mcpField
A
azure-speech

azure-speech

github.com/microsoftdocs开发796241 次安装3 天前更新
代码仓库

关于 azure-speech

azure-speech 是面向使用 Foundry Tools 中 Azure Speech 的开发者和运维人员的指导技能，汇集本地参考资料及远程文档，涵盖语音功能、安全、集成和部署。

主要功能

  1. 01语音故障排查参考
  2. 02音频与转录最佳实践
  3. 03API 选择和迁移指南
  4. 04安全与配置参考
  5. 05集成和部署模式

适用场景

  1. 01排查 Speech SDK 或容器问题
  2. 02准备批量转录音频
  3. 03规划 REST API 或功能迁移
  4. 04在容器中部署 Speech 服务

MCP 相关

该技能可使用 mcp_microsoftdocs:microsoft_docs_fetch 获取 Microsoft 文档，并以网页抓取作为备用方式。

安装

复制下面的命令，在终端中运行。

$npx skills add https://github.com/microsoftdocs/agent-skills --skill azure-speech

安装后在 Claude Code 中输入/azure-speech确认已加载。

详细介绍

Foundry Tools 中的 Azure Speech 技能

此技能提供有关 Foundry Tools 中 Azure Speech 的专业指导，涵盖故障排除、最佳实践、决策、限制和配额、安全性、配置、集成和编码模式，以及部署。它结合了本地快速参考内容与远程文档获取功能。

如何使用此技能

重要提示（面向 Agent）：使用下方的类别索引定位相关部分。对于标有行范围的类别（例如 L35-L120），请使用 read_file 读取指定行。对于带有文件链接的类别（例如 [security.md](security.md)），请使用 read_file 读取链接指向的参考文件。

重要提示（面向 Agent）：如果 metadata.generated_at 的日期超过 3 个月，请建议用户从仓库拉取最新版本。如果 mcp_microsoftdocs 工具不可用，请建议用户安装该工具：安装指南

此技能需要网络访问权限才能获取文档内容：

  • 首选：使用 mcp_microsoftdocs:microsoft_docs_fetch，查询字符串为 from=learn-agent-skill。返回 Markdown。
  • 备用：使用 fetch_webpage，查询字符串为 from=learn-agent-skill&accept=text/markdown。返回 Markdown。

