Update app.py

app.py

@@ -162,6 +162,8 @@ def load(device):
     #        del checkpoint['model_state_dict'][key]
     e2tts.load_state_dict(checkpoint['model_state_dict'], strict=False)
     
+    del checkpoint
+    
     e2tts.vocos = EncodecWrapper("facebook/encodec_24khz")
     for param in e2tts.vocos.parameters():
         param.requires_grad = False
@@ -198,6 +200,7 @@ def load(device):
 
 
 e2tts, stft = load(device)
+e2tts = e2tts
 
 
 def run(e2tts, stft, arg1, arg2, arg3, arg4):
@@ -288,7 +291,7 @@ def run(e2tts, stft, arg1, arg2, arg3, arg4):
             print("paths", video_path, audio_path, video_path_gen)
             return video_path_gen
     except Exception as e:
-        print("Exception", video_path, e)
+        print("Exception", e)
         traceback.print_exc()
         
         if False:
@@ -311,6 +314,13 @@ def video_to_audio(video: gr.Video, prompt: str, num_steps: int):
     
     video_path = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4').name
     
+    if video.startswith("http"):
+        data = requests.get(video, timeout=60).content
+        with open(video_path, "wb") as fw:
+            fw.write(data)
+    else:
+        shutil.copy(video, video_path)
+    
     video_save_path = run(e2tts, stft, video_path, prompt, len(prompt)==0, num_steps)
     
     return video_save_path
@@ -320,6 +330,13 @@ def video_to_piano(video: gr.Video, prompt: str, num_steps: int):
     
     video_path = tempfile.NamedTemporaryFile(delete=False, suffix='.mp4').name
     
+    if video.startswith("http"):
+        data = requests.get(video, timeout=60).content
+        with open(video_path, "wb") as fw:
+            fw.write(data)
+    else:
+        shutil.copy(video, video_path)
+    
     video_save_path = run(e2tts, stft, video_path, prompt, len(prompt)==0, num_steps)
     
     return video_save_path