Recognizer on northamerica-northeast1

Google Cloud Build with AI AI Infrastructure
1

When running a transcription the google cloud platform on northamerica-northeast1 with chirp_3
I get this following error

Error running recognize request. Error on retrieving SLM instance:Resource ‘projects/cloud-stt-cm-prod/locations/northamerica-northeast1/instances/310504136861-custom_models’ was not found

image
image2267×1067 152 KB

Is this normal?

1 Like
2

Hi … Let’s solve the problem.**

Observations:**

  • Is this normal? No, this is likely a backend infrastructure issue or a regional limitation. The error refers to a missing internal Resource (SLM/Speech Language Modeling instance) within the cloud-stt-cm-prod project, which is managed by Google, not the user.

  • Context: This error often occurs when the specific regional endpoint for Chirp is experiencing a synchronization issue or if the “Model Adaptation” features are pointing to a non-existent internal resource.

Recommended Troubleshooting Steps:

  1. Switch Regions: Try running the request in us-central1. This region is typically the most stable for Chirp and Speech-to-Text v2 features. If it works there, the issue is confirmed to be specific to the northamerica-northeast1 deployment.

  2. Check Recognizer Configuration: If you are using the STT v2 API, ensure you are not using a “Recognizer” that was configured with a deleted or incompatible custom model. Try creating a fresh Recognizer in the console without any model adaptations.

  3. Use the SDK instead of the Console: Sometimes the GCP Console UI experiences glitches when rendering v2 configurations. Testing via the Python SDK or gcloud CLI can often bypass UI-specific bugs.

Sample Python Test:

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

# Test with a different region if Montreal continues to fail
client = SpeechClient()
request = cloud_speech.RecognizeRequest(
    recognizer="projects/[PROJECT_ID]/locations/us-central1/recognizers/_",
    config=cloud_speech.RecognitionConfig(
        auto_decoding_config=cloud_speech.AutoDetectDecodingConfig(),
        model="long", # Chirp
        language_codes=["en-US"],
    ),
    uri="gs://[BUCKET]/[FILE].wav",
)

Has anyone else noticed regional instability with Chirp in northamerica-northeast1 recently?

thanks …

3

Hello @Émile,

From the documentation about Regional availability, northamerica-northeast1 is in preview, your project may not be able to access it.

US and EU are GA (General Availability). I recommend that you switch to US.

1 Like