使用谷歌语音转文本时遇nl-NL语言不支持视频模型报错求助

Hey there, let's break this down and fix it step by step! That error message gives us the key clue: the video model the demo page is defaulting to doesn't support Dutch (nl-NL). Google's Speech-to-Text has different model types optimized for specific use cases, and the video model has a much smaller list of supported languages compared to standard audio-focused models.

Here's how to resolve this issue: #

• Switch to a Dutch-compatible model
  The default model (optimized for general audio transcription) fully supports Dutch, along with others like command_and_search or phone_call depending on your use case. If the web demo isn't letting you explicitly choose the model, it might be auto-selecting the video model for longer audio clips. To get around this, try using the official API directly (via tools like curl or the Google Cloud SDK) where you can specify the model manually.

• Verify your audio file formatting
  Even though you exported WAV/MP3/FLAC, double-check that your files meet Google's recommended specs:

  • Preferred encoding: FLAC or 16-bit linear PCM WAV
  • Sample rate: 16kHz (the most widely supported rate for Speech-to-Text)
  • Channel count: Mono (single channel)
    Mismatched formatting can sometimes trigger the demo to fall back to unsupported models by mistake.

• Use asynchronous transcription for longer audio
  If your audio is over 1 minute long, the demo's synchronous transcription might hit limits and default to the wrong model. Asynchronous transcription (using the longrunningrecognize endpoint) is built for longer files, supports more language-model combinations, and avoids this issue entirely.

Example curl command to test with the correct model: #

# First get a valid access token with: gcloud auth application-default print-access-token
curl -H "Content-Type: application/json" \
     -H "Authorization: Bearer YOUR_ACCESS_TOKEN" \
     https://speech.googleapis.com/v1/speech:recognize \
     -d '{
         "config": {
             "encoding": "FLAC",
             "sampleRateHertz": 16000,
             "languageCode": "nl-NL",
             "model": "default",
             "enableAutomaticPunctuation": true
         },
         "audio": {
             "uri": "gs://your-cloud-storage-bucket/your-audio-file.flac"
         }
     }'

Quick extra tips: #

• If you want to stick with the web demo, try uploading a shorter audio clip (under 1 minute) — it's more likely to use the default model instead of the video model.
• Ensure your Google Cloud project has the Speech-to-Text API enabled (though this error is specifically model-related, not a permissions/quota issue).

内容的提问来源于stack exchange，提问作者Frederik Wouters