Update app.py

app.py

@@ -13,6 +13,8 @@ import cv2
 import imageio
 import os
 import ffmpeg
+from io import BytesIO
+import requests
 
 from fairseq.checkpoint_utils import load_model_ensemble_and_task_from_hf_hub
 from fairseq.models.text_to_speech.hub_interface import TTSHubInterface
@@ -88,7 +90,14 @@ def calculate(image_in, audio_in):
     waveform, sample_rate = torchaudio.load(audio_in)
     waveform = torch.mean(waveform, dim=0, keepdim=True)
     torchaudio.save("/content/audio.wav", waveform, sample_rate, encoding="PCM_S", bits_per_sample=16)
-    image = Image.open(image_in)
+    if os.path.exists(image_in):
+        image = Image.open(image_in)
+    else:
+        print("image not exists reading web image")
+        image_url = "http://labelme.csail.mit.edu/Release3.0/Images/users/DNguyen91/face/m_unsexy_gr.jpg"
+        response = requests.get(image_url)
+        image = Image.open(BytesIO(response.content))
+        
     image = pad_image(image)
     # os.system(f"rm -rf /content/image.png")
     image.save("image.png")
@@ -144,7 +153,7 @@ def one_shot_talking(image_in,audio_in):
     os.makedirs("/content/results")
   
   #Improve quality of input image
-  os.system(f"python /content/GFPGAN/inference_gfpgan.py --upscale 2 -i /content/image_pre.png -o /content/results  --bg_upsampler realesrgan")
+  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/image_pre.png'