{"id":10746,"date":"2026-08-04T19:07:16","date_gmt":"2026-08-04T11:07:16","guid":{"rendered":"https:\/\/infernews.com\/blog\/we-propose-leaptalk-a-novel-framework-that-achieves-stable-and-real-time-talking\/"},"modified":"2026-08-04T19:39:14","modified_gmt":"2026-08-04T11:39:14","slug":"leaptalk-real-time-talking","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/leaptalk-real-time-talking\/","title":{"rendered":"LeapTalk\uff1a1 \u6b65\u751f\u6210\u6703\u8a71\u6578\u5b57\u4eba"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/pasted-6aab408e3288.jpg\" alt=\"teaser\"\/><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u60f3\u505a\u6703\u8a71\u6578\u5b57\u4eba\uff1f\u5361\u4f4d\u901a\u5e38\u5514\u4fc2\u80fd\u5426\u751f\u6210\uff0c\u800c\u4fc2\u5920\u5514\u5920\u5feb\u3001\u500b\u6a23\u6703\u5514\u6703\u6108\u64ad\u6108\u8d70\u6a23\u3002LeapTalk\u5c6c\u65bc\u958b\u6e90 talking-head generation \u9805\u76ee\uff0c\u91dd\u5c0d\u7684\u6b63\u662f\u5373\u6642\u4e32\u6d41\u3001\u4eba\u81c9\u4e00\u81f4\u6027\u540c\u5634\u578b\u540c\u6b65\u4e09\u4ef6\u4e8b\u4e00\u9f4a\u8981\u517c\u9867\u7684\u96e3\u984c\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b83\u7684\u5438\u5f15\u529b\u5728\u65bc\u6bcf\u500b\u5f71\u7247\u7247\u6bb5\u53ea\u7528 <strong>1 NFE<\/strong> \u5c31\u53ef\u63a8\u9032\uff0c\u5728 Lite \u8a2d\u5b9a\u4e0b\u6a19\u793a\u53ef\u9054 <strong>200 FPS<\/strong>\uff0c\u800c\u4e14 Web Demo \u5df2\u7d93\u505a\u6210\u53ef\u8f09\u5165\u4eba\u50cf\u5716\u3001\u7528\u6587\u5b57\u6216\u54aa\u9ad8\u5cf0\u5c0d\u8a71\u7684\u4ecb\u9762\u3002\u5e95\u5c64\u4e26\u975e\u55ae\u9760\u786c\u63a8\u901f\u5ea6\uff0c\u800c\u662f\u5efa\u57fa\u65bc <strong>SoulX-FlashHead-1_3B<\/strong>\uff0c\u518d\u914d\u5408 <strong>LoRA<\/strong>\u3001<code>audio_proj_step_*.pt<\/code>\u3001<code>wav2vec2-base-960h<\/code>\uff0c\u7528 audio-driven <strong>classifier-free guidance<\/strong> \u5f37\u5316\u5c0d\u5634\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>\u4ee5 chunk-by-chunk streaming pipeline \u9023\u7e8c\u751f\u6210\u9577\u7247<\/li>\n\n\n\n<li>\u7528 <strong>Brownian bridge<\/strong> \u540c <strong>Bridge Forcing<\/strong> \u6e1b\u5c11\u9577\u7247\u8eab\u4efd\u6f02\u79fb<\/li>\n\n\n\n<li>README \u5df2\u4ea4\u4ee3\u9700\u53e6\u5916\u4e0b\u8f09 base model\u3001\u8a9e\u97f3\u6a21\u578b\u8207 LeapTalk \u6b0a\u91cd<\/li>\n\n\n\n<li>\u63a8\u8ad6\u524d\u8981\u81ea\u884c\u586b\u5165 <code>COND_IMAGE<\/code>\u3001<code>AUDIO_PATH<\/code> \u7b49\u672c\u5730\u8def\u5f91<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u540c\u985e\u505a\u6cd5\u5e38\u898b\u53d6\u6368\u662f\u901f\u5ea6\u5feb\u4f46\u756b\u9762\u7d2f\u7a4d\u8aa4\u5dee\uff0c\u6216\u756b\u8cea\u7a69\u5b9a\u4f46\u5ef6\u9072\u9ad8\uff1bLeapTalk\u660e\u986f\u7ad9\u5728\u300c\u5148\u628a\u5ef6\u9072\u58d3\u5230\u53ef\u4e92\u52d5\u300d\u9019\u4e00\u908a\uff0c\u518d\u7528 reference-anchored data-to-data transport \u88dc\u56de\u7a69\u5b9a\u5ea6\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u73fe\u9069\u5408\u505a\u6578\u5b57\u4eba\u4e92\u52d5\u3001\u865b\u64ec\u4e3b\u64ad\u3001\u5373\u6642\u5167\u5bb9\u751f\u6210\u7684\u5718\u968a\u3002200 FPS \u8207\u55ae\u5f35 H200 GPU \u7684\u8aaa\u6cd5\u4f86\u81ea\u9805\u76ee\u63d0\u4f9b\u8cc7\u8a0a\uff0c\u666e\u901a\u786c\u4ef6\u80fd\u5426\u8907\u88fd\u540c\u6a23\u6548\u679c\u3001Web Demo \u7684\u5b8c\u6574\u90e8\u7f72\u7d30\u7bc0\u662f\u5426\u9f4a\u5099\uff0c\u4ecd\u8981\u7b49\u66f4\u591a\u5be6\u6e2c\u4f50\u8b49\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/zhangrongxiang.github.io\/leaptalk-page\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/zhangrongxiang\/LeapTalk\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u8b1b\u4eba\u50cf\u76f8\u7247\u540c\u8a9e\u97f3\uff0c\u5c31\u80fd\u9023\u7e8c\u751f\u6210\u5c0d\u5634\u5f71\u7247\u3002LeapTalk\u628a\u901f\u5ea6\u62c9\u5230\u63a5\u8fd1\u5373\u6642\uff0c\u540c\u6642\u76e1\u91cf\u5b88\u4f4f\u4eba\u81c9\u4e00\u81f4\u6027\u3002<\/p>\n","protected":false},"author":8,"featured_media":10745,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,29,162,120,126,141],"tags":[],"class_list":["post-10746","post","type-post","status-publish","format-standard","hentry","category-133","category-audio","category-ai-productions","category-120","category-126","category-141"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10746","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=10746"}],"version-history":[{"count":2,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10746\/revisions"}],"predecessor-version":[{"id":10753,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10746\/revisions\/10753"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/10745"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=10746"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=10746"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=10746"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}