Rediger

Del via

Get batch transcription results

  • Artikkel

To get transcription results, first check the status of the transcription job. If the job is completed, you can retrieve the transcriptions and transcription report.

Get transcription status

To get the status of the transcription job, call the Transcriptions_Get operation of the Speech to text REST API.

Important

Batch transcription jobs are scheduled on a best-effort basis. At peak hours, it may take up to 30 minutes or longer for a transcription job to start processing. Most of the time during the execution the transcription status will be Running. This is because the job is assigned the Running status the moment it moves to the batch transcription backend system. When the base model is used, this assignment happens almost immediately; it's slightly slower for custom models. Thus, the amount of time a transcription job spends in the Running state doesn't correspond to the actual transcription time but also includes waiting time in the internal queues.

Make an HTTP GET request using the URI as shown in the following example. Replace YourTranscriptionId with your transcription ID, replace YourSubscriptionKey with your Speech resource key, and replace YourServiceRegion with your Speech resource region.

curl -v -X GET "https://YourServiceRegion.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/YourTranscriptionId" -H "Ocp-Apim-Subscription-Key: YourSubscriptionKey"

You should receive a response body in the following format:

{
  "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3",
  "model": {
    "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/models/base/aaa321e9-5a4e-4db1-88a2-f251bbe7b555"
  },
  "links": {
    "files": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files"
  },
  "properties": {
    "diarizationEnabled": false,
    "wordLevelTimestampsEnabled": false,
    "displayFormWordLevelTimestampsEnabled": true,
    "channels": [
      0,
      1
    ],
    "punctuationMode": "DictatedAndAutomatic",
    "profanityFilterMode": "Masked",
    "duration": "PT3S",
    "languageIdentification": {
      "candidateLocales": [
        "en-US",
        "de-DE",
        "es-ES"
      ]
    }
  },
  "lastActionDateTime": "2024-05-10T18:39:09Z",
  "status": "Succeeded",
  "createdDateTime": "2024-05-10T18:39:07Z",
  "locale": "en-US",
  "displayName": "My Transcription"
}

The status property indicates the current status of the transcriptions. The transcriptions and transcription report are available when the transcription status is Succeeded.

Important

Batch transcription jobs are scheduled on a best-effort basis. At peak hours, it may take up to 30 minutes or longer for a transcription job to start processing. Most of the time during the execution the transcription status will be Running. This is because the job is assigned the Running status the moment it moves to the batch transcription backend system. When the base model is used, this assignment happens almost immediately; it's slightly slower for custom models. Thus, the amount of time a transcription job spends in the Running state doesn't correspond to the actual transcription time but also includes waiting time in the internal queues.

To get the status of the transcription job, use the spx batch transcription status command. Construct the request parameters according to the following instructions:

  • Set the transcription parameter to the ID of the transcription that you want to get.

Here's an example Speech CLI command to get the transcription status:

spx batch transcription status --api-version v3.2 --transcription YourTranscriptionId

You should receive a response body in the following format:

{
  "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3",
  "model": {
    "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/models/base/aaa321e9-5a4e-4db1-88a2-f251bbe7b555"
  },
  "links": {
    "files": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files"
  },
  "properties": {
    "diarizationEnabled": false,
    "wordLevelTimestampsEnabled": false,
    "displayFormWordLevelTimestampsEnabled": true,
    "channels": [
      0,
      1
    ],
    "punctuationMode": "DictatedAndAutomatic",
    "profanityFilterMode": "Masked",
    "duration": "PT3S"
  },
  "lastActionDateTime": "2024-05-10T18:39:09Z",
  "status": "Succeeded",
  "createdDateTime": "2024-05-10T18:39:07Z",
  "locale": "en-US",
  "displayName": "My Transcription"
}

The status property indicates the current status of the transcriptions. The transcriptions and transcription report are available when the transcription status is Succeeded.

For Speech CLI help with transcriptions, run the following command:

spx help batch transcription

Get transcription results

The Transcriptions_ListFiles operation returns a list of result files for a transcription. A transcription report file is provided for each submitted batch transcription job. In addition, one transcription file (the end result) is provided for each successfully transcribed audio file.