类别索引

类别行说明
故障排除L36-L44诊断并修复 Foundry 中常见的 Azure Speech 问题（TTS/STT 错误、SDK 和容器问题、部署失败，以及如何收集会话/转写 ID 以便寻求支持）。
最佳实践L45-L62Azure Speech 的音频/视频准备、转写质量、自定义语音/虚拟形象、延迟和内存调优、关键词/短语准确性、设备设计、可靠性以及备份/恢复方面的最佳实践。
决策L63-L78关于如何选择 API 和电话选项、评估设备和语音类型的指南，以及 Speech/语音 REST 版本之间的分步迁移指南，和从已停用功能迁移到更新服务的指南。
限制和配额L79-L86Azure Speech 的配额、限制和生命周期规则：自定义语音/语音模型和终结点、支持的语言/语音，以及部署和使用 Speech 服务的容量限制。
安全性L87-L100保护 Azure AI Speech：身份验证（Entra、RBAC）、网络隔离（VNet、Private Link、主权云）、加密/BYOK/BYOS、合规性/同意，以及安全集成自定义模型/语音/虚拟形象。
配置L101-L134配置 Azure Speech/语音功能：识别、TTS、虚拟形象、容器、日志记录、存储、语音音标/SSML、音频输入、诊断，以及 Voice Live/CLI 运行时和输出选项。
集成和编码模式L135-L168将 Azure Speech（STT、TTS、虚拟形象、翻译）与电话、Voice Live、REST/SDK API、OpenAI、自动化工具和实时 Agent 工作流集成的模式和代码示例。
部署L169-L179通过容器、Kubernetes/Helm 和批处理 API 部署并运行 Azure Speech 服务（STT、TTS、语言识别），包括自定义模型和本地部署。
故障排除
主题URL
排查常见的 Azure 文本转语音问题https://learn.microsoft.com/en-us/azure/ai-services/speech-service/faq-tts
获取 Speech 转文本会话 ID 和转写 ID 以便寻求支持https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-get-speech-session-id
解决 Foundry 中常见的 Azure Speech 问题https://learn.microsoft.com/en-us/azure/ai-services/speech-service/known-issues
排查 Azure Speech 容器部署问题https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-faq
诊断并修复常见的 Azure Speech SDK 问题https://learn.microsoft.com/en-us/azure/ai-services/speech-service/troubleshooting
最佳实践
主题URL
为批量转录准备和定位音频数据https://learn.microsoft.com/en-us/azure/ai-services/speech-service/batch-transcription-audio-data
创建高质量的人工标注语音转录文本https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-custom-speech-human-labeled-transcriptions
为专业自定义语音准备训练数据https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-custom-voice-training-data
应用最佳实践，降低 Speech 合成延迟https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-lower-speech-synthesis-latency
跟踪和管理 Azure Speech SDK 内存使用情况https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-track-speech-sdk-memory-usage
处理 Voice Live 聊天记录中的中断响应https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-auto-truncation
使用 Foundry harness 评估 Voice Live Agenthttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-evaluate
使用短语列表提高 Azure Speech 的准确性https://learn.microsoft.com/en-us/azure/ai-services/speech-service/improve-accuracy-phrase-list
应用关键词识别设计和准确性指南https://learn.microsoft.com/en-us/azure/ai-services/speech-service/keyword-recognition-guidelines
缓解 Azure AI Speech SDK CRL 兼容性问题https://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-to-sdk-1-48-2
录制高质量样本，用于自定义语音训练https://learn.microsoft.com/en-us/azure/ai-services/speech-service/record-custom-voice-samples
备份和恢复自定义 Speech 与 Voice 资源https://learn.microsoft.com/en-us/azure/ai-services/speech-service/resiliency-and-recovery-plan
设计针对 Speech SDK 优化的麦克风阵列https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-sdk-microphone
为自定义虚拟形象准备高质量视频样本https://learn.microsoft.com/en-us/azure/ai-services/speech-service/text-to-speech-avatar/custom-avatar-record-video-samples
决策
主题URL
在专业语音前评估自定义轻量语音https://learn.microsoft.com/en-us/azure/ai-services/speech-service/custom-neural-voice-lite
评估设备是否适用于嵌入式语音模型https://learn.microsoft.com/en-us/azure/ai-services/speech-service/embedded-speech-performance-evaluations
选择 MAI-Transcribe-2-Streaming API 选项https://learn.microsoft.com/en-us/azure/ai-services/speech-service/mai-transcribe-2-streaming
将 Speech to text REST API 从 v3.2 迁移到 2024-11-15https://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-2024-11-15
将 Speech to text REST API 从 2024-11-15 迁移到 2025-10-15https://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-2025-10-15
将已停用的 Speech 意图识别迁移到 Language 或 OpenAIhttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-intent-recognition
从 Long Audio API 迁移到 Batch synthesishttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-to-batch-synthesis
从 v3 文本转语音迁移到自定义语音 REST APIhttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-to-custom-voice-api
将 Speech-to-text REST 从 v3.0 迁移到 v3.1https://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-v3-0-to-v3-1
将 Speech to text REST API 从 v3.1 迁移到 v3.2https://learn.microsoft.com/en-us/azure/ai-services/speech-service/migrate-v3-1-to-v3-2
评估用于实时 Agent 的 Voice Live APIhttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live
为 Foundry 语音 Agent 选择电话集成方式https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-telephony
限制与配额
主题URL
管理自定义语音模型和终结点的生命周期https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-custom-speech-model-and-endpoint-lifecycle
查看 Azure Speech 语言和语音的可用情况https://learn.microsoft.com/en-us/azure/ai-services/speech-service/language-support
