Tacotron: Towards End-to-End Speech Synthesis. Wang, Y., Skerry-Ryan, R. J., Stanton, D., Wu, Y., Weiss, R. J., Jaitly, N., Yang, Z., Xiao, Y., Chen, Z., Bengio, S., Le, Q., Agiomyrgiannakis, Y., Clark, R., & Saurous, R. A. arXiv:1703.10135 [cs], April, 2017. arXiv: 1703.10135![Tacotron: Towards End-to-End Speech Synthesis [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper abstract bibtex A text-to-speech synthesis system typically consists of multiple stages, such as a text analysis frontend, an acoustic model and an audio synthesis module. Building these components often requires extensive domain expertise and may contain brittle design choices. In this paper, we present Tacotron, an end-to-end generative text-to-speech model that synthesizes speech directly from characters. Given \textlesstext, audio\textgreater pairs, the model can be trained completely from scratch with random initialization. We present several key techniques to make the sequence-tosequence framework perform well for this challenging task. Tacotron achieves a 3.82 subjective 5-scale mean opinion score on US English, outperforming a production parametric system in terms of naturalness. In addition, since Tacotron generates speech at the frame level, it’s substantially faster than sample-level autoregressive methods.
@article{wang_tacotron_2017,
title = {Tacotron: {Towards} {End}-to-{End} {Speech} {Synthesis}},
shorttitle = {Tacotron},
url = {http://arxiv.org/abs/1703.10135},
abstract = {A text-to-speech synthesis system typically consists of multiple stages, such as a text analysis frontend, an acoustic model and an audio synthesis module. Building these components often requires extensive domain expertise and may contain brittle design choices. In this paper, we present Tacotron, an end-to-end generative text-to-speech model that synthesizes speech directly from characters. Given {\textless}text, audio{\textgreater} pairs, the model can be trained completely from scratch with random initialization. We present several key techniques to make the sequence-tosequence framework perform well for this challenging task. Tacotron achieves a 3.82 subjective 5-scale mean opinion score on US English, outperforming a production parametric system in terms of naturalness. In addition, since Tacotron generates speech at the frame level, it’s substantially faster than sample-level autoregressive methods.},
language = {en},
urldate = {2022-01-19},
journal = {arXiv:1703.10135 [cs]},
author = {Wang, Yuxuan and Skerry-Ryan, R. J. and Stanton, Daisy and Wu, Yonghui and Weiss, Ron J. and Jaitly, Navdeep and Yang, Zongheng and Xiao, Ying and Chen, Zhifeng and Bengio, Samy and Le, Quoc and Agiomyrgiannakis, Yannis and Clark, Rob and Saurous, Rif A.},
month = apr,
year = {2017},
note = {arXiv: 1703.10135},
keywords = {/unread, Computer Science - Computation and Language, Computer Science - Machine Learning, Computer Science - Sound, ⛔ No DOI found},
}
Downloads: 0
{"_id":"7grbhCrRsdjaEZFas","bibbaseid":"wang-skerryryan-stanton-wu-weiss-jaitly-yang-xiao-etal-tacotrontowardsendtoendspeechsynthesis-2017","author_short":["Wang, Y.","Skerry-Ryan, R. J.","Stanton, D.","Wu, Y.","Weiss, R. J.","Jaitly, N.","Yang, Z.","Xiao, Y.","Chen, Z.","Bengio, S.","Le, Q.","Agiomyrgiannakis, Y.","Clark, R.","Saurous, R. A."],"bibdata":{"bibtype":"article","type":"article","title":"Tacotron: Towards End-to-End Speech Synthesis","shorttitle":"Tacotron","url":"http://arxiv.org/abs/1703.10135","abstract":"A text-to-speech synthesis system typically consists of multiple stages, such as a text analysis frontend, an acoustic model and an audio synthesis module. Building these components often requires extensive domain expertise and may contain brittle design choices. In this paper, we present Tacotron, an end-to-end generative text-to-speech model that synthesizes speech directly from characters. Given \\textlesstext, audio\\textgreater pairs, the model can be trained completely from scratch with random initialization. We present several key techniques to make the sequence-tosequence framework perform well for this challenging task. Tacotron achieves a 3.82 subjective 5-scale mean opinion score on US English, outperforming a production parametric system in terms of naturalness. In addition, since Tacotron generates speech at the frame