audio preprocessing enhancement. fix dirs. add zip archive response.

cli.py

@@ -2,6 +2,7 @@ from argparse import ArgumentParser
 from utils.download_video import download_video
 from utils.transcriber import transcriber
 from utils.subtitler import subtitler
+from utils.convert_mp4_to_mp3 import convert_mp4_to_mp3
 import logging, os
 from tqdm import tqdm
 
@@ -20,26 +21,32 @@ def main(video_url:str,
         text_color:str
         ):
     INVIDEO_DIR = os.path.join('data/',invideo_filename)
-    if not os.path.exists(INVIDEO_DIR):
-        os.makedirs(INVIDEO_DIR)
+    os.makedirs(INVIDEO_DIR, exist_ok=True)
     SRT_PATH = os.path.join(INVIDEO_DIR, f"{invideo_filename}.srt")
     OUTVIDEO_PATH = os.path.join(INVIDEO_DIR, f"result_{invideo_filename}.mp4")
     with tqdm(total=100, desc="Overall Progress") as pbar:       
         if video_url != None:
-            stream_title = download_video(video_url, INVIDEO_DIR, invideo_filename)
-            pbar.update(33.33)
+            INVIDEO_PATH = download_video(video_url, INVIDEO_DIR, invideo_filename)
+            pbar.update(25)
+            INAUDIO_PATH = os.path.join(INVIDEO_DIR, f"{invideo_filename}.mp3")
+            convert_mp4_to_mp3(INVIDEO_PATH,INAUDIO_PATH)
+            pbar.update(25)
             if not os.path.exists(SRT_PATH):
-                transcriber(stream_title, SRT_PATH, max_words_per_line)
-            pbar.update(33.33)
-            subtitler(stream_title, SRT_PATH, OUTVIDEO_PATH,fontsize, font, bg_color, text_color)
-            pbar.update(33.34)
+                transcriber(INAUDIO_PATH, SRT_PATH, max_words_per_line)
+            pbar.update(25)
+            subtitler(INVIDEO_PATH, SRT_PATH, OUTVIDEO_PATH,fontsize, font, bg_color, text_color)
+            pbar.update(25)
             return
         INVIDEO_PATH = os.path.join(INVIDEO_DIR, f"{invideo_filename}.mp4")
+        INAUDIO_PATH = os.path.join(INVIDEO_DIR, f"{invideo_filename}.mp3")
+        if not os.path.exists(INAUDIO_PATH):
+            convert_mp4_to_mp3(INVIDEO_PATH,INAUDIO_PATH)
+        pbar.update(50)
         if not os.path.exists(SRT_PATH):
-            transcriber(INVIDEO_PATH, SRT_PATH, max_words_per_line)
-        pbar.update(66.66)
+            transcriber(INAUDIO_PATH, SRT_PATH, max_words_per_line)
+        pbar.update(25)
         subtitler(INVIDEO_PATH, SRT_PATH, OUTVIDEO_PATH, fontsize, font, bg_color, text_color)
-        pbar.update(33.34)
+        pbar.update(25)
 
 if __name__ == '__main__':
     parser = ArgumentParser()

main.py

@@ -2,7 +2,15 @@ from fastapi import FastAPI, UploadFile, File, HTTPException, Form
 from fastapi.responses import FileResponse, HTMLResponse
 from typing import Optional
 from utils.process_video import process_video
-import shutil, os
+from utils.zip_response import zip_response
+import shutil, os, logging
+
+logging.basicConfig(filename='main.log',
+                encoding='utf-8',
+                level=logging.DEBUG,
+                format='%(asctime)s %(levelname)s %(message)s',
+                datefmt='%m/%d/%Y %I:%M:%S %p')
+
 
 app = FastAPI()
 
