{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-23T06:40:50.084Z","headline":"通义实验室发布 Qwen-Audio-3.0-TTS 实时语音合成模型","description":"通义实验室发布 Qwen-Audio-3.0-TTS，含 Flash（首包延迟约300ms）和 Plus 两个版本。Plus 版本在 Artificial Analysis 榜单夺冠，支持16种语言和20种中文方言，平均 WER/CER 低至3.87（Flash），说话人相似度最高达82.75（Plus）。","url":"https://www.aioga.com/news/cmrt0856a09v9bitl62e4r17b/","mainEntityOfPage":"https://www.aioga.com/news/cmrt0856a09v9bitl62e4r17b/","datePublished":"2026-07-20T08:53:11.000Z","dateModified":"2026-07-20T08:53:11.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://mp.weixin.qq.com/s/INvrqTrWLMm2WCLIqhqTrg","https://aihot.virxact.com/items/cmrt0856a09v9bitl62e4r17b"],"canonicalUrl":"https://www.aioga.com/news/cmrt0856a09v9bitl62e4r17b/","directAnswer":{"@type":"Answer","text":"通义实验室发布 Qwen-Audio-3.0-TTS，提供 Flash 与 Plus 两个版本。材料称 Flash 首包延迟约300ms，Plus 在 Artificial Analysis 榜单夺冠。","url":"https://www.aioga.com/news/cmrt0856a09v9bitl62e4r17b/","dateCreated":"2026-07-20T08:53:11.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"公众号：通义实验室（千问） source article","url":"https://mp.weixin.qq.com/s/INvrqTrWLMm2WCLIqhqTrg","datePublished":"2026-07-20T08:53:11.000Z","provider":{"@type":"Organization","name":"公众号：通义实验室（千问）","url":"https://mp.weixin.qq.com/s/INvrqTrWLMm2WCLIqhqTrg"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrt0856a09v9bitl62e4r17b","datePublished":"2026-07-20T08:53:11.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrt0856a09v9bitl62e4r17b"}}],"aggregationSource":"公众号：通义实验室（千问）","originalPublisher":{"name":"公众号：通义实验室（千问）","url":"https://mp.weixin.qq.com/s/INvrqTrWLMm2WCLIqhqTrg"},"article":{"id":"cmrt0856a09v9bitl62e4r17b","slug":"cmrt0856a09v9bitl62e4r17b","url":"https://www.aioga.com/news/cmrt0856a09v9bitl62e4r17b/","title":"通义实验室发布 Qwen-Audio-3.0-TTS 实时语音合成模型","title_en":"从\"能说话\"到\"会表达\"：Qwen-Audio-3.0-TTS 发布","summary":"通义实验室发布 Qwen-Audio-3.0-TTS，含 Flash（首包延迟约300ms）和 Plus 两个版本。Plus 版本在 Artificial Analysis 榜单夺冠，支持16种语言和20种中文方言，平均 WER/CER 低至3.87（Flash），说话人相似度最高达82.75（Plus）。","source":"公众号：通义实验室（千问）","sourceUrl":"https://mp.weixin.qq.com/s/INvrqTrWLMm2WCLIqhqTrg","aiHotUrl":"https://aihot.virxact.com/items/cmrt0856a09v9bitl62e4r17b","publishedAt":"2026-07-20T08:53:11.000Z","category":"模型更新","score":70,"selected":true,"articleBody":["通义实验室发布 Qwen-Audio-3.0-TTS，含 Flash（首包延迟约300ms）和 Plus 两个版本。","Plus 版本在 Artificial Analysis 榜单夺冠，支持16种语言和20种中文方言，平均 WER/CER 低至3.87（Flash），说话人相似度最高达82.75（Plus）。"],"articleImages":[],"mediaStatus":"none","articleBodyZh":[],"translationStatus":"","bodyOrigin":"summary-fallback","editorial":{"summary":"通义实验室发布 Qwen-Audio-3.0-TTS，提供 Flash 与 Plus 两个版本。材料称 Flash 首包延迟约300ms，Plus 在 Artificial Analysis 榜单夺冠。","background":"公开材料将该产品归为实时语音合成模型，并披露其支持16种语言和20种中文方言。现有摘录未进一步说明发布时间、训练方法、部署条件或榜单评测规则。","viewpoint":"Aioga 判断，双版本设计体现了对响应速度与合成表现两类指标的区分。榜单成绩和延迟数据值得关注，但仍需结合完整评测口径与实际应用测试审慎理解。","implications":"该模型可能为多语言、中文方言及实时语音合成场景提供新的选择。材料给出的平均 WER/CER 低至3.87和相似度最高82.75，可作为初步比较依据，但不代表所有场景表现。","nextStep":"建议后续关注官方是否披露完整模型文档、Artificial Analysis 评测设置、WER/CER 计算口径、说话人相似度测试条件，以及 Flash 与 Plus 在不同语言和方言中的具体表现。