{"id":10967,"date":"2026-08-12T03:00:04","date_gmt":"2026-08-11T19:00:04","guid":{"rendered":"https:\/\/infernews.com\/blog\/nemospeech-framework\/"},"modified":"2026-08-12T03:00:04","modified_gmt":"2026-08-11T19:00:04","slug":"nemospeech-framework","status":"publish","type":"post","link":"https:\/\/infernews.com\/blog\/nemospeech-framework\/","title":{"rendered":"NeMo Speech\uff1aNVIDIA \u628a ASR\u3001TTS\u3001\u8a9e\u97f3 LLM \u6536\u9032\u540c\u4e00\u689d PyTorch \u751f\u7522\u7dda"},"content":{"rendered":"\n<figure class=\"wp-block-image size-large\"><img decoding=\"async\" src=\"https:\/\/infernews.com\/blog\/wp-content\/uploads\/2026\/08\/pasted-aef3a28d0c21.jpg\" alt=\"Repository image for NVIDIA-NeMo\/Speech\"><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">\u8a9e\u97f3 AI \u7684\u75db\u9ede\u5f80\u5f80\u4e0d\u662f\u6a21\u578b\u4e0d\u5920\u5f37\uff0c\u800c\u662f\u958b\u767c\u8005\u8981\u540c\u6642\u9762\u5c0d ASR\u3001TTS\u3001\u4e32\u6d41\u8b58\u5225\u7b49\u597d\u5e7e\u5957\u7368\u7acb\u5de5\u5177\u93c8\u3002NVIDIA NeMo Speech \u628a\u9019\u4e9b\u4efb\u52d9\u6536\u9032\u540c\u4e00\u500b PyTorch \u6846\u67b6\uff0c\u4e26\u63d0\u4f9b\u9810\u8a13\u7df4\u6b0a\u91cd\uff0c\u8b93\u7814\u7a76\u54e1\u53ef\u4ee5\u628a\u7cbe\u529b\u82b1\u5728\u5be6\u9a57\u8a2d\u8a08\uff0c\u800c\u4e0d\u662f\u5f9e\u982d\u642d\u5efa\u8a13\u7df4\u6d41\u7a0b\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u5f9e\u8fd1\u671f\u66f4\u65b0\u53ef\u4ee5\u770b\u5230\u4e09\u500b\u503c\u5f97\u7559\u610f\u7684\u65b9\u5411\uff1aMagpieTTS v2607 \u628a\u652f\u63f4\u8a9e\u8a00\u64f4\u5c55\u5230 12 \u7a2e\uff0c\u65b0\u589e\u963f\u62c9\u4f2f\u6587\u3001\u97d3\u6587\u3001\u8461\u8404\u7259\u6587\uff1bNemotron-3.5-ASR-Streaming-0.6B \u5728\u55ae\u4e00 H100 \u4e0a\u80fd\u540c\u6642\u8655\u7406\u6700\u591a 2400 \u689d\u4e32\u6d41\uff0c\u4e26\u5141\u8a31\u628a\u5ef6\u9072\u63a7\u5236\u5728 80ms \u5230 1s \u4e4b\u9593\uff1bParakeet-unified-en-0.6b \u5247\u628a\u96e2\u7dda\u8207\u4e32\u6d41\u63a8\u7406\u5408\u4f75\u6210\u4e00\u500b\u82f1\u6587\u6a21\u578b\uff0c\u6700\u77ed\u5ef6\u9072 160ms\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Nemotron 3 VoiceChat \u628a LLM\u3001\u9aa8\u5e79\u8207 TTS \u89e3\u78bc\u5668\u4e32\u6210\u5168\u96d9\u5de5\u5c0d\u8a71\uff0c\u80fd\u81ea\u7136\u8655\u7406\u6253\u65b7\u8207\u63d2\u8a71\uff0c\u9019\u5c0d\u65bc\u60f3\u5efa\u7acb\u8a9e\u97f3\u52a9\u7406\u7684\u5718\u968a\u662f\u6bd4\u8f03\u5b8c\u6574\u7684\u4e00\u689d\u8def\u3002Fastconformer \u7b49\u5feb\u53d6\u611f\u77e5\u67b6\u69cb\u662f\u80cc\u5f8c\u7684\u5de5\u7a0b\u529f\u81e3\uff0c\u8b93\u9577\u97f3\u8a0a\u4e32\u6d41\u4e0d\u9700\u8981\u72a7\u7272\u592a\u591a\u541e\u5410\u91cf\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u8a13\u7df4\u968e\u6bb5\u5fc5\u9808\u914d\u5099 NVIDIA GPU \u8207 CUDA \u74b0\u5883\uff0c\u63a8\u85a6\u4f7f\u7528 PyTorch 2.7 \u6216\u4ee5\u4e0a\u7248\u672c\uff1b\u73fe\u6642\u5009\u5eab\u6b63\u9032\u884c\u62c6\u5206\uff0c\u4e0b\u4e00\u500b\u4e3b\u8981\u7248\u672c\u9810\u5b9a 2026 \u5e74 6 \u6708\u767c\u4f48\uff0c\u77ed\u671f\u5167\u7a69\u5b9a\u4f7f\u7528\u53ef\u4ee5\u8003\u616e 26.02 NGC