{"id":10419,"date":"2026-07-26T15:20:06","date_gmt":"2026-07-26T07:20:06","guid":{"rendered":"https:\/\/infernews.com\/blog\/verbatim-automatic-speech-recognition-with-improved-word-level-timestamps-and-fi\/"},"modified":"2026-07-26T15:20:06","modified_gmt":"2026-07-26T07:20:06","slug":"verbatim-automatic-speech-recognition-with-improved-word-level-timestamps-and-fi","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/verbatim-automatic-speech-recognition-with-improved-word-level-timestamps-and-fi\/","title":{"rendered":"CrisperWhisper \u628a\u8a9e\u97f3\u9010\u5b57\u7a3f\u8b8a\u6210\u53ef\u63a7\u5236\u8f38\u51fa"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/07\/pasted-d2bf99fd13b6.jpg\" alt=\"Repository image for nyrahealth\/CrisperWhisper\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u505a\u6703\u8b70\u7d00\u9304\u3001\u8a2a\u8ac7\u6574\u7406\u6216\u81e8\u5e8a\u8a9e\u97f3\u5206\u6790\u6642\uff0c\u6700\u5927\u843d\u5dee\u5f80\u5f80\u5514\u4fc2\u8fa8\u8b58\u5230\u5e7e\u591a\u5b57\uff0c\u800c\u4fc2\u7cfb\u7d71\u7a76\u7adf\u5beb\u51fa\u300c\u8b1b\u5497\u4e5c\u300d\u5b9a\u300c\u672c\u4f86\u60f3\u8868\u9054\u4e5c\u300d\u3002CrisperWhisper \u5c6c\u65bc\u958b\u6e90\u8a9e\u97f3\u8fa8\u8b58\u6a21\u578b\u9805\u76ee\uff0c\u6838\u5fc3\u50f9\u503c\u4fc2\u5c07 verbatim \u540c intended \u5169\u7a2e\u8f49\u9304\u6a21\u5f0f\u8b8a\u6210\u53ef\u660e\u78ba\u63a7\u5236\u7684\u8f38\u51fa\uff0c\u4ee4\u9010\u5b57\u7a3f\u5514\u518d\u53d7\u8a13\u7df4\u8cc7\u6599\u98a8\u683c\u5de6\u53f3\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5462\u500b\u53d6\u5411\u540c\u4e00\u822c speech-to-text \u7cfb\u7d71\u597d\u5514\u540c\u3002\u5e38\u898b\u505a\u6cd5\u6703\u4e0d\u4e00\u81f4\u5730\u522a\u8d70 filler\u3001\u91cd\u8907\u3001\u505c\u9813\u540c cut-off\uff0cCrisperWhisper 2.0 \u5c31\u523b\u610f\u4fdd\u7559\u5462\u5572\u8a9e\u97f3\u7d30\u7bc0\uff0c\u6216\u8005\u6309\u9700\u8981\u8f38\u51fa\u6574\u7406\u5f8c\u7248\u672c\uff1b\u540c\u4e00\u6bb5\u9304\u97f3\u53ef\u4ee5\u5f97\u51fa\u5169\u4efd\u7528\u9014\u5b8c\u5168\u4e0d\u540c\u5605\u6587\u672c\u3002\u5c0d\u505a TTS \u8cc7\u6599\u6574\u7406\u3001\u91ab\u7642\u6216\u7814\u7a76\u8a2a\u8ac7\u5206\u6790\u3001\u9700\u8981\u7cbe\u6e96\u5b57\u5e55\u6642\u9593\u78bc\u5605\u5718\u968a\uff0c\u5462\u7a2e\u5206\u6d41\u6bd4\u55ae\u7d14\u8ffd\u6c42\u53ef\u8b80\u6027\u66f4\u6709\u7528\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u9805\u76ee\u53e6\u4e00\u500b\u95dc\u9375\u4f4d\u4fc2\u6642\u9593\u5c0d\u9f4a\u540c\u9577\u97f3\u8a0a\u8655\u7406\u3002\u5b83\u63d0\u4f9b word-level timings\uff0c\u8b80\u7a3f\u8a9e\u97f3\u5e73\u5747\u908a\u754c\u8aa4\u5dee\u7d04 30 ms\uff0c\u5c0d\u8a71\u8a9e\u97f3\u7d04 41 ms\uff1b\u9577\u97f3\u8a0a\u5247\u7528 conditional continuation \u907f\u514d\u4e00\u822c\u5206\u6bb5\u8f49\u9304\u5e38\u898b\u5605\u91cd\u8907\u6f0f\u5b57\u3002README \u4ea6\u63d0\u5230\u63a8\u7406\u7aef\u57fa\u65bc CTranslate2\uff0c\u914d\u5408 speculative decoding\uff0c\u540c\u6642\u6e1b\u8f15 Whisper \u5e38\u898b looping-hallucination \u554f\u984c\uff0c\u65b9\u5411\u660e\u986f\u4fc2\u671d\u4f4f production inference\u3002<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>verbatim \u8207 intended \u5169\u7a2e\u6a21\u5f0f\u5206\u958b\u63a7\u5236\uff0c\u9069\u5408\u540c\u4e00\u9304\u97f3\u5c0d\u61c9\u4e0d\u540c\u5de5\u4f5c\u6d41<\/li>\n<li>\u652f\u63f4 multilingual\uff0c\u8986\u84cb\u591a\u6578 Whisper \u652f\u63f4\u8a9e\u8a00<\/li>\n<li>\u53ef\u7528 Verbatimize \u4f9d\u64da\u97f3\u8a0a\u52a0\u4e7e\u6de8\u6587\u672c\u88dc\u56de\u771f\u5be6\u8a9e\u6c23\u8a5e\u8207\u53e3\u8aa4<\/li>\n<li>\u9577\u97f3\u8a0a\u8f49\u9304\u8457\u91cd\u9023\u7e8c\u6027\uff0c\u6e1b\u5c11 chunk \u908a\u754c\u9020\u6210\u5605\u932f\u6f0f<\/li>\n<li>Nyra Verbatim Speech Benchmark \u4ee5 disfluency F1 \u7b49\u6307\u6a19\u8861\u91cf\u4fdd\u771f\u80fd\u529b<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">\u5b89\u88dd\u8207\u90e8\u7f72\u65b9\u5411\u76f8\u5c0d\u6e05\u6670\uff1a\u6a21\u578b\u53ef\u7d93 PyPI\u3001Hugging Face \u540c\u6587\u4ef6\u4f7f\u7528\uff0c\u63a8\u7406\u8def\u7dda\u570d\u7e5e CTranslate2 runtime\uff0c\u800c\u5514\u4fc2\u53ea\u505c\u7559\u55ba\u7814\u7a76\u5c55\u793a\u3002\u8981\u7559\u610f\u5605\u53d6\u6368\u4ea6\u597d\u76f4\u63a5\uff0c\u7576\u4f60\u8981\u7684\u662f\u53ef\u8b80\u3001\u53ef\u767c\u5e03\u6587\u672c\uff0cintended \u6a21\u5f0f\u66f4\u5408\u9069\uff1b\u7576\u4f60\u8981\u4fdd\u7559\u7336\u8c6b\u3001\u91cd\u8907\u3001\u7b11\u8072\u540c\u8a9e\u97f3\u4e8b\u4ef6\uff0cverbatim \u6a21\u5f0f\u5148\u771f\u6b63\u767c\u63ee\u50f9\u503c\u3002\u5462\u500b\u9805\u76ee\u5514\u4fc2\u55ae\u9760\u66f4\u9ad8 WER \u6210\u7e3e\u53bb\u5438\u5f15\u4eba\uff0c\u800c\u4fc2\u91cd\u65b0\u754c\u5b9a\u9010\u5b57\u7a3f\u61c9\u5426\u5fe0\u65bc\u8aaa\u8a71\u8868\u9762\u5f62\u5f0f\uff0c\u4e26\u4e14\u7528 benchmark \u5c07\u5462\u4ef6\u4e8b\u91cf\u5316\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/www.nyra-labs.com\/\" rel=\"noopener noreferrer\"><strong>\u9805\u76ee\u4e3b\u9801<\/strong><\/a> \u00b7 <a href=\"https:\/\/github.com\/nyrahealth\/CrisperWhisper\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/huggingface.co\/nyralabs\/CrisperWhisper2.0_large_pro\" rel=\"noopener noreferrer\"><strong>\u6a21\u578b<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u540c\u4e00\u6bb5\u9304\u97f3\uff0c\u53ef\u4ee5\u8981\u9010\u5b57\u4fdd\u7559\u505c\u9813\u8207\u53e3\u5403\uff0c\u4e5f\u53ef\u4ee5\u8981\u4e7e\u6de8\u53ef\u8b80\u7248\u672c\u3002CrisperWhisper \u628a\u8a9e\u97f3\u8fa8\u8b58\u6700\u5e38\u6df7\u5728\u4e00\u8d77\u7684\u5169\u7a2e\u9700\u6c42\u5206\u958b\u8655\u7406\u3002<\/p>\n","protected":false},"author":8,"featured_media":10418,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,179,37,144,76,128,199],"tags":[],"class_list":["post-10419","post","type-post","status-publish","format-standard","hentry","category-133","category-nvidia","category-37","category-medical","category-76","category-128","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10419","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=10419"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10419\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/10418"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=10419"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=10419"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=10419"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}