Update app.py

app.py

@@ -152,19 +152,27 @@ def create_temp_input_json(prompt: str, cond_image_path: str, cond_audio_path_sp
 
 def infer(prompt, cond_image_path, cond_audio_path_spk1, cond_audio_path_spk2, sample_steps):
 
+    timestamp = datetime.now().strftime("%Y%m%d%H%M%S%f")
+    result_filename = f"meigen_multitalk_result_{sample_steps}_steps_{timestamp}"
+    temp_files_to_cleanup = []
+    
     if is_shared_ui:
-        
         trimmed_audio_path_spk1 = trim_audio_to_5s_temp(cond_audio_path_spk1)
-        cond_audio_path_spk1 = trimmed_audio_path_spk1
-        
+        if trimmed_audio_path_spk1 != cond_audio_path_spk1:
+            cond_audio_path_spk1 = trimmed_audio_path_spk1
+            temp_files_to_cleanup.append(trimmed_audio_path_spk1)
+
         if cond_audio_path_spk2 is not None:
             trimmed_audio_path_spk2 = trim_audio_to_5s_temp(cond_audio_path_spk2)
-            cond_audio_path_spk2 = trimmed_audio_path_spk2
-    
+            if trimmed_audio_path_spk2 != cond_audio_path_spk2:
+                cond_audio_path_spk2 = trimmed_audio_path_spk2
+                temp_files_to_cleanup.append(trimmed_audio_path_spk2)
+
     # Prepare input JSON
     input_json_path = create_temp_input_json(prompt, cond_image_path, cond_audio_path_spk1, cond_audio_path_spk2)
-
-    # Base args
+    temp_files_to_cleanup.append(input_json_path)
+    
+# Base args
     common_args = [
         "--ckpt_dir", "weights/Wan2.1-I2V-14B-480P",
         "--wav2vec_dir", "weights/chinese-wav2vec2-base",
@@ -172,7 +180,7 @@ def infer(prompt, cond_image_path, cond_audio_path_spk1, cond_audio_path_spk2, s
         "--sample_steps", str(sample_steps),
         "--mode", "streaming",
         "--use_teacache",
-        "--save_file", "multi_long_multigpu_exp"
+        "--save_file", result_filename
     ]
 
     if num_gpus > 1:
@@ -210,9 +218,16 @@ def infer(prompt, cond_image_path, cond_audio_path_spk1, cond_audio_path_spk2, s
         if process.returncode != 0:
             raise RuntimeError("Inference failed. Check inference.log for details.")
 
-        return "multi_long_multigpu_exp.mp4"
+        return f"{result_filename}.mp4"
 
-    
+    finally:
+        for f in temp_files_to_cleanup:
+            try:
+                if os.path.exists(f):
+                    os.remove(f)
+                    print(f"[INFO] Removed temporary file: {f}")
+            except Exception as e:
+                print(f"[WARNING] Could not remove {f}: {e}")  
 
 
 with gr.Blocks(title="MultiTalk Inference") as demo: