SoundTwin: A High-Similarity Fast Diffusion Autoregressive Speech Cloning Model Based on Local-Global Feature Fusion

SoundTwin: A High-Similarity Fast Diffusion Autoregressive Speech Cloning Model Based on Local-Global Feature Fusion | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article SoundTwin: A High-Similarity Fast Diffusion Autoregressive Speech Cloning Model Based on Local-Global Feature Fusion Qizhang Li, Haiping Qu, Yongqi Kang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7512915/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract In recent years, significant advances in deep learning have propelled text-to-speech (TTS) technology forward. However, there are still challenges in high-quality voice cloning with low-resource and low-latency conditions. Traditional autoregressive models suffer from high inference latency, while emerging diffusion models, although capable of generating high-fidelity speech, incur substantial computational overhead due to their multi-step sampling process. To mitigate these limitations, we propose SoundTwin, a novel speech synthesis framework integrating accelerated diffusion sampling with an autoregressive Transformer architecture. This approach significantly enhances synthesis efficiency without compromising speech naturalness. Furthermore, we design a Local-Global Squeeze-and-Excitation weighted Speaker embedding Network to efficiently extract fine-grained timbre features from limited reference audio, enabling rapid speaker adaptation. The model accepts target text, reference speech, and reference text as inputs, jointly modeling duration, pitch, and energy features to generate high-quality mel-spectrograms. Experimental results validate that our method achieves state-of-the-art speaker similarity and speech naturalness in zero-shot voice cloning tasks. Text-to-Speech Synthesis Zero-Shot Voice Cloning Fast Diffusion Models AutoRegressive Neural Networks Speaker Embedding Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 28 Nov, 2025 Reviewers invited by journal 20 Nov, 2025 Editor assigned by journal 20 Nov, 2025 Submission checks completed at journal 04 Sep, 2025 First submitted to journal 01 Sep, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7512915","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":549562325,"identity":"a6674747-67ae-4204-aae4-fa3dd02f5791","order_by":0,"name":"Qizhang Li","email":"","orcid":"","institution":"Ludong University","correspondingAuthor":false,"prefix":"","firstName":"Qizhang","middleName":"","lastName":"Li","suffix":""},{"id":549562326,"identity":"9aa38430-40a8-455b-a0cb-c61ddc91c224","order_by":1,"name":"Haiping Qu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4UlEQVRIiWNgGAWjYFACHgaGBAYJBgb25gYgycDYQLwWnoOkaAEDiUQitfDdyD344eEOizz5yIeNt3kYbGQ3HGB+9gCfFskbeckSiWckig1vJzZb8zCkGW84wGZugE+LwY0cA4nENonEjbMT26R5GA4nbjjAwyZBQIvxD7CWmQdBWv4TpcUMbMt8CUaQlgOEtUieeWNmAfRL4gaexGbLOQbJxjMPs5nh1cJ3PMf45s8ddYnz2w8fvPGmwk6273jzM7xaGA4wQOLC4ADYnUDMjFc9khb5BkIqR8EoGAWjYMQCANmCTTRn8XKyAAAAAElFTkSuQmCC","orcid":"","institution":"Ludong University","correspondingAuthor":true,"prefix":"","firstName":"Haiping","middleName":"","lastName":"Qu","suffix":""},{"id":549562327,"identity":"3279ce23-fb0b-498e-90a3-439ad4d306ca","order_by":2,"name":"Yongqi Kang","email":"","orcid":"","institution":"Ludong University","correspondingAuthor":false,"prefix":"","firstName":"Yongqi","middleName":"","lastName":"Kang","suffix":""}],"badges":[],"createdAt":"2025-09-02 03:38:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7512915/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7512915/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97136178,"identity":"306a9a21-9c2b-4c8b-9bee-18cbcc10575a","added_by":"auto","created_at":"2025-12-01 09:55:56","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":616127,"visible":true,"origin":"","legend":"","description":"","filename":"SoundTwinAHighSimilarityFastDiffusionAutoregressiveSpeechCloningModelBasedonLocalGlobalFeatureFusion.docx","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/d52fde53883a744d822eb42b.docx"},{"id":96949755,"identity":"eb56ef66-456f-4b3b-bb95-5829a5b41b89","added_by":"auto","created_at":"2025-11-27 23:24:45","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4874,"visible":true,"origin":"","legend":"","description":"","filename":"0ac6520c06164d789cf3410ed06a7c02.json","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/dd24f0f403e4f7ebd5bd31d0.json"},{"id":96949764,"identity":"53707b39-044c-499d-869e-52d793073000","added_by":"auto","created_at":"2025-11-27 23:24:45","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":150591,"visible":true,"origin":"","legend":"","description":"","filename":"0ac6520c06164d789cf3410ed06a7c021enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/054925baebfc065123f248a4.xml"},{"id":97137226,"identity":"93a5caff-ef3b-4744-aa68-ec88242be575","added_by":"auto","created_at":"2025-12-01 09:57:29","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":70072,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/4e06752c322ca8533d96db7c.png"},{"id":96949758,"identity":"a26229cd-b1ab-4921-ae63-b35deab02b29","added_by":"auto","created_at":"2025-11-27 23:24:45","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":65771,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/c56f65e7d2bb20d50ae5e064.png"},{"id":96949760,"identity":"a8766384-9c1f-46c8-80ae-ac4706060827","added_by":"auto","created_at":"2025-11-27 23:24:45","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":104829,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/7934f5902d559216111407b2.png"},{"id":96949761,"identity":"50ffeef0-c64b-4f16-b83e-d7d7f7dcacd3","added_by":"auto","created_at":"2025-11-27 23:24:45","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":285166,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/92dbec2252bf2b157c301a11.png"},{"id":97136217,"identity":"73629e23-cc79-444f-94ba-1f9b8fee3e80","added_by":"auto","created_at":"2025-12-01 