대화형 에이전트(Dialogflow CX) 에이전트에서 제한 시간 기반 고급 음성 끝 감도를 사용 설정한 경우 런타임에서 설정이 true로 설정되더라도 single_utterance가 false라고 가정하여 모델 식별자가 라우팅됩니다.
다른 설정 조합의 경우 추가 안내가 있을 때까지 자동 마이그레이션이 보류 상태로 유지됩니다(ETA 없음).
phone_call 모델 식별자를 계속 사용할 수 있지만 Conformer 기반 음성 모델을 선택하는 것이 좋습니다.
마이그레이션 개요
특정 언어 태그 및 음성 설정 조합을 사용하면 오디오가 포함된 대화형 에이전트(Dialogflow CX) 및 Dialogflow ES 트래픽이 phone_call 모델 식별자 뒤쪽의 기본 Speech-to-Text 모델에서 새로운 Conformer 기반 음성 모델로 점진적으로 라우팅됩니다. 리디렉션이 내부 전용이므로 현재 모델 식별자와 코드는 유효하게 유지되며 트래픽을 계속 처리합니다.
영향을 받는 언어 태그는 en-au, en-gb, de, de-de, es, es-es, es-us, fr, fr-ca, fr-fr, it, it-it, ja, pt-br입니다.
다음 표에는 이전이 적용될 때 발생하는 Speech-to-Text phone_call 모델의 라우팅이 나와 있습니다. 구성이 언어 태그 및 테이블의 처음 세 열과 일치하면 새 모델 식별자가 네 번째 열에 나열됩니다.
요청 수준 음성 모델 재정의 설정을 사용하는 경우를 제외하고 런타임에서 API에 지정된 음성 모델이 대화형 에이전트(Dialogflow CX) 콘솔에 지정된 음성 모델보다 우선 적용됩니다. 음성 모델의 우선순위에 대해 자세히 알아보세요.
파트너 전화 통합의 경우 음성 모델이 API에 설정되는지 전화 서비스 제공업체에 문의하세요. 대화형 에이전트(Dialogflow CX) Phone Gateway를 사용하는 경우 음성 모델 제한사항을 검토합니다.
프로젝트를 사전에 마이그레이션하면 충분한 시간을 갖고 모델을 테스트하고 일찍 개선된 정확성과 견고성을 활용할 수 있습니다. 대부분의 통합에는 single_utterance = true가 필요하므로 고급 제한 시간 기반 음성 종료 민감도를 사용하지 않는다면 latest_short를 사용하는 경우가 많습니다.
[[["이해하기 쉬움","easyToUnderstand","thumb-up"],["문제가 해결됨","solvedMyProblem","thumb-up"],["기타","otherUp","thumb-up"]],[["이해하기 어려움","hardToUnderstand","thumb-down"],["잘못된 정보 또는 샘플 코드","incorrectInformationOrSampleCode","thumb-down"],["필요한 정보/샘플이 없음","missingTheInformationSamplesINeed","thumb-down"],["번역 문제","translationIssue","thumb-down"],["기타","otherDown","thumb-down"]],["최종 업데이트: 2025-09-04(UTC)"],[[["\u003cp\u003eAs of July 2024, automatic migration to new speech models is complete for configurations where the current model is \u003ccode\u003ephone_call\u003c/code\u003e, the \u003ccode\u003emodel_variant\u003c/code\u003e is other than \u003ccode\u003eUSE_STANDARD\u003c/code\u003e, and \u003ccode\u003esingle_utterance\u003c/code\u003e is \u003ccode\u003efalse\u003c/code\u003e, directing traffic to the \u003ccode\u003etelephony\u003c/code\u003e model.\u003c/p\u003e\n"],["\u003cp\u003eAutomatic migration for other speech model settings is currently on hold with no estimated time of resumption.\u003c/p\u003e\n"],["\u003cp\u003eUsers can proactively opt in to migrate earlier by changing the speech model identifier in the Conversational Agents (Dialogflow CX) or Dialogflow ES API to take advantage of the improved accuracy.\u003c/p\u003e\n"],["\u003cp\u003eThe migration affects specific language tags, including \u003ccode\u003een-au\u003c/code\u003e, \u003ccode\u003een-gb\u003c/code\u003e, \u003ccode\u003ede\u003c/code\u003e, \u003ccode\u003ees\u003c/code\u003e, \u003ccode\u003efr\u003c/code\u003e, \u003ccode\u003eit\u003c/code\u003e, \u003ccode\u003eja\u003c/code\u003e, and \u003ccode\u003ept-br\u003c/code\u003e, redirecting traffic from the \u003ccode\u003ephone_call\u003c/code\u003e model to new conformer-based speech models.\u003c/p\u003e\n"],["\u003cp\u003eIf the advanced timeout-based end of speech sensitivity is enabled in a Dialogflow CX agent, the model identifier is routed assuming \u003ccode\u003esingle_utterance\u003c/code\u003e is \u003ccode\u003efalse\u003c/code\u003e, regardless of the actual setting.\u003c/p\u003e\n"]]],[],null,["# Speech model migration for phone_call in Q2 2024\n\nMigration status\n----------------\n\nAs of July 2024, the [automatic migration](#automatic) has been completed for\nthe following combination of settings that correspond to the second row in the\n[migration table](#overview):\n\nFor other combination of settings, the automatic migration will remain on hold\nuntil further notice (no ETA).