transcript/transcript.py

from yt_dlp import YoutubeDL
from moviepy.editor import AudioFileClip
from faster_whisper import WhisperModel
import subprocess
from pyngrok import ngrok
from flask import Flask, render_template, request, jsonify
import os

app = Flask(__name__)

def get_output_webm_paths(folder_path):
    
    for root, dirs, files in os.walk(folder_path):
        for file in files:
            if file.startswith("output.webm"):
                file_path = os.path.join(root, file)
    
                return file_path

def transcribe_video(video_path):
  youtube_link=video_path
  ydl_opts = {
      'outtmpl': 'output.webm',
  }
  with YoutubeDL(ydl_opts) as ydl:
      ydl.download([youtube_link])
  output_webm_path = get_output_webm_paths("/content/drive/MyDrive/Prometheus_RVC/transcript")
  audio = AudioFileClip(output_webm_path)
  audio.write_audiofile("/content/drive/MyDrive/Prometheus_RVC/transcript/output.wav")
  AUDIO_FILE_NAME = "/content/drive/MyDrive/Prometheus_RVC/transcript/output.wav"
  model_size = "medium"
  model = WhisperModel(model_size, device="cuda", compute_type="float16")
  segments, info = model.transcribe(AUDIO_FILE_NAME, beam_size=5)
  output_file = "/content/drive/MyDrive/Prometheus_RVC/transcript/transcription.txt"  # 出力ファイル名
  with open(output_file, "w") as file:
      for segment in segments:
          file.write("%s\n" % segment.text)
  text = open(output_file, "r").read()
  os.remove(output_webm_path)
  os.remove("/content/drive/MyDrive/Prometheus_RVC/transcript/output.wav")
  return text

@app.route('/transcript', methods=['POST'])
def transcript():
    data = request.get_json()
    video_url = data['video_url']
    transcript = transcribe_video(video_url)
    return jsonify({'transcript': transcript})

if __name__ == "__main__":
    ngrok_tunnel = ngrok.connect(5000)
    http_tunnel = ngrok.connect(addr="5000", proto="http", hostname="measured-smashing-piglet.ngrok-free.app")
    print(f"ngrok tunnel URL: {ngrok_tunnel.public_url}")
    app.run()