How to output JSON file?

[IkeDoku]

I am following https://jaimeleal.github.io/how-to-speech-synthesis to create a voice synthesis dataset. The tutorial uses Amazon Transcribe and outputs a JSON and the JSON files are then use to create a metadata.csv and filelists. My question: How can I output a JSON file directly from Whisper?

[jongwook]

If you are using the command line, you can edit transcribe.py to add something like:

import json

...

        # (after line 303)
        # save JSON
        with open(os.path.join(output_dir, audio_basename + ".json"), "w", encoding="utf-8") as f:
            json.dump(result, f)

[besimali]

Alternatively, you can try our whisper asr webservice which returns the json file. You can spin it up as a docker image and use the REST API for inference.

[IkeDoku]

Great :-) I will try it out soon.

[IkeDoku]

I used @jongwook suggestion. Works fine! :-)

[IkeDoku]

How should one proceed with the already output .txt, .SRT and .vtt files? Ideally, you would not transcribe all the audio files again (couple of thousands) to generate JSON files.

There are suggestions at https://stackoverflow.com/questions/11265575/converting-text-to-json/11265677 using Gelatin, Java or Go.

[jongwook]

There are some information in the result object that is not saved in txt/srt/vtt, like average log probability and the temperature applied for decoding. So in case you need those, you'll need to re-run the script :'(

[IkeDoku]

Ok, I guess I'll take the safe road. Thanks again :-)

[IkeDoku]

As I mentioned before I following @JaimeLeal (https://jaimeleal.github.io/how-to-speech-synthesis) blog tutorial and now stuck at extracting text the transcriptions (Step 4: metadata.csv and filelists). The tutorial uses Amazon Transcribe. Unfortunately, I wasn't able to reach the author that's why I trying it here. The code for extraction looks like this:
`import os
import json
import pandas as pd
from sklearn.model_selection import train_test_split
import numpy as np

path = '/dataset/transcripts/'
files = os.listdir(path)

Extract the transcript text from the output of Amazon Transcribe

rows = []
for index in range(1, len(files) + 1):
name = '{:04d}'.format(index)
with open(os.path.join(path, "{}.json".format(name))) as json_file:
data = json.load(json_file)
transcript = data["results"]["transcripts"][0]["transcript"]
rows.append([name, transcript])

data = pd.DataFrame(rows, columns = ["name", "transcript"])`

At line data = json.load(json_file) it returns
in <module> transcript = data["results"]["transcripts"][0]["transcript"] KeyError: 'results'

As far as I understood it tries to fetch the data from a url. Which I don't have ofc. But then, how can I make it work in my case?