Make an HTTP GET request using the "files" URI from the previous response body. Replace YourTranscriptionId with your transcription ID, replace YourSubscriptionKey with your Speech resource key, and replace YourServiceRegion with your Speech resource region.

curl -v -X GET "https://YourServiceRegion.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/YourTranscriptionId/files" -H "Ocp-Apim-Subscription-Key: YourSubscriptionKey"

You should receive a response body in the following format:

{
  "values": [
    {
      "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files/2dd180a1-434e-4368-a1ac-37350700284f",
      "name": "contenturl_0.json",
      "kind": "Transcription",
      "properties": {
        "size": 3407
      },
      "createdDateTime": "2024-05-10T18:39:09Z",
      "links": {
        "contentUrl": "YourTranscriptionUrl"
      }
    },
    {
      "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files/c027c6a9-2436-4303-b64b-e98e3c9fc2e3",
      "name": "contenturl_1.json",
      "kind": "Transcription",
      "properties": {
        "size": 8233
      },
      "createdDateTime": "2024-05-10T18:39:09Z",
      "links": {
        "contentUrl": "YourTranscriptionUrl"
      }
    },
    {
      "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files/faea9a41-c95c-4d91-96ff-e39225def642",
      "name": "report.json",
      "kind": "TranscriptionReport",
      "properties": {
        "size": 279
      },
      "createdDateTime": "2024-05-10T18:39:09Z",
      "links": {
        "contentUrl": "YourTranscriptionReportUrl"
      }
    }
  ]
}

The location of each transcription and transcription report files with more details are returned in the response body. The contentUrl property contains the URL to the transcription ("kind": "Transcription") or transcription report ("kind": "TranscriptionReport") file.

If you didn't specify a container in the destinationContainerUrl property of the transcription request, the results are stored in a container managed by Microsoft. When the transcription job is deleted, the transcription result data is also deleted.

The spx batch transcription list command returns a list of result files for a transcription. A transcription report file is provided for each submitted batch transcription job. In addition, one transcription file (the end result) is provided for each successfully transcribed audio file.

  • Set the required files flag.
  • Set the required transcription parameter to the ID of the transcription that you want to get logs.

Here's an example Speech CLI command that gets a list of result files for a transcription:

spx batch transcription list --api-version v3.2 --files --transcription YourTranscriptionId

You should receive a response body in the following format:

{
  "values": [
    {
      "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files/2dd180a1-434e-4368-a1ac-37350700284f",
      "name": "contenturl_0.json",
      "kind": "Transcription",
      "properties": {
        "size": 3407
      },
      "createdDateTime": "2024-05-10T18:39:09Z",
      "links": {
        "contentUrl": "YourTranscriptionUrl"
      }
    },
    {
      "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files/c027c6a9-2436-4303-b64b-e98e3c9fc2e3",
      "name": "contenturl_1.json",
      "kind": "Transcription",
      "properties": {
        "size": 8233
      },
      "createdDateTime": "2024-05-10T18:39:09Z",
      "links": {
        "contentUrl": "YourTranscriptionUrl"
      }
    },
    {
      "self": "https://eastus.api.cognitive.microsoft.com/speechtotext/v3.2/transcriptions/637d9333-6559-47a6-b8de-c7d732c1ddf3/files/faea9a41-c95c-4d91-96ff-e39225def642",
      "name": "report.json",
      "kind": "TranscriptionReport",
      "properties": {
        "size": 279
      },
      "createdDateTime": "2024-05-10T18:39:09Z",
      "links": {
        "contentUrl": "YourTranscriptionReportUrl"
      }
    }
  ]
}

The location of each transcription and transcription report files with more details are returned in the response body. The contentUrl property contains the URL to the transcription ("kind": "Transcription") or transcription report ("kind": "TranscriptionReport") file.

By default, the results are stored in a container managed by Microsoft. When the transcription job is deleted, the transcription result data is also deleted.

Transcription report file

One transcription report file is provided for each submitted batch transcription job.

The contents of each transcription result file are formatted as JSON, as shown in this example.

{
  "successfulTranscriptionsCount": 2,
  "failedTranscriptionsCount": 0,
  "details": [
    {
      "source": "https://crbn.us/hello.wav",
      "status": "Succeeded"
    },
    {
      "source": "https://crbn.us/whatstheweatherlike.wav",
      "status": "Succeeded"
    }
  ]
}

Transcription result file

One transcription result file is provided for each successfully transcribed audio file.

