim%2C+Sungwon+Kim%2C+and+Sungroh+Yoon%2E+2021%2E+Guided%2DTTS%3A+Text%2Dto%2Dspeech+with+untranscribed+speech%2E+arXiv+preprint+arXiv%3A+2111%2E11755%2E+%E2%80%94+%5BarXiv%3A2111%2E11755%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2111%2E11755%29)
17. Sungwon Kim, Heeseung Kim, and Sungroh Yoon. 2022. Guided-TTS 2: A diffusion model for high-quality adaptive text-to-speech with untranscribed data. arXiv preprint arXiv:2205 .15370. [find](https://scholar.google.com/scholar?q=Sungwon+Kim%2C+Heeseung+Kim%2C+and+Sungroh+Yoon%2E+2022%2E+Guided%2DTTS+2%3A+A+diffusion+model+for+high%2Dquality+adaptive+text%2Dto%2Dspeech+with+untranscribed+data%2E+arXiv+preprint+arXiv%3A2205+%2E15370%2E)
18. Jungil Kong, Jaehyeon Kim, and Jaekyoung Bae. 2020. Hifi-gan: Generative adversarial networks for efficient and high fidelity speech synthesis. NeurIPS. [find](https://scholar.google.com/scholar?q=Jungil+Kong%2C+Jaehyeon+Kim%2C+and+Jaekyoung+Bae%2E+2020%2E+Hifi%2Dgan%3A+Generative+adversarial+networks+for+efficient+and+high+fidelity+speech+synthesis%2E+NeurIPS%2E)
19. Felix Kreuk, Adam Polyak, Jade Copet, Eugene Kharitonov, Tu-Anh Nguyen, Morgane Rivière, Wei-Ning Hsu, Abdelrahman Mohamed, Emmanuel Dupoux, and Yossi Adi. 2021. Textless speech emotion conversion using decomposed and discrete representations. arXiv preprint arXiv:2111.07402. [https://doi.org](https://doi.org) /10.18… [↗](https://doi.org)
20. Kushal Lakhotia, Eugene Kharitonov, Wei-Ning Hsu, Yossi Adi, Adam Polyak, Benjamin Bolte, Tu-Anh Nguyen, Jade Copet, Alexei Baevski, Abdelrahman Mohamed, Emmanuel Dupoux. 2021. On generative spoken language modeling from raw audio. TACL. [our card](https://md.research.elamur.ai/free/en/papers/lakhotia-2021-generative-spoken-language-modeling-raw/)
21. Alon Levkovitch, Eliya Nachmani, and Lior Wolf. 2022. Zero-shot voice conditioning for denoising diffusion TTS models. arXiv preprint arXiv:2206.02246. https://doi.org/10.21437/Interspeech.2022-10045 — [doi:10.21437/interspeech.2022-10045](https://doi.org/10.21437/interspeech.2022-10045) [↗](https://doi.org/10.21437/Interspeech.2022-10045)
22. Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension. In ACL. https://doi.org/10.18653/v1 /2020.acl-ma… [our card](https://md.research.elamur.ai/free/en/papers/lewis-2020-bart-denoising-sequence-sequence-pre/)
23. Alexander H. Liu, Cheng-I Jeff Lai, Wei-Ning Hsu, Michael Auli, Alexei Baevski, and James Glass. 2022. Simple and effective unsupervised speech synthesis. arXiv preprint arXiv: 2204.02524. — [arXiv:2204.02524](https://arxiv.org/abs/2204.02524) [find](https://scholar.google.com/scholar?q=Alexander+H%2E+Liu%2C+Cheng%2DI+Jeff+Lai%2C+Wei%2DNing+Hsu%2C+Michael+Auli%2C+Alexei+Baevski%2C+and+James+Glass%2E+2022%2E+Simple+and+effective+unsupervised+speech+synthesis%2E+arXiv+preprint+arXiv%3A+2204%2E02524%2E+%E2%80%94+%5BarXiv%3A22)
24. Alexander H. Liu, Tao Tu, Hung-yi Lee, and Lin-shan Lee. 2020. Towards unsupervised speech recognition and synthesis with quantized speech representation learning. In ICASSP. [find](https://scholar.google.com/scholar?q=Alexander+H%2E+Liu%2C+Tao+Tu%2C+Hung%2Dyi+Lee%2C+and+Lin%2Dshan+Lee%2E+2020%2E+Towards+unsupervised+speech+recognition+and+synthesis+with+quantized+speech+representation+learning%2E+In+ICASSP%2E)