09:56:04","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":17836,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/d1e87c0cff2bfbd9f889d95c.png"},{"id":96949757,"identity":"4c1ec329-87c4-4bc3-95d0-0ff17d55d710","added_by":"auto","created_at":"2025-11-27 23:24:45","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":19322,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/6519b1870afd04d13fc877bd.png"},{"id":97136582,"identity":"d1f35ff4-0a4e-414a-946e-ad2de3edf09b","added_by":"auto","created_at":"2025-12-01 09:56:48","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":22831,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/3d99fa5a24a7b13304069469.png"},{"id":96949765,"identity":"98569025-5390-4f0d-aee3-924b888aeabc","added_by":"auto","created_at":"2025-11-27 23:24:46","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":61674,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/dcf8316c7191716eacf5c8f4.png"},{"id":96949766,"identity":"54cc5429-6f85-4c6c-9c1e-f141f957af9a","added_by":"auto","created_at":"2025-11-27 23:24:46","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":148079,"visible":true,"origin":"","legend":"","description":"","filename":"0ac6520c06164d789cf3410ed06a7c021structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/e0a5289a58ee0cea476084dc.xml"},{"id":96949767,"identity":"e77e95ae-f303-4c2b-abd9-9b44e08baed6","added_by":"auto","created_at":"2025-11-27 23:24:46","extension":"html","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":167108,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1/308e6a037bcd5dba558e114c.html"},{"id":97144711,"identity":"380c5f14-cbaf-44fc-8296-40f7e662bb0e","added_by":"auto","created_at":"2025-12-01 10:11:44","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":747914,"visible":true,"origin":"","legend":"","description":"","filename":"SoundTwinAHighSimilarityFastDiffusionAutoregressiveSpeechCloningModelBasedonLocalGlobalFeatureFusion.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7512915/v1_covered_a3b855a3-e4a9-413b-bcee-d4999a8855d1.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"SoundTwin: A High-Similarity Fast Diffusion Autoregressive Speech Cloning Model Based on Local-Global Feature Fusion","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"language-resources-and-evaluation","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"lrev","sideBox":"Learn more about [Language Resources and Evaluation](http://link.springer.com/journal/10579)","snPcode":"10579","submissionUrl":"https://submission.nature.com/new-submission/10579/3","title":"Language Resources and Evaluation","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Text-to-Speech Synthesis, Zero-Shot Voice Cloning, Fast Diffusion Models, AutoRegressive Neural Networks, Speaker Embedding","lastPublishedDoi":"10.21203/rs.3.rs-7512915/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7512915/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eIn recent years, significant advances in deep learning have propelled text-to-speech (TTS) technology forward. However, there are still challenges in high-quality voice cloning with low-resource and low-latency conditions. Traditional autoregressive models suffer from high inference latency, while emerging diffusion models, although capable of generating high-fidelity speech, incur substantial computational overhead due to their multi-step sampling process. To mitigate these limitations, we propose SoundTwin, a novel speech synthesis framework integrating accelerated diffusion sampling with an autoregressive Transformer architecture. This approach significantly enhances synthesis efficiency without compromising speech naturalness. Furthermore, we design a Local-Global Squeeze-and-Excitation weighted Speaker embedding Network to efficiently extract fine-grained timbre features from limited reference audio, enabling rapid speaker adaptation. The model accepts target text, reference speech, and reference text as inputs, jointly modeling duration, pitch, and energy features to generate high-quality mel-spectrograms. Experimental results validate that our method achieves state-of-the-art speaker similarity and speech naturalness in zero-shot voice cloning tasks.\u003c/p\u003e","manuscriptTitle":"SoundTwin: A High-Similarity Fast Diffusion Autoregressive Speech Cloning Model Based on Local-Global Feature Fusion","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-11-27 23:24:41","doi":"10.21203/rs.3.rs-7512915/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"221008624321908156843094000975233218355","date":"2025-11-28T20:25:24+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-11-20T14:16:20+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-11-20T12:55:43+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-09-05T02:04:48+00:00","index":"","fulltext":""},{"type":"submitted","content":"Language Resources and Evaluation","date":"2025-09-02T03:35:52+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"language-resources-and-evaluation","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"lrev","sideBox":"Learn more about [Language Resources and Evaluation](http://link.springer.com/journal/10579)","snPcode":"10579","submissionUrl":"https://submission.nature.com/new-submission/10579/3","title":"Language Resources and Evaluation","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"3b5df42a-7bb4-4625-9e7a-caf73f6539c0","owner":[],"postedDate":"November 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-11-27T23:24:41+00:00","versionOfRecord":[],"versionCreatedAt":"2025-11-27 23:24:41","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7512915","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7512915","identity":"rs-7512915","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}