{"id":8884,"date":"2026-06-06T16:38:13","date_gmt":"2026-06-06T08:38:13","guid":{"rendered":"https:\/\/infernews.com\/blog\/pilottts\/"},"modified":"2026-06-06T16:38:13","modified_gmt":"2026-06-06T08:38:13","slug":"pilottts","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/pilottts\/","title":{"rendered":"\u7528 20 \u842c\u5c0f\u6642\u8a13\u7df4\u7684\u958b\u6e90\u8a9e\u97f3\u5408\u6210\uff1aPilotTTS \u5f37\u5728\u54ea\uff1f"},"content":{"rendered":"<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/06\/Introduction-10ad4ebc2ce4.jpg\" alt=\"Introduction\"><\/figure>\n<p>PilotTTS \u662f\u9ad8\u5fb7\u5730\u5716\uff08Amap\uff09\u5718\u968a\u5728 GitHub \u4e0a\u958b\u6e90\u7684\u6587\u5b57\u8f49\u8a9e\u97f3\uff08text-to-speech, TTS\uff09\u9805\u76ee\uff0c\u4e3b\u6253\u4ee5 LLM-based \u67b6\u69cb\u914d\u5408\u56b4\u8b39\u7684\u8cc7\u6599\u5de5\u7a0b\uff0c\u7528\u8f03\u5c11\u8cc7\u6e90\u505a\u51fa\u5177\u7af6\u722d\u529b\u7684\u5408\u6210\u54c1\u8cea\u3002\u5c0d\u60f3\u7814\u7a76\u8a9e\u97f3\u751f\u6210\u3001\u537b\u6c92\u6709\u767e\u842c\u5c0f\u6642\u8cc7\u6599\u548c\u9f90\u5927\u7b97\u529b\u7684\u5718\u968a\u4f86\u8aaa\uff0c\u9019\u662f\u4e00\u500b\u503c\u5f97\u95dc\u6ce8\u7684\u9078\u64c7\u3002<\/p>\n<p>\u9019\u500b\u9805\u76ee\u8981\u89e3\u6c7a\u7684\u554f\u984c\u5f88\u76f4\u63a5\uff1a\u73fe\u4eca\u6700\u9802\u5c16\u7684 TTS \u7cfb\u7d71\u5f80\u5f80\u4f9d\u8cf4\u6578\u767e\u842c\u5c0f\u6642\u7684\u79c1\u6709\u8cc7\u6599\u548c\u8907\u96dc\u7684\u591a\u968e\u6bb5\u67b6\u69cb\uff0c\u9032\u5165\u9580\u6abb\u6975\u9ad8\u3002PilotTTS \u53cd\u5176\u9053\u800c\u884c\uff0c\u50c5\u4ee5 20 \u842c\u5c0f\u6642\u3001\u4ee5\u5168\u958b\u6e90\u5de5\u5177\u8655\u7406\u7684\u8cc7\u6599\u96c6\u9032\u884c\u8a13\u7df4\uff0c\u4e26\u91cb\u51fa\u5f9e\u54c1\u8cea\u8a55\u4f30\u3001\u6a19\u8a3b\u5230\u904e\u6ffe\u7684\u5b8c\u6574\u8cc7\u6599\u7ba1\u7dda\uff08data pipeline\uff09\uff0c\u8b93\u5176\u4ed6\u7814\u7a76\u8005\u80fd\u91cd\u73fe\u4e26\u6539\u826f\u3002<\/p>\n<p>\u5728\u529f\u80fd\u9762\u4e0a\uff0c\u9805\u76ee\u6db5\u84cb\u56db\u500b\u9762\u5411\uff1a\u96f6\u6a23\u672c\u8072\u97f3\u8907\u88fd\uff08zero-shot voice cloning\uff09\u300111 \u7a2e\u60c5\u7dd2\u5408\u6210\u30014 \u7a2e\u526f\u8a9e\u97f3\u6548\u679c\uff08\u5982\u7b11\u8072\u3001\u547c\u5438\u3001\u54b3\u55fd\u3001\u54ed\u8072\uff09\uff0c\u4ee5\u53ca 14 \u7a2e\u4e2d\u6587\u65b9\u8a00\u7684\u8de8\u65b9\u8a00\u5408\u6210\u3002\u6a21\u578b\u65b9\u9762\uff0c\u6b0a\u91cd\u5206\u70ba <code>pilot_tts.pt<\/code>\uff08\u57fa\u790e\u6a21\u578b\uff09\u8207 <code>pilot_tts_instruct.pt<\/code>\uff08\u6307\u4ee4\u63a7\u5236\u7248\u672c\uff09\uff0c\u53ef\u5f9e HuggingFace \u6216 ModelScope \u4e0b\u8f09\uff0c\u914d\u5408 <code>w2v-bert-2.0<\/code> \u7b49\u958b\u6e90\u7279\u5fb5\u63d0\u53d6\u5668\u5373\u53ef\u904b\u4f5c\u3002<\/p>\n<p>\u8a55\u4f30\u7d50\u679c\u65b9\u9762\uff0c\u5718\u968a\u5728 Seed-TTS Eval \u57fa\u6e96\u4e0a\u5831\u51fa\u4e86\u4eae\u773c\u6578\u5b57\uff1a\u82f1\u6587\u6e2c\u8a66\u96c6 WER 1.50%\u3001\u4e2d\u6587 CER 0.87%\uff0c\u5169\u7d44\u6e2c\u8a66\u7684\u8aaa\u8a71\u4eba\u76f8\u4f3c\u5ea6\uff08speaker similarity\uff09\u5206\u5225\u9054\u5230 0.862 \u8207 0.815\uff0c\u52dd\u904e\u591a\u500b\u4ee5\u66f4\u5927\u8cc7\u6599\u96c6\u8a13\u7df4\u7684\u7cfb\u7d71\u3002\u6a21\u578b\u63a1\u7528 Q-Former-based