level, it’s substantially faster than sample-level autoregressive methods.","language":"en","urldate":"2022-01-19","journal":"arXiv:1703.10135 [cs]","author":[{"propositions":[],"lastnames":["Wang"],"firstnames":["Yuxuan"],"suffixes":[]},{"propositions":[],"lastnames":["Skerry-Ryan"],"firstnames":["R.","J."],"suffixes":[]},{"propositions":[],"lastnames":["Stanton"],"firstnames":["Daisy"],"suffixes":[]},{"propositions":[],"lastnames":["Wu"],"firstnames":["Yonghui"],"suffixes":[]},{"propositions":[],"lastnames":["Weiss"],"firstnames":["Ron","J."],"suffixes":[]},{"propositions":[],"lastnames":["Jaitly"],"firstnames":["Navdeep"],"suffixes":[]},{"propositions":[],"lastnames":["Yang"],"firstnames":["Zongheng"],"suffixes":[]},{"propositions":[],"lastnames":["Xiao"],"firstnames":["Ying"],"suffixes":[]},{"propositions":[],"lastnames":["Chen"],"firstnames":["Zhifeng"],"suffixes":[]},{"propositions":[],"lastnames":["Bengio"],"firstnames":["Samy"],"suffixes":[]},{"propositions":[],"lastnames":["Le"],"firstnames":["Quoc"],"suffixes":[]},{"propositions":[],"lastnames":["Agiomyrgiannakis"],"firstnames":["Yannis"],"suffixes":[]},{"propositions":[],"lastnames":["Clark"],"firstnames":["Rob"],"suffixes":[]},{"propositions":[],"lastnames":["Saurous"],"firstnames":["Rif","A."],"suffixes":[]}],"month":"April","year":"2017","note":"arXiv: 1703.10135","keywords":"/unread, Computer Science - Computation and Language, Computer Science - Machine Learning, Computer Science - Sound, ⛔ No DOI found","bibtex":"@article{wang_tacotron_2017,\n\ttitle = {Tacotron: {Towards} {End}-to-{End} {Speech} {Synthesis}},\n\tshorttitle = {Tacotron},\n\turl = {http://arxiv.org/abs/1703.10135},\n\tabstract = {A text-to-speech synthesis system typically consists of multiple stages, such as a text analysis frontend, an acoustic model and an audio synthesis module. Building these components often requires extensive domain expertise and may contain brittle design choices. In this paper, we present Tacotron, an end-to-end generative text-to-speech model that synthesizes speech directly from characters. Given {\\textless}text, audio{\\textgreater} pairs, the model can be trained completely from scratch with random initialization. We present several key techniques to make the sequence-tosequence framework perform well for this challenging task. Tacotron achieves a 3.82 subjective 5-scale mean opinion score on US English, outperforming a production parametric system in terms of naturalness. In addition, since Tacotron generates speech at the frame level, it’s substantially faster than sample-level autoregressive methods.},\n\tlanguage = {en},\n\turldate = {2022-01-19},\n\tjournal = {arXiv:1703.10135 [cs]},\n\tauthor = {Wang, Yuxuan and Skerry-Ryan, R. J. and Stanton, Daisy and Wu, Yonghui and Weiss, Ron J. and Jaitly, Navdeep and Yang, Zongheng and Xiao, Ying and Chen, Zhifeng and Bengio, Samy and Le, Quoc and Agiomyrgiannakis, Yannis and Clark, Rob and Saurous, Rif A.},\n\tmonth = apr,\n\tyear = {2017},\n\tnote = {arXiv: 1703.10135},\n\tkeywords = {/unread, Computer Science - Computation and Language, Computer Science - Machine Learning, Computer Science - Sound, ⛔ No DOI found},\n}\n\n","author_short":["Wang, Y.","Skerry-Ryan, R. J.","Stanton, D.","Wu, Y.","Weiss, R. J.","Jaitly, N.","Yang, Z.","Xiao, Y.","Chen, Z.","Bengio, S.","Le, Q.","Agiomyrgiannakis, Y.","Clark, R.","Saurous, R. A."],"key":"wang_tacotron_2017","id":"wang_tacotron_2017","bibbaseid":"wang-skerryryan-stanton-wu-weiss-jaitly-yang-xiao-etal-tacotrontowardsendtoendspeechsynthesis-2017","role":"author","urls":{"Paper":"http://arxiv.org/abs/1703.10135"},"keyword":["/unread","Computer Science - Computation and Language","Computer Science - Machine Learning","Computer Science - Sound","⛔ No DOI found"],"metadata":{"authorlinks":{}},"html":""},"bibtype":"article","biburl":"https://bibbase.org/zotero/victorjhu","dataSources":["CmHEoydhafhbkXXt5"],"keywords":["/unread","computer science - computation and language","computer science - machine learning","computer science - sound","⛔ no doi found"],"search_terms":["tacotron","towards","end","end","speech","synthesis","wang","skerry-ryan","stanton","wu","weiss","jaitly","yang","xiao","chen","bengio","le","agiomyrgiannakis","clark","saurous"],"title":"Tacotron: Towards End-to-End Speech Synthesis","year":2017}