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-07-23T00:29:50.217Z","sourceHash":"bfccf902071c40e2","review":{"approved":true,"groundedness":96,"clarity":92,"duplicationRisk":24,"blockingIssues":[],"notes":["“双版本设计体现了对响应速度与合成表现两类指标的区分”属于合理推断，且已用“Aioga 判断”明确标为观点。","候选内容已对榜单成绩、延迟和效果指标补充评测口径与适用范围限制，避免将官方披露的数据泛化为所有场景的实际表现。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["模型更新","公众号：通义实验室（千问）"],"translations":{"zh-CN":{"title":"通义实验室发布 Qwen-Audio-3.0-TTS 实时语音合成模型","summary":"通义实验室发布 Qwen-Audio-3.0-TTS，含 Flash（首包延迟约300ms）和 Plus 两个版本。Plus 版本在 Artificial Analysis 榜单夺冠，支持16种语言和20种中文方言，平均 WER/CER 低至3.87（Flash），说话人相似度最高达82.75（Plus）。","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"通义实验室发布 Qwen-Audio-3.0-TTS 实时语音合成模型 - Aioga AI资讯","description":"通义实验室发布 Qwen-Audio-3.0-TTS，含 Flash（首包延迟约300ms）和 Plus 两个版本。Plus 版本在 Artificial Analysis 榜单夺冠，支持16种语言和20种中文方言，平均 WER/CER 低至3.87（Flash），说话人相似度最高达82.75（Plus）。","url":"https://www.aioga.com/news/cmrt0856a09v9bitl62e4r17b/"},"en":{"title":"Tongyi Lab releases Qwen-Audio-3.0-TTS real-time speech synthesis model","summary":"Tongyi Laboratory released Qwen-Audio-3.0-TTS, which includes two versions: Flash (with a first-packet delay of about 300ms) and Plus. The Plus version topped the Artificial Analysis rankings, supports 16 languages and 20 Chinese dialects, with an average WER/CER as low as 3.87 (Flash), and a speaker similarity of up to 82.75 (Plus).","category":"Models","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Tongyi Lab releases Qwen-Audio-3.0-TTS real-time speech synthesis model - Aioga AI News","description":"Tongyi Laboratory released Qwen-Audio-3.0-TTS, which includes two versions: Flash (with a first-packet delay of about 300ms) and Plus. The Plus version topped the Artificial Analys...","url":"https://www.aioga.com/en/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:25:14.140Z"},"ja":{"title":"通義実験室が Qwen-Audio-3.0-TTS リアルタイム音声合成モデルを公開","summary":"通義実験室は Qwen-Audio-3.0-TTS を発表し、Flash（初回パック遅延約300ms）と Plus の2つのバージョンを含む。Plus バージョンは Artificial Analysis ランキングで首位を獲得し、16言語と20の中国語方言をサポート、平均 WER/CER は Flash で3.87まで低下、話者の類似度は Plus で最高82.75に達する。","category":"モデル更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"通義実験室が Qwen-Audio-3.0-TTS リアルタイム音声合成モデルを公開 - Aioga AIニュース","description":"通義実験室は Qwen-Audio-3.0-TTS を発表し、Flash（初回パック遅延約300ms）と Plus の2つのバージョンを含む。Plus バージョンは Artificial Analysis ランキングで首位を獲得し、16言語と20の中国語方言をサポート、平均 WER/CER は Flash で3.87まで低下、話者の類似度は Plus で最高...","url":"https://www.aioga.com/ja/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:25:34.173Z"},"ko":{"title":"통의 실험실, Qwen-Audio-3.0-TTS 실시간 음성 합성 모델 발표","summary":"통의 실험실에서 Qwen-Audio-3.0-TTS를 발표했으며, Flash(첫 패킷 지연 약 300ms)와 Plus 두 가지 버전이 있습니다. Plus 버전은 Artificial Analysis 순위에서 1위를 차지했으며, 16개 언어와 20개 중국어 방언을 지원하고, 평균 WER/CER는 Flash 기준 3.87까지 낮으며, 화자 유사도는 Plus 기준 최고 82.75에 달합니다.","category":"모델 업데이트","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"통의 실험실, Qwen-Audio-3.0-TTS 실시간 음성 합성 모델 발표 - Aioga AI 뉴스","description":"통의 실험실에서 Qwen-Audio-3.0-TTS를 발표했으며, Flash(첫 패킷 지연 약 300ms)와 Plus 두 가지 버전이 있습니다. Plus 버전은 Artificial Analysis 순위에서 1위를 차지했으며, 16개 언어와 20개 중국어 방언을 지원하고, 평균 WER/CER는 Flash 기준 3.87까지 낮...","url":"https://www.aioga.com/ko/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:26:18.914Z"},"es":{"title":"El