conditioning\uff0c\u900f\u904e\u8de8\u6a23\u672c\u914d\u5c0d\u8a13\u7df4\u628a\u8aaa\u8a71\u4eba\u8eab\u4efd\u8207\u8a9e\u6c23\u98a8\u683c\u89e3\u8026\uff0c\u9019\u662f\u5b83\u5728\u7cbe\u7c21\u67b6\u69cb\u4e0b\u4ecd\u80fd\u4fdd\u6301\u9ad8\u8868\u73fe\u7684\u95dc\u9375\u8a2d\u8a08\u4e4b\u4e00\u3002<\/p>\n<p>\u5c0d\u4e00\u822c\u8b80\u8005\u800c\u8a00\uff0c\u9019\u500b\u9805\u76ee\u8f03\u9069\u5408\u5f9e\u4e8b\u8a9e\u97f3\u5408\u6210\u3001LLM \u591a\u6a21\u614b\u61c9\u7528\u6216\u4e2d\u6587\u65b9\u8a00\u7814\u7a76\u7684\u958b\u767c\u8005\u8207\u5b78\u751f\uff1b\u5c0d\u60f3\u6253\u9020\u6709\u8072\u5167\u5bb9\u3001\u914d\u97f3\u5de5\u5177\u6216\u7121\u969c\u7919\u8a9e\u97f3\u4ecb\u9762\u7684\u7522\u54c1\u5718\u968a\uff0c\u5b83\u4e5f\u63d0\u4f9b\u4e86\u53ef\u76f4\u63a5\u6574\u5408\u7684\u958b\u6e90\u6b0a\u91cd\u8207\u63a8\u7406\u6d41\u7a0b\u3002<\/p>\n<p><strong>\u91cd\u9ede\u6458\u8981<\/strong><\/p>\n<ul>\n<li><strong>\u6975\u7c21\u67b6\u69cb<\/strong>\uff1aLLM-based \u81ea\u8ff4\u6b78\u6a21\u578b\uff0c\u4ee5 20 \u842c\u5c0f\u6642\u958b\u6e90\u8cc7\u6599\u8655\u7406\u5f8c\u7684\u8a13\u7df4\u96c6\u9054\u5230\u9802\u5c16\u57fa\u6e96\u6210\u7e3e\u3002<\/li>\n<li><strong>\u5b8c\u6574\u8cc7\u6599\u7ba1\u7dda<\/strong>\uff1a\u54c1\u8cea\u8a55\u4f30\u3001\u6a19\u8a3b\u3001\u904e\u6ffe\u5168\u90e8\u4f7f\u7528\u516c\u958b\u5de5\u5177\uff0c\u53ef\u91cd\u73fe\u4e14\u6210\u672c\u8f03\u4f4e\u3002<\/li>\n<li><strong>\u591a\u7dad\u5ea6\u63a7\u5236<\/strong>\uff1a\u652f\u63f4 11 \u7a2e\u60c5\u7dd2\u30014 \u7a2e\u526f\u8a9e\u97f3\u300114 \u7a2e\u4e2d\u6587\u65b9\u8a00\u7684\u8de8\u65b9\u8a00\u5408\u6210\u3002<\/li>\n<li><strong>\u9802\u5c16\u6307\u6a19<\/strong>\uff1aSeed-TTS Eval \u53d6\u5f97\u6700\u9ad8\u8aaa\u8a71\u4eba\u76f8\u4f3c\u5ea6\uff0c\u4e2d\u6587 CER \u50c5 0.87%\u3002<\/li>\n<li><strong>\u5b8c\u6574\u958b\u6e90<\/strong>\uff1a\u6a21\u578b\u6b0a\u91cd\u3001\u8655\u7406\u7ba1\u7dda\u8207\u7a0b\u5f0f\u78bc\u5747\u65bc GitHub\u3001HuggingFace\u3001ModelScope \u91cb\u51fa\u3002<\/li>\n<\/ul>\n<p><strong>GitHub\uff1a<\/strong> <a href=\"https:\/\/github.com\/AMAPVOICE\/PilotTTS\" rel=\"noopener noreferrer\">https:\/\/github.com\/AMAPVOICE\/PilotTTS<\/a><\/p>\n<p><strong>\u9805\u76ee\uff1a<\/strong> <a href=\"https:\/\/amapvoice.github.io\/PilotTTS\/\" rel=\"noopener noreferrer\">https:\/\/amapvoice.github.io\/PilotTTS\/<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u9ad8\u5fb7\u5730\u5716\u65d7\u4e0b\u5718\u968a\u767c\u4f48\u7684 LLM-based \u6587\u5b57\u8f49\u8a9e\u97f3\u9805\u76ee PilotTTS\uff0c\u4ee5\u7cbe\u7c21\u67b6\u69cb\u914d\u5408\u56b4\u8b39\u7684\u8cc7\u6599\u5de5\u7a0b\uff0c\u5728 Seed-TTS \u57fa\u6e96\u53d6\u5f97\u9802\u5c16\u6210\u7e3e\uff0c\u4e26\u5b8c\u6574\u958b\u6e90\u8cc7\u6599\u8655\u7406\u6d41\u7a0b\u8207\u6a21\u578b\u6b0a\u91cd\u3002<\/p>\n","protected":false},"author":8,"featured_media":8883,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","wpai_meta_description":"","footnotes":""},"categories":[133,171,76,126],"tags":[],"class_list":["post-8884","post","type-post","status-publish","format-standard","hentry","category-133","category-171","category-76","category-126"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8884","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=8884"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/8884\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/8883"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=8884"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=8884"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=8884"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}