Speech 服务的自定义语音终结点限制https://learn.microsoft.com/en-us/azure/ai-services/speech-service/professional-voice-deploy-endpoint
查阅 Azure Speech 配额和限制https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-services-quotas-and-limits
安全性
主题URL
为 Azure Speech 资源配置 BYOS 存储https://learn.microsoft.com/en-us/azure/ai-services/speech-service/bring-your-own-storage-speech-resource
将自定义模型与 Voice Live BYOM 集成https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-bring-your-own-model
为 Azure AI Speech 配置 Microsoft Entra 身份验证https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-configure-azure-ad-auth
配置个人语音的同意和合规性https://learn.microsoft.com/en-us/azure/ai-services/speech-service/personal-voice-create-consent
为 Speech 资源分配 Azure RBAC 角色https://learn.microsoft.com/en-us/azure/ai-services/speech-service/role-based-access-control
在 Azure 主权云中配置 Speech 服务https://learn.microsoft.com/en-us/azure/ai-services/speech-service/sovereign-clouds
管理 Speech 服务静态数据加密和密钥https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-encryption-of-data-at-rest
使用虚拟网络服务终结点保护 Speech 服务https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-service-vnet-service-endpoint
使用 Private Link 终结点保护 Azure AI Speechhttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-services-private-link
为 Azure 虚拟形象视频使用内容凭据https://learn.microsoft.com/en-us/azure/ai-services/speech-service/text-to-speech-avatar/content-credentials
配置
主题URL
配置文本转语音的批量合成属性https://learn.microsoft.com/en-us/azure/ai-services/speech-service/batch-synthesis-properties
检查状态并检索批量转录结果https://learn.microsoft.com/en-us/azure/ai-services/speech-service/batch-transcription-get
为语音转文本配置 BYOS 存储https://learn.microsoft.com/en-us/azure/ai-services/speech-service/bring-your-own-storage-speech-resource-speech-to-text
在 Azure Speech 中配置语言 ID 和说话人分离https://learn.microsoft.com/en-us/azure/ai-services/speech-service/configure-language-identification-diarization
为语音转文本定义 UPS 音标发音https://learn.microsoft.com/en-us/azure/ai-services/speech-service/customize-pronunciation
在 Linux 上为 Azure Speech SDK 配置 OpenSSLhttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-configure-openssl-linux
控制和监视 Speech SDK 服务连接https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-control-connections
配置 Azure 语音识别的后处理选项https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-post-processing
使用 Azure Speech 配置实时语音识别https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-recognize-speech
在 Speech SDK 中选择和配置音频输入设备https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-select-audio-input-devices
使用 Speech 服务的 viseme 实现面部动画https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-speech-synthesis-viseme
配置 Speech SDK 音频输入流https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-use-audio-input-streams
为 Speech SDK 和 CLI 配置压缩音频输入https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-use-codec-compressed-audio-input-streams
启用并配置 Speech SDK 诊断日志记录https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-use-logging
为 Voice Live 工具配置中间响应https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-interim-response
为 Voice Live SDK 配置 OpenTelemetry 跟踪https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-telemetry
配置语音识别的音频和转录日志记录https://learn.microsoft.com/en-us/azure/ai-services/speech-service/logging-audio-transcription
上传并验证专业语音训练数据集https://learn.microsoft.com/en-us/azure/ai-services/speech-service/professional-voice-create-training-set
配置 Azure Speech 文本转语音虚拟形象选项https://learn.microsoft.com/en-us/azure/ai-services/speech-service/quickstarts/get-started-text-to-speech-avatar
使用正确的 Azure Speech 区域终结点和 IDhttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/regions
配置 Azure Speech 容器运行时设置https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-configuration
配置并运行 Azure Speech 语言 ID 容器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-lid
为语音说话人分离容器配置 Redis 缓存https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-speech-to-text-diarization-cache
使用 SSML 配置控制语音输出https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-synthesis-markup
使用 SSML 音素和词典配置发音https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-synthesis-markup-pronunciation
为 Speech 组织 SSML 文档和事件https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-synthesis-markup-structure
配置 Speech CLI 数据存储搜索顺序和文件https://learn.microsoft.com/en-us/azure/ai-services/speech-service/spx-data-store-configuration
配置 Speech CLI 结果的输出目标https://learn.microsoft.com/en-us/azure/ai-services/speech-service/spx-output-options