The contents of each transcription result file are formatted as JSON, as shown in this example.

{
  "source": "...",
  "timestamp": "2023-07-10T14:28:16Z",
  "durationInTicks": 25800000,
  "duration": "PT2.58S",
  "combinedRecognizedPhrases": [
    {
      "channel": 0,
      "lexical": "hello world",
      "itn": "hello world",
      "maskedITN": "hello world",
      "display": "Hello world."
    }
  ],
  "recognizedPhrases": [
    {
      "recognitionStatus": "Success",
      "channel": 0,
      "offset": "PT0.76S",
      "duration": "PT1.32S",
      "offsetInTicks": 7600000.0,
      "durationInTicks": 13200000.0,
      "nBest": [
        {
          "confidence": 0.5643338,
          "lexical": "hello world",
          "itn": "hello world",
          "maskedITN": "hello world",
          "display": "Hello world.",
          "displayWords": [
            {
              "displayText": "Hello",
              "offset": "PT0.76S",
              "duration": "PT0.76S",
              "offsetInTicks": 7600000.0,
              "durationInTicks": 7600000.0
            },
            {
              "displayText": "world.",
              "offset": "PT1.52S",
              "duration": "PT0.56S",
              "offsetInTicks": 15200000.0,
              "durationInTicks": 5600000.0
            }
          ]
        },
        {
          "confidence": 0.1769063,
          "lexical": "helloworld",
          "itn": "helloworld",
          "maskedITN": "helloworld",
          "display": "helloworld"
        },
        {
          "confidence": 0.49964225,
          "lexical": "hello worlds",
          "itn": "hello worlds",
          "maskedITN": "hello worlds",
          "display": "hello worlds"
        },
        {
          "confidence": 0.4995761,
          "lexical": "hello worm",
          "itn": "hello worm",
          "maskedITN": "hello worm",
          "display": "hello worm"
        },
        {
          "confidence": 0.49418187,
          "lexical": "hello word",
          "itn": "hello word",
          "maskedITN": "hello word",
          "display": "hello word"
        }
      ]
    }
  ]
}

Depending in part on the request parameters set when you created the transcription job, the transcription file can contain the following result properties.

Property Description
channel The channel number of the results. For stereo audio streams, the left and right channels are split during the transcription. A JSON result file is created for each input audio file.
combinedRecognizedPhrases The concatenated results of all phrases for the channel.
confidence The confidence value for the recognition.
display The display form of the recognized text. Added punctuation and capitalization are included.
displayWords The timestamps for each word of the transcription. The displayFormWordLevelTimestampsEnabled request property must be set to true, otherwise this property isn't present.

Note: This property is only available with Speech to text REST API version 3.1 and later.
duration The audio duration. The value is an ISO 8601 encoded duration.
durationInTicks The audio duration in ticks (one tick is 100 nanoseconds).
itn The inverse text normalized (ITN) form of the recognized text. Abbreviations such as "Doctor Smith" to "Dr Smith", phone numbers, and other transformations are applied.
lexical The actual words recognized.
locale The locale identified from the input the audio. The languageIdentification request property must be set, otherwise this property isn't present.

Note: This property is only available with Speech to text REST API version 3.1 and later.
maskedITN The ITN form with profanity masking applied.
nBest A list of possible transcriptions for the current phrase with confidences.
offset The offset in audio of this phrase. The value is an ISO 8601 encoded duration.
offsetInTicks The offset in audio of this phrase in ticks (one tick is 100 nanoseconds).
recognitionStatus The recognition state. For example: "Success" or "Failure".
recognizedPhrases The list of results for each phrase.
source The URL that was provided as the input audio source. The source corresponds to the contentUrls or contentContainerUrl request property. The source property is the only way to confirm the audio input for a transcription.
speaker The identified speaker. The diarization and diarizationEnabled request properties must be set, otherwise this property isn't present.
timestamp The creation date and time of the transcription. The value is an ISO 8601 encoded timestamp.
words A list of results with lexical text for each word of the phrase. The wordLevelTimestampsEnabled request property must be set to true, otherwise this property isn't present.

Next steps