Update app.py

app.py

@@ -90,6 +90,7 @@ def calculate(image_in, audio_in):
     torchaudio.save("/content/audio.wav", waveform, sample_rate, encoding="PCM_S", bits_per_sample=16)
     image = Image.open(image_in)
     image = pad_image(image)
+    os.system(f"rm -rf /content/image.png")
     image.save("image.png")
 
     pocketsphinx_run = subprocess.run(['pocketsphinx', '-phone_align', 'yes', 'single', '/content/audio.wav'], check=True, capture_output=True)
@@ -97,6 +98,7 @@ def calculate(image_in, audio_in):
     with open("test.json", "w") as f:
         f.write(jq_run.stdout.decode('utf-8').strip())
     # device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+    os.system(f"rm -rf /content/image_audio.mp4")
     os.system(f"cd /content/one-shot-talking-face && python3 -B test_script.py --img_path /content/image.png --audio_path /content/audio.wav --phoneme_path /content/test.json --save_dir /content/train")
     return "/content/train/image_audio.mp4"
 
@@ -116,7 +118,7 @@ def merge_frames():
   filenames.sort() # this iteration technique has no built in order, so sort the frames
   print(filenames)
   images = list(map(lambda filename: imageio.imread("/content/video_results/restored_imgs/"+filename), filenames))
-
+  os.system(f"rm -rf /content/video_output.mp4")
   imageio.mimsave('/content/video_output.mp4', images, fps=25.0) # modify the frame duration as needed
   return "/content/video_output.mp4"
 
@@ -140,6 +142,8 @@ def one_shot_talking(image_in,audio_in):
   os.system(f"python /content/GFPGAN/inference_gfpgan.py --upscale 2 -i /content/image_pre.png -o /content/results  --bg_upsampler realesrgan")
   # time.sleep(60)
   image_in_one_shot='/content/results/restored_imgs/image_pre.png'
+  os.system(f"rm -rf /content/results/restored_imgs/image_pre.png")
+  
   #One Shot Talking Face algorithm
   calculate(image_in_one_shot,audio_in)
 
@@ -149,6 +153,8 @@ def one_shot_talking(image_in,audio_in):
   os.system(f"python /content/PyVideoFramesExtractor/extract.py --video=/content/train/image_audio.mp4")
 
   #2. Improve image quality using GFPGAN on each frames
+  os.system(f"rm -rf /content/extracted_frames/image_audio_frames")
+  os.system(f"rm -rf /content/video_results/")
   os.system(f"python /content/GFPGAN/inference_gfpgan.py --upscale 2 -i /content/extracted_frames/image_audio_frames -o /content/video_results  --bg_upsampler realesrgan")
 
   #3. Merge all the frames to a one video using imageio