The ML.TRANSCRIBE function

This document describes the ML.TRANSCRIBE function, which lets you transcribe audio files from an object table by using the Speech-to-Text API.


  MODEL `project_id.dataset.model_name`,
  TABLE `project_id.dataset.object_table`,
  [RECOGNITION_CONFIG => ( JSON 'recognition_config')]


ML.TRANSCRIBE takes the following arguments:

  • project_id: Your project ID.

  • dataset: The BigQuery dataset that contains the model.

  • model: The name of a remote model with a REMOTE_SERVICE_TYPE of CLOUD_AI_SPEECH_TO_TEXT_V2.

  • object_table: The name of the object table that contains URIs of the audio files.

    The audio files in the object table must be of a supported type. An error is returned for any row that contains an audio files of an unsupported type.

  • recognition_config: a STRING value that contains a RecognitionConfig resource in JSON format.

    If a recognizer has been specified for the remote model by using the SPEECH_RECOGNIZER option, you can optionally specify a recognition_config value to override the default configuration of the specified recognizer.

    This argument is required if no recognizer has been specified for the remote model by using the SPEECH_RECOGNIZER option.


ML.TRANSCRIBE returns the following columns:

  • transcripts: a STRING value that contains the transcripts from processing the audio files.
  • ml_transcribe_result: a JSON value that contains the result from the Speech-to-Text API.
  • ml_transcribe_status: a STRING value that contains the API response status for the corresponding row. This value is empty if the operation was successful.
  • The object table columns.


See Cloud AI service functions quotas and limits.

Known issues

Sometimes after a query job that uses this function finishes successfully, some returned rows contain the following error message:

A retryable error occurred: RESOURCE EXHAUSTED error from <remote endpoint>

This issue occurs because BigQuery query jobs finish successfully even if the function fails for some of the rows. The function fails when the volume of API calls to the remote endpoint exceeds the quota limits for that service. This issue occurs most often when you are running multiple parallel batch queries. BigQuery retries these calls, but if the retries fail, the resource exhausted error message is returned.

To iterate through inference calls until all rows are successfully processed, you can use the BigQuery remote inference SQL scripts or the BigQuery remote inference pipeline Dataform package.


ML.TRANSCRIBE must run in the same region as the remote model that the function references. For more information about supported locations for models based on the Speech-to-Text, see Locations for remote models.


The function can't process audio files that are longer than 1 minute. Any row that contains such a file returns an error.


The following example transcribes the audio files represented by the audio table:

Create the model:

# Create model
REMOTE WITH CONNECTION `myproject.myregion.myconnection`
OPTIONS (remote_service_type = 'CLOUD_AI_SPEECH_TO_TEXT_V2',
speech_recognizer = 'projects/project_number/locations/recognizer_location/recognizer/recognizer_id');

Transcribe the audio files without overriding the recognizer's default configuration:

  MODEL `myproject.mydataset.transcribe_model`,
  TABLE ``

Transcribe the audio files and override the recognizer's default configuration:

  MODEL `myproject.mydataset.transcribe_model`,
  TABLE ``,
  recognition_config => ( JSON '{"language_codes": ["en-US" ],"model": "telephony","auto_decoding_config": {}}')

The result is similar to the following:

transcripts ml_transcribe_result ml_transcribe_status uri ...
OK Google stream stranger things from Netflix to my TV. Okay, stranger things from Netflix playing on t v smart home and it's just... {"metadata":{"total_billed_duration":{"seconds":56}},"results":[{"alternatives":[{"confidence":0.738729,"transcript"... gs://mybucket/audio_files

What's next