Laboratorio Tongyi lanza el modelo de síntesis de voz en tiempo real Qwen-Audio-3.0-TTS","summary":"Tongyi Laboratory lanzó Qwen-Audio-3.0-TTS, con dos versiones: Flash (con un retraso inicial de aproximadamente 300 ms) y Plus. La versión Plus lidera la lista de Artificial Analysis, soporta 16 idiomas y 20 dialectos chinos, con un WER/CER promedio tan bajo como 3.87 (Flash), y la similitud de orador más alta alcanza 82.75 (Plus).","category":"Modelos","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"El Laboratorio Tongyi lanza el modelo de síntesis de voz en tiempo real Qwen-Audio-3.0-TTS - Aioga Noticias de IA","description":"Tongyi Laboratory lanzó Qwen-Audio-3.0-TTS, con dos versiones: Flash (con un retraso inicial de aproximadamente 300 ms) y Plus. La versión Plus lidera la lista de Artificial Analys...","url":"https://www.aioga.com/es/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:26:15.634Z"},"fr":{"title":"Le laboratoire Tongyi a publié le modèle de synthèse vocale en temps réel Qwen-Audio-3.0-TTS","summary":"Le laboratoire Tongyi a publié Qwen-Audio-3.0-TTS, disponible en deux versions : Flash (le premier paquet a un délai d'environ 300 ms) et Plus. La version Plus a remporté le premier prix du classement Artificial Analysis, prend en charge 16 langues et 20 dialectes chinois, avec un WER/CER moyen aussi bas que 3,87 (Flash), et une similitude vocale maximale atteignant 82,75 (Plus).","category":"Modèles","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Le laboratoire Tongyi a publié le modèle de synthèse vocale en temps réel Qwen-Audio-3.0-TTS - Aioga Actualités IA","description":"Le laboratoire Tongyi a publié Qwen-Audio-3.0-TTS, disponible en deux versions : Flash (le premier paquet a un délai d'environ 300 ms) et Plus. La version Plus a remporté le premie...","url":"https://www.aioga.com/fr/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:27:08.476Z"},"de":{"title":"Tongyi Laboratory veröffentlicht das Echtzeit-Sprachsynthesemodell Qwen-Audio-3.0-TTS","summary":"Tongyi Laboratory hat Qwen-Audio-3.0-TTS veröffentlicht, mit zwei Versionen: Flash (erste Paketzustellung Verzögerung ca. 300ms) und Plus. Die Plus-Version erreichte den ersten Platz auf der Artificial Analysis Rangliste, unterstützt 16 Sprachen und 20 chinesische Dialekte, mit durchschnittlichen WER/CER bis zu 3,87 (Flash), und die Sprecherähnlichkeit erreicht bis zu 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Tongyi Laboratory veröffentlicht das Echtzeit-Sprachsynthesemodell Qwen-Audio-3.0-TTS - Aioga KI-News","description":"Tongyi Laboratory hat Qwen-Audio-3.0-TTS veröffentlicht, mit zwei Versionen: Flash (erste Paketzustellung Verzögerung ca. 300ms) und Plus. Die Plus-Version erreichte den ersten Pla...","url":"https://www.aioga.com/de/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:27:00.484Z"},"pt-BR":{"title":"O Laboratório Tongyi lançou o modelo de síntese de voz em tempo real Qwen-Audio-3.0-TTS","summary":"O Laboratório Tongyi lançou o Qwen-Audio-3.0-TTS, contendo duas versões: Flash (latência inicial de cerca de 300ms) e Plus. A versão Plus liderou o ranking Artificial Analysis, suporta 