\n\nYou can continue to use the `phone_call` model identifier but are encouraged to\n[opt in](#opt-in) to use conformer-based speech models.\n\nMigration overview\n------------------\n\nFor certain [language tag](/dialogflow/cx/docs/reference/language) and speech\nsetting combinations, the Conversational Agents (Dialogflow CX) and Dialogflow ES traffic with audio will\ngradually route away from the classic Speech-to-Text model behind the\n`phone_call` model identifier to the new conformer-based [speech\nmodels](/speech-to-text/docs/transcription-model). As the redirection will be\ninternal only, the current model identifiers and code will remain valid and\ncontinue to serve traffic.\n\nThe affected language tags are: `en-au`, `en-gb`, `de`, `de-de`, `es`, `es-es`,\n`es-us`, `fr`, `fr-ca`, `fr-fr`, `it`, `it-it`, `ja`, `pt-br`.\n\nThe following table shows the routing of Speech-to-Text\n`phone_call` model that will take place when the migration takes effect. If your\nconfiguration matches the language tag and the first three columns in the table,\nthe new model identifier is listed in the fourth column.\n\nMigration options\n-----------------\n\nThere are three migration options: early opt-in, automatic, and temporary\nopt-out.\n\n### Opt in and migrate earlier\n\nTo opt in proactively, change the speech model identifier that you have been\nusing in the [Conversational Agents (Dialogflow CX)\nAPI](/dialogflow/cx/docs/reference/rest/v3/ConversationTurn#inputaudioconfig)\nand [Dialogflow ES\nAPI](/dialogflow/es/docs/reference/rpc/google.cloud.dialogflow.v2#inputaudioconfig)\nwith the updated one. For Conversational Agents (Dialogflow CX), you can also change the [model\nidentifier](/dialogflow/cx/docs/concept/agent#stt-model-selection) in the\n**Advanced speech settings** under **Speech and IVR settings** in the\nConversational Agents (Dialogflow CX) console.\n\nSpeech model specified in the API at runtime takes precedence over the speech\nmodel specified in the Conversational Agents (Dialogflow CX) console, unless you enable the [**Override\nrequest-level speech\nmodel**](/dialogflow/cx/docs/concept/agent#stt-model-selection) setting. Read\nmore about precedence in [speech\nmodels](/dialogflow/cx/docs/concept/speech-models).\n\nFor partner telephony integrations, check with your telephony provider if the\nspeech model is being set in the API. If you are using the Conversational Agents (Dialogflow CX) Phone\nGateway, review the speech model\n[limitations](/dialogflow/cx/docs/concept/integration/phone-gateway#limits).\n\nMigrating your project proactively gives you time to test the models and take\nadvantage of the improved accuracy and robustness earlier. Most integrations\nrequire `single_utterance = true`, so if you are not using [advanced\ntimeout-based end of speech\nsensitivity](/dialogflow/cx/docs/concept/agent#advanced-end-of-speech-sensitivity)\nyou'll likely want to use `latest_short`.\n\nIf you experience issues with the new models, [open a Google Cloud support\ncase](/support/docs/customer-care-procedures) with issue details.\n\n### Migrate automatically\n\n| **Note:** See [migration status](#status).\n\nTo be migrated automatically, no action is required on your part. The existing\nSpeech-to-Text model identifiers are valid and we will shift traffic\ninternally.\n\n### Opt out and migrate later\n\nDue to the suspended [migration status](#status), new opt-out requests are not\naccepted. If your projects were opted out before July 2024, they will remain in\nthis state until further announcement."]]