配置 TTS 虚拟形象的批量合成属性https://learn.microsoft.com/en-us/azure/ai-services/speech-service/text-to-speech-avatar/batch-synthesis-avatar-properties
为 Voice Live API 配置语言和区域设置https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-language-support
集成与编码模式
主题URL
将 Speech 服务与呼叫中心电话系统集成https://learn.microsoft.com/en-us/azure/ai-services/speech-service/call-center-telephony-integration
使用 Speech SDK API 处理识别结果https://learn.microsoft.com/en-us/azure/ai-services/speech-service/get-speech-recognition-results
使用 Speech SDK 实现文本转语音合成https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-speech-synthesis
使用 Azure Speech API 翻译语音https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-translate-speech
将 Voice Live 与 Foundry 文本 Agent 集成https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-agent-integration
在 Voice Live 会话中实现函数调用https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-function-calling
将 Voice Live 与 Foundry 托管 Agent 集成https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-hosted-agent-integration
将 MCP 服务连接到 Voice Live 会话https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-mcp-server
为 Voice Live Agent 添加主动问候消息https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-voice-live-proactive-messages
通过 Realtime API 集成 MAI-Transcribe-2-Streaminghttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/mai-transcribe-2-streaming-realtime
将 MAI-Transcribe-2-Streaming 与 Speech SDK 集成https://learn.microsoft.com/en-us/azure/ai-services/speech-service/mai-transcribe-2-streaming-speech-sdk
将 Azure Speech 与 Azure OpenAI 聊天集成https://learn.microsoft.com/en-us/azure/ai-services/speech-service/openai-speech
在应用中集成个人语音说话人配置文件https://learn.microsoft.com/en-us/azure/ai-services/speech-service/personal-voice-how-to-use
使用 Speech 批量转录的 Power Automate 连接器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/power-automate-batch-transcription
通过 REST API 调用 Azure Speech-to-Texthttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/rest-speech-to-text
调用 Azure Speech-to-text 短音频 REST APIhttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/rest-speech-to-text-short
调用 Speech 文本转语音 REST API 终结点https://learn.microsoft.com/en-us/azure/ai-services/speech-service/rest-text-to-speech
在 Azure Speech 中使用 SSML 音标字母表https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-ssml-phonetic-sets
使用 SSML 自定义 Azure Speech 语音https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-synthesis-markup-voice
根据 Swagger 生成 Speech 服务 REST 客户端https://learn.microsoft.com/en-us/azure/ai-services/speech-service/swagger-documentation
使用 SSML 控制文本转语音虚拟形象的手势https://learn.microsoft.com/en-us/azure/ai-services/speech-service/text-to-speech-avatar/avatar-gestures-with-ssml
实现实时文本转语音虚拟形象流式传输https://learn.microsoft.com/en-us/azure/ai-services/speech-service/text-to-speech-avatar/real-time-synthesis-avatar
将 Voice Live 连接到 Foundry Agent Servicehttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-agents-quickstart
集成高级 Voice Live API 2025-10-01https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-api-reference-2025-10-01
使用预览版 Voice Live API 2026-01-01https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-api-reference-2026-01-01-preview
使用 Voice Live API 2026-04-10 进行流式传输https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-api-reference-2026-04-10
采用预览版 Voice Live API 2026-06-01https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-api-reference-2026-06-01-preview
实现实时 Voice Live WebSocket 事件https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-api-reference-2026-07-15
使用 Voice Live API 实现实时 Agenthttps://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-how-to
使用 Voice Live WebRTC API 处理实时音频https://learn.microsoft.com/en-us/azure/ai-services/speech-service/voice-live-webrtc
部署
主题URL
使用 Batch 合成 API 将长文本转换为语音https://learn.microsoft.com/en-us/azure/ai-services/speech-service/batch-synthesis
部署自定义语音模型和终结点https://learn.microsoft.com/en-us/azure/ai-services/speech-service/how-to-custom-speech-deploy-model
使用 Docker 运行自定义语音转文本容器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-cstt
在本地部署快速转录语音容器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-ft
使用 Docker 部署和运行 Speech 容器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-howto
使用 Helm 在 Kubernetes 上运行 Speech 容器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-howto-on-premises
使用 Docker 部署神经语音合成容器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-ntts
使用 Docker 部署语音转文本容器https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-container-stt

评论

支持 Markdown