Update app.py

app.py

@@ -32,6 +32,7 @@ def set_tts():
 def create_kd_talker():
     return Inferencer() 
 
+@spaces.GPU
 def predict(prompt, upload_reference_audio, microphone_reference_audio, reference_audio_type):
     global result_dir
     output_file_path = os.path.join(result_dir, 'output.wav')
@@ -49,6 +50,10 @@ def predict(prompt, upload_reference_audio, microphone_reference_audio, referenc
     )
     return gr.Audio(value=output_file_path, type='filepath')
 
+@spaces.GPU
+def generate(upload_driven_audio, tts_driven_audio, driven_audio_type, source_image, smoothed_pitch, smoothed_yaw, smoothed_roll, smoothed_t):
+    return Inferencer().generate_with_audio_img(upload_driven_audio, tts_driven_audio, driven_audio_type, source_image,
+                                    smoothed_pitch, smoothed_yaw, smoothed_roll, smoothed_t)
 
 
 @spaces.GPU
@@ -108,9 +113,8 @@ def main():
                         smoothed_roll = gr.Slider(minimum=0, maximum=1, step=0.1, label="Roll", value=0.8)
                         smoothed_t = gr.Slider(minimum=0, maximum=1, step=0.1, label="T", value=0.8)
                         kd_submit = gr.Button("Generate", elem_id="kdtalker_generate", variant="primary")
-                        kdtalker = Inferencer()
                         kd_submit.click(
-                                fn=kdtalker.generate_with_audio_img,
+                                fn=generate,
                                 inputs=[
                                     upload_driven_audio, tts_driven_audio, driven_audio_type, source_image,
                                     smoothed_pitch, smoothed_yaw, smoothed_roll, smoothed_t