[{"data":1,"prerenderedAt":374},["ShallowReactive",2],{"navigation":3,"\u002Fconcepts\u002Flatency":224,"\u002Fconcepts\u002Flatency-surround":369},[4,24,60,88,120,152,196],{"title":5,"path":6,"stem":7,"children":8,"icon":23},"Start","\u002Fstart","1.start\u002F1.index",[9,11,15,19],{"title":10,"path":6,"stem":7},"Overview",{"title":12,"path":13,"stem":14},"Quickstart","\u002Fstart\u002Fquickstart","1.start\u002F2.quickstart",{"title":16,"path":17,"stem":18},"Authentication","\u002Fstart\u002Fauthentication","1.start\u002F3.authentication",{"title":20,"path":21,"stem":22},"Choose an API","\u002Fstart\u002Fchoose-an-api","1.start\u002F5.choose-an-api","i-lucide-rocket",{"title":25,"path":26,"stem":27,"children":28,"icon":59},"Realtime","\u002Frealtime","2.realtime\u002F1.index",[29,31,35,39,43,47,51,55],{"title":30,"path":26,"stem":27},"Realtime Overview",{"title":32,"path":33,"stem":34},"Realtime Quickstart","\u002Frealtime\u002Fquickstart","2.realtime\u002F2.quickstart",{"title":36,"path":37,"stem":38},"Connection Lifecycle","\u002Frealtime\u002Fconnection-lifecycle","2.realtime\u002F3.connection-lifecycle",{"title":40,"path":41,"stem":42},"Audio Input","\u002Frealtime\u002Faudio-input","2.realtime\u002F4.audio-input",{"title":44,"path":45,"stem":46},"Transcript Lifecycle","\u002Frealtime\u002Ftranscript-lifecycle","2.realtime\u002F5.transcript-lifecycle",{"title":48,"path":49,"stem":50},"Realtime Translation","\u002Frealtime\u002Frealtime-translation","2.realtime\u002F6.realtime-translation",{"title":52,"path":53,"stem":54},"Session Configuration","\u002Frealtime\u002Fsession-configuration","2.realtime\u002F7.session-configuration",{"title":56,"path":57,"stem":58},"Handling Interruptions & Silence","\u002Frealtime\u002Finterruptions-silence","2.realtime\u002F8.interruptions-silence","i-lucide-radio",{"title":61,"path":62,"stem":63,"children":64,"icon":87},"Recorded","\u002Frecorded","3.recorded\u002F1.index",[65,67,71,75,79,83],{"title":66,"path":62,"stem":63},"Recorded Overview",{"title":68,"path":69,"stem":70},"Transcribe Audio","\u002Frecorded\u002Ftranscribe-audio","3.recorded\u002F2.transcribe-audio",{"title":72,"path":73,"stem":74},"Timestamps & Speakers","\u002Frecorded\u002Ftimestamps-speakers","3.recorded\u002F3.timestamps-speakers",{"title":76,"path":77,"stem":78},"Subtitles","\u002Frecorded\u002Fsubtitles","3.recorded\u002F4.subtitles",{"title":80,"path":81,"stem":82},"Recorded Translation","\u002Frecorded\u002Ftranslation","3.recorded\u002F5.translation",{"title":84,"path":85,"stem":86},"Structured Results","\u002Frecorded\u002Fstructured-results","3.recorded\u002F6.structured-results","i-lucide-file-audio",{"title":89,"path":90,"stem":91,"children":92,"icon":119},"Concepts","\u002Fconcepts","4.concepts\u002F1.index",[93,95,99,103,107,111,115],{"title":94,"path":90,"stem":91},"Voice Context Layer",{"title":96,"path":97,"stem":98},"Stable vs. Partial Text","\u002Fconcepts\u002Fstable-vs-partial-text","4.concepts\u002F2.stable-vs-partial-text",{"title":100,"path":101,"stem":102},"StableStream","\u002Fconcepts\u002Fstablestream","4.concepts\u002F3.stablestream",{"title":104,"path":105,"stem":106},"Context-Aware Processing","\u002Fconcepts\u002Fcontext-aware-processing","4.concepts\u002F4.context-aware-processing",{"title":108,"path":109,"stem":110},"Understanding Latency","\u002Fconcepts\u002Flatency","4.concepts\u002F5.latency",{"title":112,"path":113,"stem":114},"Live Translation","\u002Fconcepts\u002Flive-translation","4.concepts\u002F6.live-translation",{"title":116,"path":117,"stem":118},"Trinity Engine","\u002Fconcepts\u002Ftrinity-engine","4.concepts\u002F7.trinity-engine","i-lucide-brain",{"title":121,"path":122,"stem":123,"children":124,"icon":151},"Guides","\u002Fguides","5.guides\u002F1.index",[125,127,131,135,139,143,147],{"title":126,"path":122,"stem":123},"Browser