16 idiomas e 20 dialetos chineses, com WER/CER médio tão baixo quanto 3,87 (Flash), e a similaridade do locutor alcançando até 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"O Laboratório Tongyi lançou o modelo de síntese de voz em tempo real Qwen-Audio-3.0-TTS - Aioga Notícias de IA","description":"O Laboratório Tongyi lançou o Qwen-Audio-3.0-TTS, contendo duas versões: Flash (latência inicial de cerca de 300ms) e Plus. A versão Plus liderou o ranking Artificial Analysis, sup...","url":"https://www.aioga.com/pt-BR/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:27:50.337Z"},"ru":{"title":"Лаборатория Tongyi выпустила модель синтеза речи в реальном времени Qwen-Audio-3.0-TTS","summary":"Лаборатория Tongyi выпустила Qwen-Audio-3.0-TTS, включающий версии Flash (первый пакет задержки около 300 мс) и Plus. Версия Plus заняла первое место в рейтинге Artificial Analysis, поддерживает 16 языков и 20 китайских диалектов, средний WER/CER достигает всего 3,87 (Flash), а максимальное сходство с говорящим достигает 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Лаборатория Tongyi выпустила модель синтеза речи в реальном времени Qwen-Audio-3.0-TTS - Aioga Новости ИИ","description":"Лаборатория Tongyi выпустила Qwen-Audio-3.0-TTS, включающий версии Flash (первый пакет задержки около 300 мс) и Plus. Версия Plus заняла первое место в рейтинге Artificial Analysis...","url":"https://www.aioga.com/ru/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:27:56.103Z"},"ar":{"title":"أصدرت مختبرات تونغي نموذج التوليد الصوتي الفوري Qwen-Audio-3.0-TTS","summary":"أصدرت مختبرات تونغي نموذج Qwen-Audio-3.0-TTS، ويحتوي على نسختين: Flash (تأخير الحزمة الأولى حوالي 300 مللي ثانية) وPlus. فازت نسخة Plus بالمركز الأول في قائمة التحليل الصناعي، وتدعم 16 لغة و20 لهجة صينية، بمتوسط معدل خطأ الكلمات/الحروف منخفض يصل إلى 3.87 (Flash)، وتشابه المتحدثين يصل إلى 82.75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"أصدرت مختبرات تونغي نموذج التوليد الصوتي الفوري Qwen-Audio-3.0-TTS - Aioga أخبار الذكاء الاصطناعي","description":"أصدرت مختبرات تونغي نموذج Qwen-Audio-3.0-TTS، ويحتوي على نسختين: Flash (تأخير الحزمة الأولى حوالي 300 مللي ثانية) وPlus. فازت نسخة Plus بالمركز الأول في قائمة التحليل الصناعي، وتدع...","url":"https://www.aioga.com/ar/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:28:43.217Z"},"hi":{"title":"通义实验室 ने Qwen-Audio-3.0-TTS रीयल-टाइम वॉइस सिंथेसिस मॉडल जारी किया","summary":"通义实验室 ने Qwen-Audio-3.0-TTS जारी किया, जिसमें Flash (पहली पैकेज की विलंबता लगभग 300ms) और Plus दो संस्करण शामिल हैं। Plus संस्करण Artificial Analysis रैंकिंग में शीर्ष स्थान पर रहा, यह 16 भाषाओं और 20 चीनी बोलियों का समर्थन करता है, औसत WER/CER 3.87 तक कम है (Flash), और वक्ता समानता अधिकतम 82.75 तक है (Plus)।","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"通义实验室 ने Qwen-Audio-3.0-TTS रीयल-टाइम वॉइस सिंथेसिस मॉडल जारी किया - Aioga AI समाचार","description":"通义实验室 ने Qwen-Audio-3.0-TTS जारी किया, जिसमें Flash (पहली पैकेज की विलंबता लगभग 300ms) और Plus दो संस्करण शामिल हैं। Plus संस्करण Artificial Analysis रैंकिंग में शीर्ष स्थान पर रहा...","url":"https://www.aioga.com/hi/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:28:53.528Z"},"it":{"title":"Il