PropertyId Enum

The subscription key used with Speech service endpoints. If you are using an intent recognizer, you need to specify the LUIS endpoint key for your particular LUIS app. Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromSubscription(String, String).

The Speech service endpoint, a URL. Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromEndpoint(Uri, String), or FromEndpoint(Uri). NOTE: This endpoint is not the same as the endpoint used to obtain an access token.

The Speech service region associated with the subscription key. Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromSubscription(String, String), FromEndpoint(Uri, String), FromEndpoint(Uri), FromHost(Uri, String), FromHost(Uri), FromAuthorizationToken(String, String).

The Speech service authorization token (aka access token). Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromAuthorizationToken(String, String), AuthorizationToken, AuthorizationToken, AuthorizationToken.

Unused. The Speech service authorization type.

The Custom Speech or Custom Voice Service endpoint id. Under normal circumstances, you shouldn't have to use this property directly. Instead use FromEndpoint(Uri, String), or FromEndpoint(Uri). NOTE: The endpoint id is available in the Custom Speech Portal, listed under Endpoint Details.

The Speech service host (url). Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromHost(Uri, String), or FromHost(Uri).

The host name of the proxy server used to connect to the Speech service. Under normal circumstances, you shouldn't have to use this property directly. Instead use SetProxy(String, Int32, String, String). Added in 1.1.0

The port of the proxy server used to connect to the Speech service. Under normal circumstances, you shouldn't have to use this property directly. Instead use SetProxy(String, Int32, String, String). Added in 1.1.0

The user name of the proxy server used to connect to the Speech service. Under normal circumstances, you shouldn't have to use this property directly. Instead use SetProxy(String, Int32, String, String). Added in 1.1.0

The password of the proxy server used to connect to the Speech service. Under normal circumstances, you shouldn't have to use this property directly. Instead use SetProxy(String, Int32, String, String). Added in 1.1.0

The URL string built from speech configuration. This property is read-only. The SDK uses this value internally. Added in 1.5.0

Specifies the list of hosts for which proxies should not be used. This setting overrides all other configurations. Hostnames are separated by commas and are matched in a case-insensitive manner. Wildcards are not supported.

The list of comma separated languages (in BCP-47 format) used as target translation languages. Under normal circumstances, you shouldn't have to use this property directly. Instead, use AddTargetLanguage(String) and the read-only TargetLanguages collection.

The name of the voice used for Text-to-speech. Under normal circumstances, you shouldn't have to use this property directly. Instead use VoiceName. Find valid voice names here.

Translation features. For internal use.

The Language Understanding Service Region. Under normal circumstances, you shouldn't have to use this property directly. Instead use LanguageUnderstandingModel.

The Speech service recognition mode. Can be INTERACTIVE, CONVERSATION, DICTATION. This property is read-only. The SDK uses it internally.

The spoken language to be recognized (in BCP-47 format). Under normal circumstances, you shouldn't have to use this property directly. Instead, use SpeechRecognitionLanguage.

The session id. This id is a universally unique identifier (aka UUID) representing a specific binding of an audio input stream and the underlying speech recognition instance to which it is bound. Under normal circumstances, you shouldn't have to use this property directly. Instead use SessionId.

The string to specify the backend to be used for speech recognition; allowed options are online and offline. Under normal circumstances, you shouldn't use this property directly. Currently the offline option is only valid when EmbeddedSpeechConfig is used. Added in version 1.19.0

The name of the model to be used for speech recognition. Under normal circumstances, you shouldn't use this property directly. Currently this is only valid when EmbeddedSpeechConfig is used. Added in version 1.19.0

The spoken language to be synthesized (e.g. en-US). Added in 1.4.0

The name of the voice to be used for Text-to-speech. Added in 1.4.0

The string to specify speech synthesis output audio format (e.g. riff-16khz-16bit-mono-pcm) Added in 1.4.0

Indicates whether to use compressed audio format for speech synthesis audio transmission. This property only matters when SpeechServiceConnection_SynthOutputFormat is set to a pcm format. If this property is not set to false and GStreamer is available, SDK will use compressed format for synthesized audio transmission, and decode it. You can set this property to false to use raw pcm format for transmission on wire. Added in 1.16.0

The string to specify TTS backend; valid options are online and offline. Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromPath(String) or FromPaths(String[]). to set the synthesis backend to offline. Added in version 1.19.0

The data file path(s) for offline synthesis engine; only valid when synthesis backend is offline. Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromPath(String) or FromPaths(String[]). Added in version 1.19.0