Live Captions",{"title":128,"path":129,"stem":130},"Server-side Streaming","\u002Fguides\u002Fserver-side-streaming","5.guides\u002F2.server-side-streaming",{"title":132,"path":133,"stem":134},"Build a Stable Caption UI","\u002Fguides\u002Fstable-caption-ui","5.guides\u002F3.stable-caption-ui",{"title":136,"path":137,"stem":138},"Build Live Translation","\u002Fguides\u002Flive-translation","5.guides\u002F4.live-translation",{"title":140,"path":141,"stem":142},"Save a Transcript","\u002Fguides\u002Fsave-transcript","5.guides\u002F5.save-transcript",{"title":144,"path":145,"stem":146},"Reconnect a Live Session","\u002Fguides\u002Freconnection","5.guides\u002F6.reconnection",{"title":148,"path":149,"stem":150},"Client-VAD speech segments","\u002Fguides\u002Fclient-vad-segments","5.guides\u002F7.client-vad-segments","i-lucide-book-open",{"title":153,"path":154,"stem":155,"children":156,"icon":195},"Reference","\u002Fapi-reference","6.api-reference\u002F1.index",[157,159,163,167,171,175,179,183,187,191],{"title":158,"path":154,"stem":155},"API Reference Overview",{"title":160,"path":161,"stem":162},"Admin API","\u002Fapi-reference\u002Fadmin","6.api-reference\u002F10.admin",{"title":164,"path":165,"stem":166},"Realtime API","\u002Fapi-reference\u002Frealtime-api","6.api-reference\u002F2.realtime-api",{"title":168,"path":169,"stem":170},"Client Messages","\u002Fapi-reference\u002Fclient-messages","6.api-reference\u002F3.client-messages",{"title":172,"path":173,"stem":174},"Server Events","\u002Fapi-reference\u002Fserver-events","6.api-reference\u002F4.server-events",{"title":176,"path":177,"stem":178},"Segment Transcription API","\u002Fapi-reference\u002Fsegment-transcription","6.api-reference\u002F5.segment-transcription",{"title":180,"path":181,"stem":182},"Batch Transcription Jobs","\u002Fapi-reference\u002Fbatch-jobs","6.api-reference\u002F6.batch-jobs",{"title":184,"path":185,"stem":186},"Languages & Models","\u002Fapi-reference\u002Flanguages","6.api-reference\u002F7.languages",{"title":188,"path":189,"stem":190},"Errors","\u002Fapi-reference\u002Ferrors","6.api-reference\u002F8.errors",{"title":192,"path":193,"stem":194},"Rate Limits","\u002Fapi-reference\u002Frate-limits","6.api-reference\u002F9.rate-limits","i-lucide-code",{"title":197,"path":198,"stem":199,"children":200,"icon":223},"Production","\u002Fproduction","7.production\u002F1.index",[201,203,207,211,215,219],{"title":202,"path":198,"stem":199},"Production Checklist",{"title":204,"path":205,"stem":206},"Reconnection & Retries","\u002Fproduction\u002Freconnection-retries","7.production\u002F2.reconnection-retries",{"title":208,"path":209,"stem":210},"Latency Best Practices","\u002Fproduction\u002Flatency-best-practices","7.production\u002F3.latency-best-practices",{"title":212,"path":213,"stem":214},"Security & Privacy","\u002Fproduction\u002Fsecurity-privacy","7.production\u002F4.security-privacy",{"title":216,"path":217,"stem":218},"Data Retention","\u002Fproduction\u002Fdata-retention","7.production\u002F5.data-retention",{"title":220,"path":221,"stem":222},"Changelog","\u002Fproduction\u002Fchangelog","7.production\u002F6.changelog","i-lucide-server",{"id":225,"title":108,"body":226,"description":361,"draft":362,"extension":363,"links":364,"meta":365,"navigation":366,"path":109,"seo":367,"status":364,"stem":110,"__hash__":368},"docs\u002F4.concepts\u002F5.latency.md",{"type":227,"value":228,"toc":352},"minimark",[229,233,236,241,244,255,258,262,265,268,271,275,278,281,284,290,293,301,305,308,311,314,317,321,324,327,349],[230,231,232],"p",{},"There is no single latency number for a live speech system.",[230,234,235],{},"What