laboratorio Tongyi rilascia il modello di sintesi vocale in tempo reale Qwen-Audio-3.0-TTS","summary":"Il laboratorio Tongyi ha rilasciato Qwen-Audio-3.0-TTS, disponibile in due versioni: Flash (primo pacchetto con un ritardo di circa 300 ms) e Plus. La versione Plus ha conquistato il primo posto nella classifica di Artificial Analysis, supporta 16 lingue e 20 dialetti cinesi, con un WER/CER medio fino a 3,87 (Flash) e una somiglianza del parlante fino all'82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Il laboratorio Tongyi rilascia il modello di sintesi vocale in tempo reale Qwen-Audio-3.0-TTS - Aioga Notizie IA","description":"Il laboratorio Tongyi ha rilasciato Qwen-Audio-3.0-TTS, disponibile in due versioni: Flash (primo pacchetto con un ritardo di circa 300 ms) e Plus. La versione Plus ha conquistato...","url":"https://www.aioga.com/it/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:29:48.839Z"},"nl":{"title":"Tongyi Lab brengt Qwen-Audio-3.0-TTS realtime spraaksynthesemodel uit","summary":"Tongyi Laboratory heeft Qwen-Audio-3.0-TTS uitgebracht, inclusief twee versies: Flash (eerste pakketvertraging ongeveer 300 ms) en Plus. De Plus-versie behaalde de eerste plaats op de Artificial Analysis-lijst, ondersteunt 16 talen en 20 Chinese dialecten, met een gemiddelde WER/CER van slechts 3,87 (Flash), en een sprekersgelijkenis die opliep tot 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Tongyi Lab brengt Qwen-Audio-3.0-TTS realtime spraaksynthesemodel uit - Aioga AI-nieuws","description":"Tongyi Laboratory heeft Qwen-Audio-3.0-TTS uitgebracht, inclusief twee versies: Flash (eerste pakketvertraging ongeveer 300 ms) en Plus. De Plus-versie behaalde de eerste plaats op...","url":"https://www.aioga.com/nl/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:29:38.264Z"},"tr":{"title":"Tongyi Laboratuvarı, Qwen-Audio-3.0-TTS gerçek zamanlı ses sentezleme modelini yayınladı","summary":"Tongyi Laboratuvarı, Flash (ilk paket gecikmesi yaklaşık 300 ms) ve Plus olmak üzere iki sürüm içeren Qwen-Audio-3.0-TTS'yi yayınladı. Plus sürümü Artificial Analysis listesinin zirvesine yerleşti, 16 dil ve 20 Mandarin lehçesini destekliyor, ortalama WER/CER Flash'ta 3,87 kadar düşük, konuşmacı benzerliği en yüksek 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Tongyi Laboratuvarı, Qwen-Audio-3.0-TTS gerçek zamanlı ses sentezleme modelini yayınladı - Aioga AI Haberleri","description":"Tongyi Laboratuvarı, Flash (ilk paket gecikmesi yaklaşık 300 ms) ve Plus olmak üzere iki sürüm içeren Qwen-Audio-3.0-TTS'yi yayınladı. Plus sürümü Artificial Analysis listesinin zi...","url":"https://www.aioga.com/tr/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:30:44.781Z"},"vi":{"title":"Phòng thí nghiệm Tongyi phát hành mô hình tổng hợp giọng nói thời gian thực Qwen-Audio-3.0-TTS","summary":"Phòng thí nghiệm Tongyi ra mắt Qwen-Audio-3.0-TTS, gồm hai phiên bản Flash (gói đầu tiên độ trễ khoảng 300ms) và Plus. Phiên bản Plus dẫn đầu bảng xếp hạng Artificial Analysis, hỗ trợ 16 ngôn ngữ và 20 phương ngữ tiếng Trung, trung bình WER/CER thấp tới 3,87 (Flash), độ tương đồng giọng nói cao nhất đạt 