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.

Syntax

ML.TRANSCRIBE(
  MODEL `PROJECT_ID.DATASET.MODEL_NAME`,
  TABLE `PROJECT_ID.DATASET.OBJECT_TABLE`,
  [RECOGNITION_CONFIG => ( JSON 'RECOGNITION_CONFIG')]
)

Arguments

ML.TRANSCRIBE takes the following arguments:

  • PROJECT_ID: the project that contains the resource.

  • DATASET: the BigQuery dataset that contains the resource.

  • MODEL_NAME: 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't specify a RECOGNITION_CONFIG value.

    If no recognizer has been specified for the remote model by using the SPEECH_RECOGNIZER option, you must specify a RECOGNITION_CONFIG value. This value is used to provide a configuration for the default recognizer. You can only use the chirp transcription model in the speech recognizer or RECOGNITION_CONFIG value that you provide.

Output

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.

Quotas

See Cloud AI service functions quotas and limits.

Known issues

This section contains information about known issues.

Resource exhausted errors

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.

Invalid argument errors

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

INVALID_ARGUMENT: The audio file cannot be processed in time.

This issue occurs because one of the audio files being processed is too long. Check your input audio files to make sure they are all 30 minutes or less.

Locations

You can run the ML.TRANSCRIBE function in the following locations:

  • asia-southeast1
  • europe-west4
  • us-central1
  • US
  • EU

ML.TRANSCRIBE must run in the same region as the remote model that the function references.

Limitations

The function can't process audio files that are longer than 30 minutes. Any row that contains such a file returns an error.

Example

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

Create the model:

# Create model
CREATE OR REPLACE MODEL
`myproject.mydataset.transcribe_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:

SELECT *
FROM ML.TRANSCRIBE(
  MODEL `myproject.mydataset.transcribe_model`,
  TABLE `myproject.mydataset.audio`
);

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

SELECT *
FROM ML.TRANSCRIBE(
  MODEL `myproject.mydataset.transcribe_model`,
  TABLE `myproject.mydataset.audio`,
  recognition_config => ( JSON '{"language_codes": ["en-US" ],"model": "chirp","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