Use models

Preview

This feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the Service Specific Terms. Pre-GA features are available "as is" and might have limited support. For more information, see the launch stage descriptions.

Use a trained Custom Speech-to-Text model in your production application or benchmarking workflows. As soon as you deploy your model through a dedicated endpoint, you automatically get programmatic access through a recognizer object, which can be used directly through the Speech-to-Text V2 API or in the Google Cloud console.

Before you begin

Ensure you have signed up for a Google Cloud account, created a project, trained a custom speech model, and deployed it using an endpoint.

Perform inference in V2

For a Custom Speech-to-Text model to be ready for use, the state of the model in the Models tab should be Active, and the dedicated endpoint in the Endpoints tab must be Deployed.

In our example, where a Google Cloud project ID is custom-models-walkthrough, the endpoint that corresponds to the Custom Speech-to-Text model quantum-computing-lectures-custom-model is quantum-computing-lectures-custom-model-prod-endpoint. The region that it's available is us-east1, and the batch transcription request is the following:

from google.api_core import client_options
from google.cloud.speech_v2 import SpeechClient
from google.cloud.speech_v2.types import cloud_speech

def quickstart_v2(
    project_id: str,
    audio_file: str,
) -> cloud_speech.RecognizeResponse:
    """Transcribe an audio file."""
    # Instantiates a client
    client = SpeechClient(
    client_options=client_options.ClientOptions(
      api_endpoint="us-east1-speech.googleapis.com"
    )
  )

    # Reads a file as bytes
    with open(audio_file, "rb") as f:
        content = f.read()

    config = cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        language_codes=["en-US"],
        model="projects/custom-models-walkthrough/locations/us-east1/endpoints/quantum-computing-lectures-custom-model-prod-endpoint",
    )
    request = cloud_speech.RecognizeRequest(
        recognizer=f"projects/custom-models-walkthrough/locations/us-east1/recognizers/_",
        config=config,
        content=content,
    )

    # Transcribes the audio into text
    response = client.recognize(request=request)

    for result in response.results:
        print(f"Transcript: {result.alternatives[0].transcript}")

    return response

Note: If you try to create a recognizer object in a different region than the one that the endpoint is created in, the request will fail.

What's next

Follow the resources to take advantage of custom speech models in your application. See Evaluate your custom models.