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Phòng thí nghiệm Tongyi phát hành mô hình tổng hợp giọng nói thời gian thực Qwen-Audio-3.0-TTS - Tin tức AI Aioga","description":"Phòng thí nghiệm Tongyi ra mắt Qwen-Audio-3.0-TTS, gồm hai phiên bản Flash (gói đầu tiên độ trễ khoảng 300ms) và Plus. Phiên bản Plus dẫn đầu bảng xếp hạng Artificial Analysis, hỗ...","url":"https://www.aioga.com/vi/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:30:41.462Z"},"id":{"title":"Tongyi Laboratory merilis model sintesis suara waktu nyata Qwen-Audio-3.0-TTS","summary":"Tongyi Laboratory merilis Qwen-Audio-3.0-TTS, termasuk dua versi, Flash (paket pertama dengan delay sekitar 300ms) dan Plus. Versi Plus memimpin di peringkat Artificial Analysis, mendukung 16 bahasa dan 20 dialek Tionghoa, dengan rata-rata WER/CER serendah 3,87 (Flash), dan tingkat kesamaan pembicara hingga 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Tongyi Laboratory merilis model sintesis suara waktu nyata Qwen-Audio-3.0-TTS - Berita AI Aioga","description":"Tongyi Laboratory merilis Qwen-Audio-3.0-TTS, termasuk dua versi, Flash (paket pertama dengan delay sekitar 300ms) dan Plus. Versi Plus memimpin di peringkat Artificial Analysis, m...","url":"https://www.aioga.com/id/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:31:30.128Z"},"th":{"title":"ห้องแล็บ Tongyi เปิดตัวโมเดลสังเคราะห์เสียงพูดแบบเรียลไทม์ Qwen-Audio-3.0-TTS","summary":"Tongyi Laboratory เปิดตัว Qwen-Audio-3.0-TTS มีสองรุ่นคือ Flash (แพ็กเกจแรกหน่วงประมาณ 300ms) และ Plus รุ่น Plus ครองอันดับหนึ่งในรายการ Artificial Analysis รองรับ 16 ภาษาและ 20 สำเนียงภาษาจีน ค่า WER/CER เฉลี่ยต่ำสุดถึง 3.87 (Flash) และความคล้ายคลึงของผู้พูดสูงสุดถึง 82.75 (Plus)","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"ห้องแล็บ Tongyi เปิดตัวโมเดลสังเคราะห์เสียงพูดแบบเรียลไทม์ Qwen-Audio-3.0-TTS - ข่าว AI Aioga","description":"Tongyi Laboratory เปิดตัว Qwen-Audio-3.0-TTS มีสองรุ่นคือ Flash (แพ็กเกจแรกหน่วงประมาณ 300ms) และ Plus รุ่น Plus ครองอันดับหนึ่งในรายการ Artificial Analysis รองรับ 16 ภาษาและ 20 สำ...","url":"https://www.aioga.com/th/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:31:37.258Z"},"pl":{"title":"Laboratorium Tongyi opublikowało model syntezy mowy w czasie rzeczywistym Qwen-Audio-3.0-TTS","summary":"Tongyi Laboratory wydało Qwen-Audio-3.0-TTS, zawierający dwie wersje: Flash (pierwsza paczka opóźnienia około 300ms) i Plus. Wersja Plus zdobyła pierwsze miejsce w rankingu Artificial Analysis, obsługuje 16 języków i 20 chińskich dialektów, średni WER/CER wynosi tylko 3,87 (Flash), a podobieństwo głosu mówcy osiąga maksymalnie 82,75 (Plus).","category":"模型更新","source":"公众号：通义实验室（千问）","aggregationSource":"公众号：通义实验室（千问）","pageTitle":"Laboratorium Tongyi opublikowało model syntezy mowy w czasie rzeczywistym Qwen-Audio-3.0-TTS - Aioga Wiadomości AI","description":"Tongyi Laboratory wydało Qwen-Audio-3.0-TTS, zawierający dwie wersje: Flash (pierwsza paczka opóźnienia około 300ms) i Plus. Wersja Plus zdobyła pierwsze miejsce w rankingu Artific...","url":"https://www.aioga.com/pl/news/cmrt0856a09v9bitl62e4r17b/","contentTranslated":true,"sourceHash":"8b55e07fd6ebfc1a","translatedAt":"2026-07-22T17:32:23.425Z"}}}}