run_for_gradio

app.py

@@ -41,7 +41,7 @@ os.makedirs("pretrained/v2a/mmaudio", exist_ok=True)
 setup_eval_logging()
 pipeline = Pipeline(
     step0_model_dir=repo_local_path, 
-    step1_mode='mmaudio_small_44k', 
+    step1_mode='mmaudio_medium_44k', 
     step2_model_dir=repo_local_path,
     step2_mode='cot',
     step3_mode='bs_roformer',

pipeline/pipeline.py

@@ -89,7 +89,7 @@ class Pipeline:
                         "temp_final_video_path": None, 
                         'log': ''}
 
-        step0_resp = self.step0.run(video_input)
+        step0_resp = self.step02.run_step0(video_input)
         step0_resp_list = re.findall(r'(Step\d:.*?)(?=Step\d:|$)', step0_resp, re.DOTALL)
         step_infos = [step_info.strip().split("\n")[0] for step_info in step0_resp_list]
         step3_temp_dir = os.path.join(output_dir, "remove_vo")
@@ -108,7 +108,7 @@ class Pipeline:
                 yield step_results
 
             elif step_info == 'Step2: Given a video and its generated audio, determine whether the audio contains voice-over.':
-                is_vo = self.step2.run(str(step_results["step1_video_path"]))
+                is_vo = self.step02.run_step2(str(step_results["step1_video_path"]))
                 step_results["is_vo"] = is_vo
                 step_results['log'] = f"Step2 completed. Contain voice-over? {'Yes' if is_vo else 'No'}"
                 yield step_results