subtitles.py

import os
import re
from rev_ai import apiclient

def timestamp2frame(ts, fps=10):
    ts = [int(x) for x in ts]
    frame = (ts[0] * 3600 + ts[1] * 60 + ts[2] + ts[3] / 1000) * fps
    return str(int(frame))

def audio2subtitle(rev_ai_token, fps=10):
    speech_file_path = "audio_out.mp3"
    client = apiclient.RevAiAPIClient(rev_ai_token)
    job = client.submit_job_local_file(speech_file_path)
    job_details = client.get_job_details(job.id)

    while str(job_details.status) != "JobStatus.TRANSCRIBED":
        job_details = client.get_job_details(job.id)
    transcript_srt = client.get_captions(job.id)

    with open("audio_out.srt", "w") as f:
        f.write(transcript_srt)
    f.close()

    timestamp_re = r"(\d+:\d+:\d+,\d+)"
    string_re = r"^[a-zA-Z][^>]+$"
    last_num_re = r"\d+"
    last_timestamp = ""

    deforum_str = ""

    with open("audio_out.srt", "r") as f:
        for line in f:
            timestamps = re.findall(timestamp_re, line)
            if timestamps:
                timestamp = re.split(":|,", timestamps[0])
                last_timestamp = re.split(":|,", timestamps[1])
                deforum_str += timestamp2frame(timestamp)
                deforum_str += ": "
            strings = re.findall(string_re, line)
            if strings:
                deforum_str += line
                deforum_str += " | "
    f.close()

    deforum_str = deforum_str.replace("\n", "").strip()
    deforum_str = re.sub(r"\|\s+(?=[a-zA-Z])", r"", deforum_str)
    deforum_str = deforum_str[:-1]

    last_timestamp = timestamp2frame(last_timestamp)

    return deforum_str, int(last_timestamp)