{"id":12121,"date":"2026-09-10T23:37:17","date_gmt":"2026-09-10T15:37:17","guid":{"rendered":"https:\/\/infernews.com\/blog\/an-open-source-foundational-model-for-speech-generation-and-editing\/"},"modified":"2026-09-10T23:38:27","modified_gmt":"2026-09-10T15:38:27","slug":"an-open-source-foundational-model-for-speech-generation-and-editing","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/an-open-source-foundational-model-for-speech-generation-and-editing\/","title":{"rendered":"\u9a30\u8a0a\u6df7\u5143\u958b\u6e90 AuK\uff1a1.5B \u6a21\u578b\u7d71\u4e00\u8a9e\u97f3\u751f\u6210\u8207\u7de8\u8f2f"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/09\/performance.jpg\" alt=\"AuK performance across speech generation, editing, enhancement, and separation benchmarks\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u8a9e\u97f3\u6a21\u578b\u4e00\u76f4\u5b58\u5728\u4e00\u500b\u5c37\u5c2c\uff1a\u96f6\u6a23\u672c\u5408\u6210\u3001\u8a9e\u97f3\u514b\u9686\u3001\u97f3\u8272\u66ff\u63db\u3001\u5206\u96e2\u3001\u964d\u566a\u5f80\u5f80\u662f\u5e7e\u5957\u7368\u7acb\u7cfb\u7d71\uff0c\u8981\u4e32\u8d77\u4f86\u5c31\u5f97\u5806 pipeline\u3002\u9a30\u8a0a Hunyuan \u958b\u6e90\u7684 AuK \u60f3\u6253\u7834\u9019\u500b\u5206\u5de5\uff0c\u4ee5 1.5B \u53c3\u6578\u7684\u57fa\u790e\u6a21\u578b\u70ba\u6838\u5fc3\uff0c\u8b93\u4e00\u53e5\u81ea\u7136\u8a9e\u8a00\u6307\u4ee4\u76f4\u63a5\u5c0d\u61c9\u5230\u7de8\u8f2f\u5f8c\u6216\u751f\u6210\u7684\u97f3\u8a0a\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">AuK \u7531\u4e09\u90e8\u5206\u7d44\u6210\uff1a\u8ca0\u8cac\u8a9e\u7fa9\u689d\u4ef6\u7684\u591a\u6a21\u614b\u8a9e\u8a00\u6a21\u578b\u3001\u63d0\u4f9b\u8072\u5b78\u6f5b\u5728\u7a7a\u9593\u7684 50 Hz VAE\uff0c\u4ee5\u53ca\u4e00\u500b\u6df7\u5408 rectified-flow Transformer\uff0c\u7528\u96d9\u6d41 MMDiT \u584a\u878d\u5408\u5169\u7a2e\u689d\u4ef6\u5f8c\u518d\u505a\u55ae\u6d41 DiT \u751f\u6210\u3002\u8a13\u7df4\u6578\u64da\u898f\u6a21\u76f8\u7576\u53ef\u89c0\u2014\u2014\u7d04 30.3 \u5104\u689d\u6307\u4ee4-\u97f3\u8a0a\u914d\u5c0d\uff0c\u52a0\u4e0a\u7d04 195 \u842c\u5c0f\u6642\u7684\u6709\u6548\u76e3\u7763\uff0c\u8986\u84cb\u4e94\u5927\u4efb\u52d9\u65cf\uff1a\u8a9e\u97f3\u751f\u6210\u3001\u5167\u5bb9\u7de8\u8f2f\u3001\u589e\u5f37\u8207\u5206\u96e2\u3001\u526f\u8a9e\u8a00\u8cc7\u8a0a\u7de8\u8f2f\u3001\u8072\u5b78\u7de8\u8f2f\u3002<\/p>\n\n\n<div\n    \tclass=\"align wp-block-vpb-video\"    id='vpbpVideoPlayer-1'\n    