update app.py

app.py

@@ -18,12 +18,12 @@ converter = opencc.OpenCC('s2t.json')
 
 # 加载模型
 cosyvoice_base = CosyVoice2(
-    'ASLP-lab/Cosyvoice2-Yue-ZoengJyutGaai',
+    'ASLP-lab/Cosyvoice2-Yue',
     load_jit=False, load_trt=False, load_vllm=False, fp16=False
 )
 print('load model 1')
 cosyvoice_zjg = CosyVoice2(
-    'ASLP-lab/Cosyvoice2-Yue',
+    'ASLP-lab/Cosyvoice2-Yue-ZoengJyutGaai',
     load_jit=False, load_trt=False, load_vllm=False, fp16=False
 )
 print('load model 2')
@@ -31,42 +31,26 @@ print('load model 2')
 #     'pretrained_models/CosyVoice2-yue-biaobei',
 #     load_jit=False, load_trt=False, load_vllm=False, fp16=False
 # )
-import gradio as gr
-
-# 预设音频字典
-preset_audio_options = {
-    "张悦楷示例": "asset/sg_017_090.wav",
-    "精品女音示例": "asset/F01_中立_20054.wav",
-}
-
-def preview_preset_audio(choice):
-    if choice is None:
-        return None
-    return choice  # 返回文件路径即可被 gr.Audio 播放
 
 @spaces.GPU 
-def tts_inference(model_choice, text, uploaded_audio, preset_audio_choice):
-    # 选择模型
+def tts_inference(model_choice, text, prompt_audio):
+    # 选择模型和默认音频
     if model_choice == "CosyVoice2-张悦楷粤语评书":
         model = cosyvoice_zjg
-        default_prompt = "asset/sg_017_090.wav"
+        prompt_audio = "asset/sg_017_090.wav"
+    elif model_choice == "CosyVoice2-精品女音":
+        model = cosyvoice_base
+        prompt_audio = "asset/F01_中立_20054.wav"
     elif model_choice == "CosyVoice2-base":
         model = cosyvoice_base
-        default_prompt = None
+        if prompt_audio is None:
+            return None, "请上传参考音频"
     else:
         return None, "未知模型"
-
-    # 决定使用哪条音频
-    if preset_audio_choice is not None:
-        prompt_audio = preset_audio_options[preset_audio_choice]
-    elif uploaded_audio is not None:
-        prompt_audio = uploaded_audio
-    elif default_prompt is not None:
-        prompt_audio = default_prompt
-    else:
-        return None, "请上传参考音频或选择预设音频"
-
+    
     model.model.cuda()
+
+    # 繁简转换
     text = converter.convert(text)
     prompt_speech_16k = load_wav(prompt_audio, 16000)
 
@@ -84,29 +68,24 @@ def tts_inference(model_choice, text, uploaded_audio, preset_audio_choice):
 
     return (sample_rate, audio_numpy), f"生成成功：{text}"
 
-# ---- Gradio 界面 ----
-with gr.Blocks() as demo:
-    gr.Markdown("## TTS Demo")
-    
-    model_choice = gr.Dropdown(["CosyVoice2-base", "CosyVoice2-张悦楷粤语评书"], label="选择模型")
-    text_input = gr.Textbox(lines=2, label="输入文本")
-    uploaded_audio = gr.Audio(source="upload", type="filepath", label="上传参考音频（仅 CosyVoice2-base 必需）")
-    
-    preset_audio_choice = gr.Radio(list(preset_audio_options.keys()), label="选择预设音频（可试听）", value=None)
-    preview_audio = gr.Audio(type="filepath", label="试听预设音频")
-    
-    # 预览按钮
-    preset_audio_choice.change(fn=preview_preset_audio, inputs=preset_audio_choice, outputs=preview_audio)
-    
-    # TTS 生成按钮
-    generate_btn = gr.Button("生成 TTS")
-    output_audio = gr.Audio(type="numpy", label="生成的语音")
-    status = gr.Textbox(label="状态信息")
-    
-    generate_btn.click(
-        fn=tts_inference,
-        inputs=[model_choice, text_input, uploaded_audio, preset_audio_choice],
-        outputs=[output_audio, status]
-    )
+
+# ---- Gradio Interface ----
+demo = gr.Interface(
+    fn=tts_inference,
+    inputs=[
+        gr.Dropdown(
+            ["CosyVoice2-base", "CosyVoice2-张悦楷粤语评书"],
+            # ["CosyVoice2-base", "CosyVoice2-张悦楷粤语评书", "CosyVoice2-精品女音"],
+            label="选择模型", value="CosyVoice2-base"
+        ),
+        gr.Textbox(lines=2, label="输入文本"),
+        # gr.Audio(source="upload", type="filepath", label="上传参考音频（仅 CosyVoice2-base 必需）")
+        gr.Audio(sources=["upload"], type="filepath", label="上传参考音频（仅 CosyVoice2-base 必需）")
+    ],
+    outputs=[
+        gr.Audio(type="numpy", label="生成的语音"),
+        gr.Textbox(label="状态信息")
+    ]
+)
 
 demo.launch()