Optimize Amazon Comprehend Analysis Job use

I have an audio recording file, and I want to extract my custom entities recognise data from the audio file. I searched and noticed Amazon Comprehend service could help me.

To do this, I have to do the following steps:

1. I should submit to server and transcribe the file using AWS Transcribe service - it takes about 15~20 seconds.
2. I should get the transcribed text and save it to s3 bucket for the Comprehend Analysis Job, because when we create a job, they ask it as input file.
3. I should create an Analysis Job with the input file.
4. I should wait till the job is finished - it is very slow (about 2 ~ 7 minutes).
5. After the job is finished, the output file is output.tar.gz. They give us compressed file, not plain text.
6. I should pull the into the local server and the unzip the file and then get the content.
7. I should parse the file content as json data.

It takes about 5~15 minutes to do the whole steps. Especially step 4 takes pretty much times. I want to optimize it as much as possible. Can you please help me?