container\u3002\u5c0d\u505a\u5ba2\u670d\u3001\u6703\u8b70\u8a18\u9304\u3001\u5a92\u9ad4\u5b57\u5e55\u6216\u6709\u8072\u66f8\u751f\u6210\u7684\u5718\u968a\uff0c\u9019\u5957\u6846\u67b6\u80fd\u628a\u8a9e\u97f3\u6a21\u578b\u5f9e\u539f\u578b\u8d70\u5230\u90e8\u7f72\u7684\u8ddd\u96e2\u660e\u986f\u7e2e\u77ed\u3002<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u91cd\u9ede\u6458\u8981\uff1a<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>\u55ae\u4e00\u6846\u67b6\u8986\u84cb\u4e09\u5927\u4efb\u52d9<\/strong>\uff1aASR\u3001TTS \u8207 Speech LLM \u90fd\u5728 NeMo Speech \u5167\uff0c\u6e1b\u5c11\u5207\u63db\u5de5\u5177\u93c8\u7684\u6210\u672c\u3002<\/li>\n<li><strong>\u4e32\u6d41\u6548\u80fd\u7a81\u51fa<\/strong>\uff1aNemotron-3.5-ASR-Streaming-0.6B \u652f\u63f4 40 \u7a2e\u8a9e\u8a00\uff0c\u55ae\u5f35 H100 \u53ef\u4e26\u884c 2400 \u689d\u6d41\uff0c\u5ef6\u9072\u53ef\u8abf\u3002<\/li>\n<li><strong>\u591a\u8a9e TTS \u64f4\u5f35<\/strong>\uff1aMagpieTTS v2607 \u8986\u84cb 12 \u7a2e\u8a9e\u8a00\uff0c\u4e26\u63d0\u4f9b Hugging Face \u7dda\u4e0a demo\u3002<\/li>\n<li><strong>\u5168\u96d9\u5de5\u8a9e\u97f3\u52a9\u7406<\/strong>\uff1aNemotron 3 VoiceChat \u628a LLM \u8207 TTS \u89e3\u78bc\u5668\u7d50\u5408\uff0c\u652f\u63f4\u81ea\u7136\u6253\u65b7\u8207\u4f4e\u5ef6\u9072\u5c0d\u8a71\u3002<\/li>\n<li><strong>\u786c\u9ad4\u8981\u6c42\u660e\u78ba<\/strong>\uff1a\u81f3\u5c11\u914d\u5099 80 GB \u8a18\u61b6\u9ad4\u3002\u8a13\u7df4\u9700 NVIDIA GPU \u8207 CUDA\uff0cPyTorch 2.7 \u6216\u4ee5\u4e0a\u7248\u672c\uff0c\u63a8\u7406\u53ef\u5728 CPU \u6216 GPU \u57f7\u884c\u3002<\/li>\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/github.com\/NVIDIA-NeMo\/Speech\" rel=\"noopener noreferrer\"><strong>GitHub<\/strong><\/a> \u00b7 <a href=\"https:\/\/huggingface.co\/nvidia\/NVIDIA-NemotronLabs-VoiceChat-11B\" rel=\"noopener noreferrer\"><strong>\u6a21\u578b<\/strong><\/a><\/p>\n","protected":false},"excerpt":{"rendered":"<p>NVIDIA \u628a\u8a9e\u97f3\u7814\u7a76\u6700\u5e38\u78b0\u5230\u7684 ASR\u3001TTS \u8207 Speech LLM \u6574\u5408\u6210\u55ae\u4e00\u6846\u67b6\uff0c\u7814\u7a76\u54e1\u548c\u5de5\u7a0b\u5e2b\u4e0d\u7528\u518d\u6771\u62fc\u897f\u6e4a\uff0c\u4e5f\u80fd\u7528\u9810\u8a13\u7df4\u6b0a\u91cd\u5feb\u901f\u5fae\u8abf\u8207\u90e8\u7f72\u3002<\/p>\n","protected":false},"author":8,"featured_media":10966,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ai_generated_summary":"","footnotes":""},"categories":[133,179,37,116,152,128,197,199],"tags":[],"class_list":["post-10967","post","type-post","status-publish","format-standard","hentry","category-133","category-nvidia","category-37","category-agentic","category-python","category-128","category-framework","category-dataset-"],"_links":{"self":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10967","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=10967"}],"version-history":[{"count":0,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/posts\/10967\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media\/10966"}],"wp:attachment":[{"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/media?parent=10967"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/categories?post=10967"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/infernews.com\/blog\/wp-json\/wp\/v2\/tags?post=10967"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}