python – Why do google cloud ml ml voice services cut the transcript in a minute?

I'm trying to use nlp voice services from gcloud to create a transcript of a 3 minute video. Since voice services require files to be less than 10 MB, I divide the video into two overlapping videos and convert them to flac format. However, although it uses the following command to call a long-term asynchronous request, it is always disabled by the minute.

                id_a =! gcloud ml speech long-term recognition $ uri_1 --async --language-code = "en-US" --sample-rate = 48000
id_b =! gcloud ml speech recognize-long duration $ uri_2 --async --language-code = "en-US" --sample-rate = 48000

id_a = int (json.loads ("". join (id_a[1:4]))['name'])
id_b = int (json.loads ("". join (id_b[1:4]))['name'])

print ("pending completion")
writings_1 =! gcloud ml voice operations wait $ id_a
writings_2 =! gcloud ml voice operations wait $ id_b

This should use an asynchronous request for long-term voice services. However, the transcripts I receive are always cut off at minute 1, although they are longer, do you have an idea of ​​what's wrong? In addition, I call the application for a Jupyter notebook running on my compute instance.