Troubleshooting of Google cloud speech to text

From LemonWiki共筆
Jump to navigation Jump to search

Category:NLP > Speech to text > Troubleshooting of Google cloud speech to text

Troubleshooting of Google Cloud Speech-to-Text - 語音辨識

Brief instruction

  • If the audio file's duration is longer than 1 minute, (1) Use the uri: speech:longrunningrecognize NOT speech:recognize (2) Upload files to Google cloud storage (gcs)
  • If the audio file's duration is shorter than 1 minute, (1) Use the uri: speech:longrunningrecognize. Or choose to use speech:recognize (2) Use the files are located on the computer. Or choose to upload files to Google cloud storage (gcs)

ERROR: (gcloud.auth.application-default.print-access-token) The Application Default Credentials are not available

input & output

$ gcloud auth application-default print-access-token
ERROR: (gcloud.auth.application-default.print-access-token) The Application Default Credentials are not available. They are available if running in Google Compute Engine. Otherwise, the environment variable GOOGLE_APPLICATION_CREDENTIALS must be defined pointing to a file defining the credentials. See https://developers.google.com/accounts/docs/application-default-credentials for more information.

Solution[1][2]: Key-in the following command. And then the browser will be opened automatically. Follow the steps on the web page.

$ gcloud auth application-default login

Invalid audio channel count

error output

  {
    "error": {
      "code": 400,
      "message": "Invalid audio channel count",
      "status": "INVALID_ARGUMENT"
    }
  }

Solution: convert the audio file from stereo to mono

Invalid recognition 'config': bad encoding

error output

  {
    "error": {
      "code": 400,
      "message": "Invalid recognition 'config': bad encoding..",
      "status": "INVALID_ARGUMENT"
    }
  }

Solution: Specify the encoding of audio file. For details, see Introduction to Audio Encoding  |  Cloud Speech-to-Text API  |  Google Cloud & RecognitionConfig  |  Cloud Speech-to-Text API  |  Google Cloud. You may use VLC player to view the encoding of audio file[3]. If the codec (encoding) of audio file is not in the allowed list on page, the codec (encoding) of audio file should be converted by audio converter.

If the audio file's duration is longer than 1 minute use LongRunningRecognize with a 'uri' parameter

input

$ curl -s -H "Content-Type: application/json" \
    -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
    https://speech.googleapis.com/v1p1beta1/speech:recognize \
    -d @sync-request.json

file content of sync-request.json[4]

{
  "config": {
      "encoding":"FLAC",
      "sampleRateHertz": 44100,
      "languageCode": "cmn-Hant-TW",
      "alternativeLanguageCodes": ["en-US"],
      "enableWordTimeOffsets": false
  },
  "audio": {
      "uri":"gs://<bucket_name>/<audio file name>"
  }
}

error message[5]

{
  "error": {
    "code": 400,
    "message": "Sync input too long. For audio longer than 1 min use LongRunningRecognize with a 'uri' parameter.",
    "status": "INVALID_ARGUMENT"
  }
}

Solution: (1) If the audio file's duration is shorter than 1 min, use the uri: speech:recognize. (2) If the audio file's duration is longer than 1 min. Upload files to Google cloud storage (gcs). Modify the uri from speech:recognize to speech:longrunningrecognize.

$ curl -s -H "Content-Type: application/json" \
    -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
    https://speech.googleapis.com/v1p1beta1/speech:longrunningrecognize \
    -d @sync-request.json

Request payload size exceeds the limit: 10485760 bytes

error output

  {
    "error": {
      "code": 400,
      "message": "Request payload size exceeds the limit: 10485760 bytes.",
      "status": "INVALID_ARGUMENT"
    }
  }

Solution: (1) Use the short audio file which shorter than 1 min or (2) Modify the uri from speech:recognize to speech:longrunningrecognize for long audio file which longer than 1 min


sample_rate_hertz (16000) in RecognitionConfig must either be unspecified or match the value in the FLAC header

input & output

$ curl -s -H "Content-Type: application/json" \
    -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
    https://speech.googleapis.com/v1/speech:longrunningrecognize \
    -d @sync-request.json

{
  "error": {
    "code": 400,
    "message": "sample_rate_hertz (16000) in RecognitionConfig must either be unspecified or match the value in the FLAC header (44100).",
    "status": "INVALID_ARGUMENT"
  }
}

Solution: verify the sample rate of audio file

Invalid JSON payload received. Unknown name \"alternative_language_codes\" at 'config': Cannot find field

input & output

$ curl -s -H "Content-Type: application/json" \
    -H "Authorization: Bearer "$(gcloud auth application-default print-access-token) \
    https://speech.googleapis.com/v1/speech:longrunningrecognize \
    -d @sync-request.json

{
  "error": {
    "code": 400,
    "message": "Invalid JSON payload received. Unknown name \"alternative_language_codes\" at 'config': Cannot find field.",
    "status": "INVALID_ARGUMENT",
    "details": [
      {
        "@type": "type.googleapis.com/google.rpc.BadRequest",
        "fieldViolations": [
          {
            "field": "config",
            "description": "Invalid JSON payload received. Unknown name \"alternative_language_codes\" at 'config': Cannot find field."
          }
        ]
      }
    ]
  }
}

Solution: modify the uri from https://speech.googleapis.com/v1/speech:longrunningrecognize to https://speech.googleapis.com/v1p1beta1/speech:longrunningrecognize

Related

References


Troubleshooting of ...

Template