25. Tu Anh Nguyen, Eugene Kharitonov, Jade Copet, Yossi Adi, Wei-Ning Hsu, Ali Elkahky, Paden Tomasello, Robin Algayres, Benoı̂t Sagot, Abdelrahman Mohamed, and Emmanuel Dupoux. 2023. Generative spoken dialogue language modeling. TACL. https://doi.org/10.1162/tacl_a_00545 [↗](https://doi.org/10.1162/tacl_a_00545) [our card](https://md.research.elamur.ai/free/en/papers/nguyen-2023-generative-spoken-dialogue-language-modeling/)
26. Vassil Panayotov, Guoguo Chen, Daniel Povey, and Sanjeev Khudanpur. 2015. LibriSpeech: An ASR corpus based on public domain audio books. In ICASSP. https://doi.org/10 .1109/ICASSP.2015.7178964 [find](https://scholar.google.com/scholar?q=Vassil+Panayotov%2C+Guoguo+Chen%2C+Daniel+Povey%2C+and+Sanjeev+Khudanpur%2E+2015%2E+LibriSpeech%3A+An+ASR+corpus+based+on+public+domain+audio+books%2E+In+ICASSP%2E+https%3A%2F%2Fdoi%2Eorg%2F10+%2E1109%2FICASSP%2E2015%2E7178964)
27. Kyubyong Park and Jongseok Kim. 2019. g2p. [https://github.com/Kyubyong/g2p](https://github.com/Kyubyong/g2p). [↗](https://github.com/Kyubyong/g2p)
28. Aidan Pine, Dan Wells, Nathan Brinklow, Patrick Littell, and Korin Richmond. 2022. Requirements and motivations of low-resource speech synthesis for language revitalization. In ACL. https://doi.org/10.18653/v1/2022.acl-long.507 model that can speak and listen. arXiv preprint arXiv: arXiv:2306.12925… [↗](https://doi.org/10.18653/v1/2022.acl-long.507) [our card](https://md.research.elamur.ai/free/en/papers/rubenstein-2023-audiopalm-large-language-model-that/)
29. Adam Polyak, Yossi Adi, Jade Copet, Eugene Kharitonov, Kushal Lakhotia, Wei-Ning Hsu, Abdelrahman Mohamed, and Emmanuel Dupoux. 2021. Speech resynthesis from discrete disentangled self-supervised representations. In Interspeech. https://doi.org/10.21437/Interspeech.2021-475 — [doi:10.21437/interspeech.2021-475](https://doi.org/10.21437/interspeech.2021-475) [↗](https://doi.org/10.21437/Interspeech.2021-475)
30. Rico Sennrich, Barry Haddow, and Alexandra Birch. 2016. Improving neural machine translation models with monolingual data. In ACL. https://doi.org/10.18653/v1/P16 -1009 [↗](https://doi.org/10.18653/v1/P16)
31. Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, and Peter J. Liu. 2020. Exploring the limits of transfer learning with a unified text-to-text transformer. JMLR, 21. [our card](https://md.research.elamur.ai/free/en/papers/raffel-2019-exploring-limits-transfer-learning-unified/)
32. Chandan K. A. Reddy, Vishak Gopal, and Ross Cutler. 2021. DNSMOS: A non-intrusive perceptual objective speech quality metric to evaluate noise suppressors. In ICASSP. [find](https://scholar.google.com/scholar?q=Chandan+K%2E+A%2E+Reddy%2C+Vishak+Gopal%2C+and+Ross+Cutler%2E+2021%2E+DNSMOS%3A+A+non%2Dintrusive+perceptual+objective+speech+quality+metric+to+evaluate+noise+suppressors%2E+In+ICASSP%2E)
33. Yi Ren, Chenxu Hu, Xu Tan, Tao Qin, Sheng Zhao, Zhou Zhao, and Tie-Yan Liu. 2020. FastSpeech 2: Fast and high-quality end-to-end text to speech. In ICLR. [find](https://scholar.google.com/scholar?q=Yi+Ren%2C+Chenxu+Hu%2C+Xu+Tan%2C+Tao+Qin%2C+Sheng+Zhao%2C+Zhou+Zhao%2C+and+Tie%2DYan+Liu%2E+2020%2E+FastSpeech+2%3A+Fast+and+high%2Dquality+end%2Dto%2Dend+text+to+speech%2E+In+ICLR%2E)