@@ -42,17 +50,19 @@ async def process_video_api(video_file: UploadFile = File(...),
     try:
         if not str(video_file.filename).endswith('.mp4'):
             raise HTTPException(status_code=400, detail="Invalid file type. Please upload an MP4 file.")
-        # Create temp dir
+        logging.info("Creating temporary directories")
         temp_dir = os.path.join(os.getcwd(),"temp")
         os.makedirs(temp_dir, exist_ok=True)
-        temp_input_path = os.path.join(temp_dir, video_file.filename)
-        # Copy video UploadFile to the temp_input_path
+        temp_vid_dir = os.path.join(temp_dir,video_file.filename.split('.')[0])
+        os.makedirs(temp_vid_dir, exist_ok=True)
+        temp_input_path = os.path.join(temp_vid_dir, video_file.filename)
+        logging.info("Copying video UploadFile to the temp_input_path")
         with open(temp_input_path, 'wb') as buffer:
             try:
                 shutil.copyfileobj(video_file.file, buffer)
             finally:
                 video_file.file.close()
-        # Copy srt UploadFile to the temp_input_path
+        logging.info("Copying SRT UploadFile to the temp_input_path")
         if srt_file.size > 0:
             SRT_PATH = os.path.abspath(f"{temp_input_path.split('.')[0]}.srt")
             with open(SRT_PATH, 'wb') as buffer:
@@ -60,14 +70,16 @@ async def process_video_api(video_file: UploadFile = File(...),
                     shutil.copyfileobj(srt_file.file, buffer)
                 finally:
                     srt_file.file.close()
-            # Process the video
+            logging.info("Processing the video...")
             output_path = process_video(temp_input_path, SRT_PATH, max_words_per_line, fontsize, font, bg_color, text_color)
-            return FileResponse(output_path, media_type="video/mp4", filename=f"result_{video_file.filename}")
-        # Process the video
+            logging.info("Archiving response...")
+            zip_path = zip_response(os.path.join(temp_vid_dir,"archive.zip"), [SRT_PATH, output_path])
+            return FileResponse(zip_path, media_type='application/zip', filename=f'result_{video_file.filename}.zip')
+        logging.info("Processing the video...")
         output_path = process_video(temp_input_path, None, max_words_per_line, fontsize, font, bg_color, text_color)
-        # FileResponse(output_path, media_type="text/srt", filename=f"result_{video_file.filename.split('.')[0]}.srt")
-        # Return the processed video file
-        return FileResponse(output_path, media_type="video/mp4", filename=f"result_{video_file.filename}")
+        logging.info("Archiving response...")
+        zip_path = zip_response(os.path.join(temp_vid_dir,"archive.zip"), [SRT_PATH, output_path])
+        return  FileResponse(zip_path, media_type='application/zip', filename=f'result_{video_file.filename}.zip')
                 
     except Exception as e:
         raise HTTPException(status_code=500, detail=str(e))

utils/convert_mp4_to_mp3.py

@@ -0,0 +1,16 @@
+from moviepy.editor import VideoFileClip
+
+def convert_mp4_to_mp3(mp4_file_path, mp3_file_path):
+    # Load the video file
+    video_clip = VideoFileClip(mp4_file_path)
+
+    # Extract the audio from the video clip
+    audio_clip = video_clip.audio
+
+    # Save the audio clip as an MP3 file
+    audio_clip.write_audiofile(mp3_file_path)
+
+    # Close the clips
+    audio_clip.close()
+    video_clip.close()
+    return mp3_file_path

utils/process_video.py

@@ -1,6 +1,7 @@
 # Import necessary modules
 from utils.transcriber import transcriber
 from utils.subtitler import subtitler
+from utils.convert_mp4_to_mp3 import convert_mp4_to_mp3
 import logging, os
 
 # Set up logging
@@ -19,13 +20,16 @@ def process_video(invideo_filename:str,
                   bg_color:str,
                   text_color:str
                   ):
-    OUTVIDEO_PATH = os.path.join("temp/", "result.mp4")
+    VIDEO_NAME = invideo_filename.split('\\')[-1] 
+    OUTVIDEO_PATH = os.path.join(invideo_filename.split('\\')[-3], invideo_filename.split('\\')[-2], f"result_{VIDEO_NAME}")
     if srt_path:
         subtitler(invideo_filename, srt_path, OUTVIDEO_PATH, fontsize, font, bg_color, text_color)
         return OUTVIDEO_PATH
-    SRT_PATH = os.path.abspath(f"{invideo_filename.split('.')[0]}.srt")
+    INAUDIO_PATH = os.path.abspath(f"{invideo_filename.split('.')[0]}.mp3")
+    convert_mp4_to_mp3(invideo_filename, INAUDIO_PATH)
+    SRT_PATH = os.path.abspath(f"{invideo_filename.split('.')[0]}.srt") 
     if not os.path.exists(SRT_PATH):
-        transcriber(invideo_filename, SRT_PATH, max_words_per_line)
+        transcriber(INAUDIO_PATH, SRT_PATH, max_words_per_line)
     logging.info("Transcription Complete")
     subtitler(invideo_filename, SRT_PATH, OUTVIDEO_PATH, fontsize, font, bg_color, text_color)
     logging.info("Subtitling Complete")

utils/zip_response.py

@@ -0,0 +1,10 @@
+import zipfile, os
+
+def zip_response(archive_name: str, files: list):
+    # Create a new zip file
+    with zipfile.ZipFile(archive_name, 'w') as zipf:
+    # Add specific files to the zip file
+        for file in files:
+            zipf.write(file, arcname=os.path.basename(file))
+            zipf.write(file, arcname=os.path.basename(file))
+    return archive_name