diff --git a/webui.py b/webui.py
index e437414..3552cd9 100644
--- a/webui.py
+++ b/webui.py
@@ -147,7 +147,7 @@ def main():
         with gr.Row():
             mode_checkbox_group = gr.Radio(choices=inference_mode_list, label='选择推理模式', value=inference_mode_list[0])
             instruction_text = gr.Text(label="操作步骤", value=instruct_dict[inference_mode_list[0]], scale=0.5)
-            sft_dropdown = gr.Dropdown(choices=sft_spk, label='选择预训练音色', value=sft_spk[0] if len(sft_spk) != 0 else '', scale=0.25)
+            sft_dropdown = gr.Dropdown(choices=sft_spk, label='选择预训练音色', value=sft_spk[0], scale=0.25)
             stream = gr.Radio(choices=stream_mode_list, label='是否流式推理', value=stream_mode_list[0][1])
             speed = gr.Number(value=1, label="速度调节(仅支持非流式推理)", minimum=0.5, maximum=2.0, step=0.1)
             with gr.Column(scale=0.25):
@@ -193,6 +193,8 @@ if __name__ == '__main__':
             raise TypeError('no valid model_type!')
 
     sft_spk = cosyvoice.list_available_spks()
+    if len(sft_spk) == 0:
+        sft_spk = ['']
     prompt_sr = 16000
     default_data = np.zeros(cosyvoice.sample_rate)
     main()