The name of the offline TTS voice to be used for speech synthesis. Under normal circumstances, you shouldn't use this property directly. Instead, use SetSpeechSynthesisVoice(String, String). Added in version 1.19.0

The Cognitive Services Speech Service voices list API endpoint (url). Under normal circumstances, you don't need to specify this property, SDK will construct it based on the region/host/endpoint of SpeechConfig. Added in 1.16.0

The initial silence timeout value (in milliseconds) used by the service. Added in 1.5.0

The end silence timeout value (in milliseconds) used by the service. Added in 1.5.0

A boolean value specifying whether audio logging is enabled in the service or not. Audio and content logs are stored either in Microsoft-owned storage, or in your own storage account linked to your Cognitive Services subscription (Bring Your Own Storage (BYOS) enabled Speech resource). Added in 1.5.0.

The speech service connection language identifier mode. Can be "AtStart" (the default), or "Continuous". See Language Identification document. Added in 1.25.0

The speech service connection translation categoryId.

The auto detect source languages. Added in 1.9.0

The auto detect source language result. Added in 1.9.0

The requested Speech service response output format (OutputFormat.Simple or OutputFormat.Detailed). Under normal circumstances, you shouldn't have to use this property directly. Instead, use OutputFormat.

Unused. The requested Speech service response output profanity level.

The requested Speech service response output profanity setting. Allowed values are masked, removed, and raw. Added in 1.5.0

A string value specifying which post processing option should be used by service. Allowed value: TrueText. Added in 1.5.0

A boolean value specifying whether to include word-level timestamps in the response result. Added in 1.5.0

The number of times a word has to be in partial results to be returned. Added in 1.5.0

A string value specifying the output format option in the response result. Internal use only. Added in 1.5.0

A boolean value specifying whether to include SNR (signal to noise ratio) in the response result. Added in version 1.18.0

A boolean value to request for stabilizing translation partial results by omitting words in the end. Added in 1.5.0

A boolean value specifying whether to request WordBoundary events. Added in version 1.21.0.

A boolean value specifying whether to request punctuation boundary in WordBoundary Events. Default is true. Added in version 1.21.0.

A boolean value specifying whether to request sentence boundary in WordBoundary Events. Default is false. Added in version 1.21.0.

A boolean value specifying whether the SDK should synchronize synthesis metadata events, (e.g. word boundary, viseme, etc.) to the audio playback. This only takes effect when the audio is played through the SDK. Default is true. If set to false, the SDK will fire the events as they come from the service, which may be out of sync with the audio playback. Added in version 1.31.0.

The Speech service response output (in JSON format). This property is available on recognition result objects only.

The Speech service error details (in JSON format). Under normal circumstances, you shouldn't have to use this property directly. Instead use ErrorDetails.

The recognition latency in milliseconds. Read-only, available on final speech/translation/intent results. This measures the latency between when an audio input is received by the SDK, and the moment the final result is received from the service. The SDK computes the time difference between the last audio fragment from the audio input that is contributing to the final result, and the time the final result is received from the speech service. Added in 1.3.0

The recognition backend. Read-only, available on speech recognition results. This indicates whether cloud (online) or embedded (offline) recognition was used to produce the result.

The speech synthesis first byte latency in milliseconds. Read-only, available on final speech synthesis results. This measures the latency between when the synthesis is started to be processed, and the moment the first byte audio is available. Added in version 1.17.0.

The speech synthesis all bytes latency in milliseconds. Read-only, available on final speech synthesis results. This measures the latency between when the synthesis is started to be processed, and the moment the whole audio is synthesized. Added in version 1.17.0.

The underrun time for speech synthesis in milliseconds. Read-only, available on results in SynthesisCompleted events. This measures the total underrun time from AudioConfig_PlaybackBufferLengthInMs is filled to synthesis completed. Added in version 1.17.0.

The speech synthesis connection latency in milliseconds. Read-only, available on final speech synthesis results. This measures the latency between when the synthesis is started to be processed, and the moment the HTTP/WebSocket connection is established. Added in version 1.26.0.

The speech synthesis network latency in milliseconds. Read-only, available on final speech synthesis results. This measures the network round trip time. Added in version 1.26.0.

The speech synthesis service latency in milliseconds. Read-only, available on final speech synthesis results. This measures the service processing time to synthesize the first byte of audio. Added in version 1.26.0.

Indicates which backend the synthesis is finished by. Read-only, available on speech synthesis results, except for the result in SynthesisStarted event. Added in version 1.19.0.

Determines if intermediate results contain speaker identification.

Unused. The cancellation reason.

Unused. The cancellation text.

Unused. The cancellation detailed text.