users experience as \"latency\" is the result of several different stages.",[237,238,240],"h2",{"id":239},"the-latency-pipeline","The latency pipeline",[230,242,243],{},"A simplified real-time pipeline looks like:",[245,246,252],"pre",{"className":247,"code":249,"language":250,"meta":251},[248],"language-text","speaker\n  ↓\naudio capture\n  ↓\nnetwork transport\n  ↓\nspeech recognition\n  ↓\ncontext processing\n  ↓\noptional translation\n  ↓\napplication rendering\n","text","",[253,254,249],"code",{"__ignoreMap":251},[230,256,257],{},"Each stage contributes to the final experience.",[237,259,261],{"id":260},"first-result-latency","First-result latency",[230,263,264],{},"The time between incoming speech and the first usable recognition result.",[230,266,267],{},"Lower first-result latency generally makes an interface feel more responsive.",[230,269,270],{},"However, extremely early results may contain greater uncertainty.",[237,272,274],{"id":273},"stabilization-latency","Stabilization latency",[230,276,277],{},"The time required before evolving speech becomes sufficiently stable for a particular use.",[230,279,280],{},"This is different from first-result latency.",[230,282,283],{},"For example:",[245,285,288],{"className":286,"code":287,"language":250,"meta":251},[248],"300 ms → first interpretation appears\n900 ms → surrounding context resolves ambiguity\n1.2 s → segment becomes stable\n",[253,289,287],{"__ignoreMap":251},[230,291,292],{},"These numbers are illustrative only.",[230,294,295,296,300],{},"The important point is that ",[297,298,299],"strong",{},"responsiveness and stability are different measurements",".",[237,302,304],{"id":303},"translation-latency","Translation latency",[230,306,307],{},"Live translation introduces another dependency.",[230,309,310],{},"Translation quality improves when more linguistic context is available, while live experiences require output before the full conversation is known.",[230,312,313],{},"This creates another latency-quality tradeoff.",[230,315,316],{},"LansonAI therefore treats live translation as a streaming context problem rather than simply translating a completed transcript.",[237,318,320],{"id":319},"measure-the-user-experience","Measure the user experience",[230,322,323],{},"For live applications, useful latency measurements should reflect what the user actually experiences.",[230,325,326],{},"Depending on the application, this may include:",[328,329,330,334,337,340,343,346],"ul",{},[331,332,333],"li",{},"time to first readable text",[331,335,336],{},"time to stable text",[331,338,339],{},"time to translated text",[331,341,342],{},"correction frequency",[331,344,345],{},"visible reflow",[331,347,348],{},"end-to-end interaction latency",[230,350,351],{},"Optimizing only one number can make another part of the experience worse.",{"title":251,"searchDepth":353,"depth":354,"links":355},1,2,[356,357,358,359,360],{"id":239,"depth":354,"text":240},{"id":260,"depth":354,"text":261},{"id":273,"depth":354,"text":274},{"id":303,"depth":354,"text":304},{"id":319,"depth":354,"text":320},"There is no single latency number for a live speech system. What users experience as latency is the result of several different stages.",false,"md",null,{},true,{"title":108,"description":361},"4g_-0v-DcTRsDZwBLSJwlKZX6Sb91t6VezlT2Lr7zoc",[370,372],{"title":104,"path":105,"stem":106,"description":371,"children":-1},"Speech is ambiguous when interpreted in isolation. Context is part of recognition, not a post-processing step.",{"title":112,"path":113,"stem":114,"description":373,"children":-1},"Live translation is different from translating a finished transcript. It must balance responsiveness, linguistic context, translation quality, and stability.",1789433608271]