Issue with Google Speech to Text API for audio files exceeding one minute

Other Questions API
1

I’m attempting to transcribe an audio file with specific characteristics using the Google Speech to Text API. Here are the file details:

  1. Format: Raw Audio
  2. Sample Rate: 16000 Hz
  3. Bit Depth: 16
  4. Channel: Mono

I implemented the following Python code to generate the text:

request = api.speech().async_recognize(
    data={
        'configuration': {
            'format': 'LINEAR16',  # raw 16-bit signed little-endian samples
            'sample_rate_hertz': 16000,  # 16 kHz
            'language': 'en-US',  # BCP-47 language tag
        },
        'recording': {
            'uri':'gs://somebucket/audio.raw'
            }
        })
result = request.execute()
print(json.dumps(result))

The code seems to function correctly, but it appears that the transcription only processes one minute of audio, disregarding any additional content. Can anyone explain this issue?

2

I encountered a similar problem where longer audio files were getting cut short. The crux of the issue might relate to the specific limits on audio duration enforced by the API. Google’s Speech-to-Text API imposes certain restrictions on maximum audio lengths for synchronous and asynchronous requests. While synchronous requests have a short duration limit, asynchronous requests should handle files longer than a minute without trimming. Make sure you are really using asynchronous processing. Moreover, check if your bucket permissions or a temporary network blip could be affecting the complete upload of the audio file.