The Language Understanding Service response output (in JSON format). Available via Properties.

The device name for audio render. Under normal circumstances, you shouldn't have to use this property directly. Instead, use FromSpeakerOutput(String). Added in version 1.17.0

Playback buffer length in milliseconds, default is 50 milliseconds. Added in version 1.17.0

The file name to write logs. Added in 1.4.0

A duration of detected silence, measured in milliseconds, after which speech-to-text will determine a spoken phrase has ended and generate a final Recognized result. Configuring this timeout may be helpful in situations where spoken input is significantly faster or slower than usual and default segmentation behavior consistently yields results that are too long or too short. Segmentation timeout values that are inappropriately high or low can negatively affect speech-to-text accuracy; this property should be carefully configured and the resulting behavior should be thoroughly validated as intended.

For more information about timeout configuration that includes discussion of default behaviors, please visit https://aka.ms/csspeech/timeouts.

The maximum length of a spoken phrase when using the Time segmentation strategy.

The strategy used to determine when a spoken phrase has ended and a final Recognized result should be generated. Allowed values are "Default", "Time", and "Semantic".

Identifier used to connect to the backend service. Added in 1.5.0

Type of dialog backend to connect to. Added in 1.7.0

Silence timeout for listening. Added in 1.5.0

The from identifier to add to speech recognition activities. Added in 1.5.0

ConversationId for the session. Added in 1.8.0

Comma separated list of custom voice deployment ids. Added in 1.8.0

Speech activity template, stamp properties from the template on the activity generated by the service for speech. See SpeechActivityTemplate Added in 1.10.0

Gets your identifier in the conversation. Added in 1.13.0

A boolean value that specifies whether or not the client should receive turn status messages and generate corresponding TurnStatusReceived events. Defaults to true. Added in 1.15.0

Additional identifying information, such as a Direct Line token, used to authenticate with the backend service. Added in 1.16.0

The time stamp associated to data buffer written by client when using Pull/Push audio mode streams. The time stamp is a 64-bit value with a resolution of 90 kHz. The same as the presentation timestamp in an MPEG transport stream. See https://en.wikipedia.org/wiki/Presentation_timestamp. Added in 1.5.0

The user identifier associated to data buffer written by client when using Pull/Push audio mode streams. Added in 1.5.0

The reference text of the audio for pronunciation evaluation. For this and the following pronunciation assessment parameters, see Pronunciation assessment parameters for details. Under normal circumstances, you shouldn't have to use this property directly. Added in 1.14.0

The point system for pronunciation score calibration (FivePoint or HundredMark). Under normal circumstances, you shouldn't have to use this property directly. Added in 1.14.0

The pronunciation evaluation granularity (Phoneme, Word, or FullText). Under normal circumstances, you shouldn't have to use this property directly. Added in 1.14.0

Indicates miscue calculation state. When enabled, the pronounced words will be compared to the reference text, and will be marked with omission/insertion based on the comparison. The default setting is false. Under normal circumstances, you shouldn't have to use this property directly. Added in 1.14.0

The pronunciation evaluation phoneme alphabet. The valid values are "SAPI" (default) and "IPA" Under normal circumstances, you shouldn't have to use this property directly. Instead, use PhonemeAlphabet. Added in version 1.20.0

The pronunciation evaluation nbest phoneme count. Under normal circumstances, you shouldn't have to use this property directly. Instead, use NBestPhonemeCount. Added in version 1.20.0

Whether to enable prosody assessment. Under normal circumstances, you shouldn't have to use this property directly. Instead, use EnableProsodyAssessment(). Added in version 1.33.0

The JSON string of pronunciation assessment parameters. Under normal circumstances, you shouldn't have to use this property directly. Added in 1.14.0

Pronunciation assessment parameters. This property is read-only. Added in 1.14.0

The content topic of the pronunciation assessment. Under normal circumstances, you shouldn't have to use this property directly. Instead, use EnableContentAssessmentWithTopic(String). Added in version 1.33.0

Speaker recognition API version. Added in 1.18.0

The name of a model to be used for speech translation. Do not use this property directly. Currently this is only valid when EmbeddedSpeechConfig is used.

The name of a model to be used for keyword recognition. Do not use this property directly. Currently this is only valid when EmbeddedSpeechConfig is used.

Enable the collection of embedded speech performance metrics which can be used to evaluate the capability of a device to use embedded speech. The collected data is included in results from specific scenarios like speech recognition. The default setting is "false". Note that metrics may not be available from all embedded speech scenarios.

The pitch of the synthesized speech.

The rate of the synthesized speech.

The volume of the synthesized speech.