34. Morgane Riviere, Armand Joulin, Pierre-Emmanuel Mazaré, and Emmanuel Dupoux. 2020. Unsupervised pretraining transfers well across languages. In ICASSP 2020-2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 7414–7418. IEEE. [https://doi.org](https://doi.org) /10.1109/ICASSP4… — [doi:10.1109/icassp40776.2020.9054548](https://doi.org/10.1109/icassp40776.2020.9054548) [↗](https://doi.org)
35. Paul K. Rubenstein, Chulayuth Asawaroengchai, Duc Dung Nguyen, Ankur Bapna, Zalán Borsos, Félix de Chaumont Quitry, Peter Chen, Dalia El Badawy, Wei Han, Eugene Kharitonov, Hannah Muckenhirn, Dirk Padfield, James Qin, Danny Rozenberg, Tara Sainath, Johan Schalkwyk, Matt Sharifi, Michelle Tadmor, R… [find](https://scholar.google.com/scholar?q=Paul+K%2E+Rubenstein%2C+Chulayuth+Asawaroengchai%2C+Duc+Dung+Nguyen%2C+Ankur+Bapna%2C+Zala%CC%81n+Borsos%2C+Fe%CC%81lix+de+Chaumont+Quitry%2C+Peter+Chen%2C+Dalia+El+Badawy%2C+Wei+Han%2C+Eugene+Kharitonov%2C+Hannah+Muckenhirn%2C+Dirk+P)
36. Jonathan Shen, Ruoming Pang, Ron J. Weiss, Mike Schuster, Navdeep Jaitly, Zongheng Yang, Zhifeng Chen, Yu Zhang, Yuxuan Wang, Rj Skerrv-Ryan, Rif A. Saurous, Yannis Agiomvrgiannakis, and Yonghui Wu. 2018. Natural TTS synthesis by conditioning Wave-Net on mel spectrogram predictions. In ICASSP. https… — [doi:10.1109/icassp.2018](https://doi.org/10.1109/icassp.2018) [find](https://scholar.google.com/scholar?q=Jonathan+Shen%2C+Ruoming+Pang%2C+Ron+J%2E+Weiss%2C+Mike+Schuster%2C+Navdeep+Jaitly%2C+Zongheng+Yang%2C+Zhifeng+Chen%2C+Yu+Zhang%2C+Yuxuan+Wang%2C+Rj+Skerrv%2DRyan%2C+Rif+A%2E+Saurous%2C+Yannis+Agiomvrgiannakis%2C+and+Yonghui+Wu%2E+2)
37. Masao Utiyama and Hitoshi Isahara. 2007. A comparison of pivot methods for phrase-based statistical machine translation. In NAACL. [our card](https://md.research.elamur.ai/free/en/papers/utiyama-2007-comparison-pivot-methods-phrase-based/)
38. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. NIPS. [our card](https://md.research.elamur.ai/free/en/papers/vaswani-2017-attention-all-need/)
39. Chengyi Wang, Sanyuan Chen, Yu Wu, Ziqiang Zhang, Long Zhou, Shujie Liu, Zhuo Chen, Yanqing Liu, Huaming Wang, Jinyu Li, Lei He, Sheng Zhao, and Furu Wei. 2023. Neural codec language models are zero-shot text to speech synthesizers. arXiv preprint arXiv:2301.02111. [our card](https://md.research.elamur.ai/free/en/papers/wang-2023-neural-codec-language-models-zero/)
40. Neil Zeghidour, Alejandro Luebs, Ahmed Omran, Jan Skoglund, and Marco Tagliasacchi. 2021. SoundStream: An end-to-end neural audio codec. IEEE/ACM Transactions on Audio, Speech, and Language Processing. https://doi.org/10.1109/TASLP.2021.3129994 [↗](https://doi.org/10.1109/TASLP.2021.3129994) [our card](https://md.research.elamur.ai/free/en/papers/zeghidour-2021-soundstream-end-end-neural-audio/)
41. Heiga Zen, Viet Dang, Rob Clark, Yu Zhang, Ron J. Weiss, Ye Jia, Zhifeng Chen, and Yonghui Wu. 2019. LibriTTS: A corpus derived from LibriSpeech for text-to-speech. arXiv preprint arXiv:1904.02882. https://doi.org/10.21437/Interspeech.2019 -2441 [↗](https://doi.org/10.21437/Interspeech.2019) [our card](https://md.research.elamur.ai/free/en/papers/zen-2019-libritts-corpus-derived-librispeech-text/)
42. Downsample factor B Computational Cost CER (%) [find](https://scholar.google.com/scholar?q=Downsample+factor+B+Computational+Cost+CER+%28%25%29)

*Updated 24 October 2026*