Postgres + Sqlalchemy double counting results

Question

I am encountering duplicate results in my sqlalchemy+postgres query. I have a model structured like below:

class Audio(db.Model):
    __tablename__ = "audio"

    id = db.Column(db.Integer, primary_key=True)
    file_location = db.Column(db.String(), unique=True)
    upload_time = db.Column(ArrowType, default=arrow.utcnow())
    keyword = db.Column(JSON)
    transcript = db.Column(JSON)
    diarization = db.Column(JSON)

    user_id = db.Column(db.Integer, db.ForeignKey('user.id', onupdate='cascade', ondelete='cascade'))
    company_id = db.Column(db.Integer, db.ForeignKey('company.id', onupdate='cascade', ondelete='cascade'))
    client_id = db.Column(db.Integer, db.ForeignKey('client.id', onupdate='cascade', ondelete='cascade'))

    def __init__(self, file_location, upload_time, keyword, transcript, diarization, user_id, company_id, client_id):
        self.file_location = file_location
        self.upload_time = upload_time
        self.keyword = keyword
        self.transcript = transcript
        self.diarization = diarization
        self.user_id = user_id
        self.company_id = company_id
        self.client_id = client_id

    def __repr__(self):
        return '<Audio ID %r>' % self.id

And under transcript, the input is a json object:

{"transcript": [
    {"p": 0, "s": 0, "e": 320, "c": 0.545, "w": "This"}, 
    {"p": 1, "s": 320, "e": 620, "c": 0.825, "w": "call"}, 
    {"p": 2, "s": 620, "e": 780, "c": 0.909, "w": "is"}, 
    {"p": 3, "s": 780, "e": 1010, "c": 0.853, "w": "being"}
    ...
    ]}

I am trying to filter entries based on the value from "w" and query its corresponding Audio.id and "p". I have tried the following:

  transcript_subquery = s.query(func.json_array_elements(Audio.transcript['transcript']).label('transcript')).subquery()
    temp = transcript_subquery.c.transcript.op('->>')('w').cast(String)
    temp1 = transcript_subquery.c.transcript.op('->>')('s').cast(Integer)
    query = s.query(temp1, Audio.id).filter(temp.ilike("all"))

The problem that I have is that right now I only have 5 data entries and one of the data entry's transcript field is blank. However, I got a result that contains the following:

(665610, 5), (736413, 5), (907230, 5), (942340, 5), (1020852, 5), (1023942, 5), (1037101, 5), (1078521, 5), (1105581, 5), (1117551, 5), (1372730, 5), (1501960, 5), (1508410, 5)

This is really odd to be given that the entry where Audio.id =5 doesn't even have transcript json object yet it produces such result. Moreover, all the entries, each has 112 results. And I suspect that 112 is the total number that the word appears and it somewhat is double counting everything for each Audio ID. And I don't really know how to fix my sqlalchemy query.