data-attributes='{&quot;source&quot;:&quot;https:\\\/\\\/auk-project.github.io\\\/assets\\\/video\\\/auk-overview.mp4&quot;,&quot;repeat&quot;:true,&quot;autoplay&quot;:true,&quot;muted&quot;:true,&quot;align&quot;:&quot;&quot;,&quot;poster&quot;:&quot;&quot;,&quot;controls&quot;:{&quot;play-large&quot;:true,&quot;restart&quot;:false,&quot;rewind&quot;:true,&quot;play&quot;:true,&quot;fast-forward&quot;:true,&quot;progress&quot;:true,&quot;current-time&quot;:true,&quot;duration&quot;:false,&quot;mute&quot;:true,&quot;volume&quot;:true,&quot;pip&quot;:false,&quot;airplay&quot;:false,&quot;captions&quot;:false,&quot;settings&quot;:true,&quot;download&quot;:false,&quot;fullscreen&quot;:true},&quot;captions&quot;:[],&quot;ratio&quot;:&quot;16:9&quot;,&quot;accentColor&quot;:&quot;&quot;,&quot;lazyLoad&quot;:false,&quot;width&quot;:&quot;100%&quot;,&quot;radius&quot;:&quot;0px&quot;,&quot;resetOnEnd&quot;:false,&quot;autoHideControl&quot;:true,&quot;isSetup&quot;:false}'\n><\/div>\n\n\n<p class=\"wp-block-paragraph\">\u540c\u6b65\u91cb\u51fa\u7684 AuK-Flash \u8d70\u901f\u5ea6\u8def\u7dda\uff1a\u900f\u904e\u4e00\u81f4\u6027\u521d\u59cb\u5316\u52a0\u4e0a\u4efb\u52d9\u8def\u7531\u7684 Decoupled DMD \u84b8\u993e\uff0c\u505a\u5230 NFE=4\u3001\u7121 CFG \u7684 4 \u6b65\u63a8\u8ad6\uff0c\u5b98\u65b9\u6307\u5728\u5339\u914d\u689d\u4ef6\u4e0b\u5c0d\u6bd4\u5b8c\u6574 AuK \u6709\u7d04 4.5 \u500d wall-clock \u52a0\u901f\uff0c\u8cea\u91cf\u63a5\u8fd1\u6559\u5e2b\u6a21\u578b\u3002\u9019\u500b\u5206\u5c64\u7b56\u7565\u5c0d\u9700\u8981\u5373\u6642\u8a9e\u97f3\u751f\u6210\u7684\u61c9\u7528\u5834\u666f\uff08\u5982\u5c0d\u8a71 agent\u3001\u5be6\u6642\u914d\u97f3\uff09\u6709\u76f4\u63a5\u610f\u7fa9\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u90e8\u7f72\u5c64\u9762\uff0c\u5b98\u65b9\u540c\u6642\u63d0\u4f9b Hugging Face Space\u3001ModelScope Space\u3001Gradio \u4e92\u52d5\u4ecb\u9762\u3001ComfyUI \u7bc0\u9ede\u3001Python API\uff0c\u4ee5\u53ca uv \u548c Conda \u5169\u7a2e\u4f9d\u8cf4\u7ba1\u7406\u65b9\u5f0f\uff0c\u4e26\u5df2\u7372 SGLang-Omni Day 0 \u652f\u63f4\u3002\u5c0d\u65bc\u672c\u5730\u8cc7\u6e90\u6709\u9650\u7684\u5718\u968a\uff0cAuK-Flash \u914d\u5408 SGLang-Omni \u662f\u8f03\u5408\u7406\u7684\u5207\u5165\u9ede\uff1b\u5982\u679c\u8ffd\u6c42\u6700\u9ad8\u8cea\u91cf\u3001\u4e14\u4e0d\u5728\u610f\u63a8\u8ad6\u5ef6\u9072\uff0c\u5247\u53ef\u9078 AuK Base\uff0c\u4e26\u900f\u904e\u53ef\u914d\u7f6e\u7684 NFE \u8207 CFG \u505a\u54c1\u8cea\/\u901f\u5ea6\u53d6\u6368\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>\u91cd\u9ede\u6458\u8981\uff1a<\/strong><\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>1.5B \u7d71\u4e00\u6a21\u578b<\/strong>\uff1a\u7528\u81ea\u7136\u8a9e\u8a00\u6307\u4ee4\u7d71\u4e00\u96f6\u6a23\u672c TTS\u3001\u8a9e\u97f3\u7de8\u8f2f\u3001\u5206\u96e2\u3001\u589e\u5f37\u3001\u526f\u8a9e\u8a00\u8207\u8072\u5b78\u7de8\u8f2f\u3002<\/li>\n\n\n\n<li><strong>\u4e09\u6a21\u7d44\u67b6\u69cb<\/strong>\uff1a\u591a\u6a21\u614b\u8a9e\u8a00\u6a21\u578b + 50 Hz \u97f3\u8a0a VAE + \u6df7\u5408 rectified-flow Transformer\uff0c\u63a1\u96d9\u6d41 MMDiT \u63a5\u55ae\u6d41 DiT\u3002<\/li>\n\n\n\n<li><strong>AuK-Flash \u84b8\u993e<\/strong>\uff1a\u4e00\u81f4\u6027\u521d\u59cb\u5316 + \u4efb\u52d9\u8def\u7531 DMD\uff0c4 \u6b65\u63a8\u8ad6\u7121\u9700 CFG\uff0cwall-clock \u52a0\u901f\u7d04 4.5 \u500d\u3002<\/li>\n\n\n\n<li><strong>\u5f8c\u8a13\u7df4\u7b56\u7565<\/strong>\uff1a\u8a9e\u97f3\u751f\u6210\u7528\u734e\u52f5\u5f37\u5316\u5b78\u7fd2\uff0c\u958b\u653e\u5f0f\u7de8\u8f2f\u7528\u4eba\u985e\u504f\u597d\u512a\u5316\u3002<\/li>\n\n\n\n<li><strong>\u90e8\u7f72\u652f\u63f4\u5b8c\u6574<\/strong>\uff1aHugging Face\u3001ModelScope\u3001Gradio\u3001ComfyUI\u3001Python API\u3001SGLang-Omni \u9f4a\u5099\u3002<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/auk-project.github.io\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/Tencent-Hunyuan\/AuK\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/huggingface.co\/tencent\/AuK\" rel=\"noopener noreferrer\"><strong>\u6a21\u578b<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u9a30\u8a0a\u6df7\u5143\u628a\u96f6\u6a23\u672c TTS\u3001\u8a9e\u97f3\u7de8\u8f2f\u3001\u5206\u96e2\u8207\u589e\u5f37\u6536\u9032\u540c\u4e00\u500b\u81ea\u7136\u8a9e\u8a00\u6307\u4ee4\u4ecb\u9762\uff0c\u4e26\u540c\u6b65\u91cb\u51fa\u8ffd\u6c42\u901f\u5ea6\u7684\u84b8\u993e\u7248\u672c AuK-Flash\u3002<\/p>\n","protected":false},"author":8,"featured_media":12120,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"wpai_generated_summary":"","footnotes":""},"categories":[133,172,37,125,116,76,119,127,137,29,152,128],"tags":[],"class_list":["post-12121","post","type-post","status-publish","format-standard","hentry","category-133","category-172","category-37","category-comfyui","category-agentic","category-76","category-119","category-127","category-api","category-audio","category-python","category-128"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12121","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/comments?post=12121"}],"version-history":[{"count":1,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12121\/revisions"}],"predecessor-version":[{"id":12123,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/12121\/revisions\/12123"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/12120"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=12121"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=12121"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=12121"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}