Update app_multi.py

app_multi.py

@@ -36,6 +36,33 @@ from infer_pack.models import (
     SynthesizerTrnMs768NSFsid_nono
 )
 from vc_infer_pipeline import VC
+
+# SadTalker
+
+import os, sys
+from src.gradio_demo import SadTalker  
+
+
+try:
+    import webui  # in webui
+    in_webui = True
+except:
+    in_webui = False
+
+
+def toggle_audio_file(choice):
+    if choice == False:
+        return gr.update(visible=True), gr.update(visible=False)
+    else:
+        return gr.update(visible=False), gr.update(visible=True)
+    
+def ref_video_fn(path_of_ref_video):
+    if path_of_ref_video is not None:
+        return gr.update(value=True)
+    else:
+        return gr.update(value=False)
+
+sad_talker = SadTalker("checkpoints", "src/config", lazy_load=True)
     
 # Reference: https://huggingface.co/spaces/zomehwh/rvc-models/blob/main/app.py#L21  # noqa
 in_hf_space = getenv('SYSTEM') == 'spaces'
@@ -629,14 +656,91 @@ with app:
     as_audio_submit.click(fn=audio_separated, inputs=[as_audio_input], outputs=[as_audio_vocals, as_audio_no_vocals, as_audio_message], show_progress=True, queue=True)
                     
     with gr.Row():
-        with gr.Column():
-            with gr.Tab('🎶 - 歌声转换'):
+        with gr.Tab('🎶 - 歌声转换'):
+            with gr.Column():
                 input_audio = as_audio_vocals
                 vc_convert_btn = gr.Button('进行歌声转换吧！', variant='primary')
                 full_song = gr.Button("加入歌曲伴奏吧！", variant="primary")
                 new_song = gr.Audio(label="AI歌手+伴奏", type="filepath")
 
-            with gr.Tab('🎙️ - 文本转语音'):
+                pitch_adjust = gr.Slider(
+                    label='Pitch',
+                    minimum=-24,
+                    maximum=24,
+                    step=1,
+                    value=0
+                )
+                f0_method = gr.Radio(
+                    label='f0 methods',
+                    choices=['pm', 'rmvpe'],
+                    value='pm',
+                    interactive=True
+                )
+
+                with gr.Accordion('更多设置', open=False):
+                    feat_ratio = gr.Slider(
+                        label='Feature ratio',
+                        minimum=0,
+                        maximum=1,
+                        step=0.1,
+                        value=0.6,
+                        visible=False
+                    )
+                    filter_radius = gr.Slider(
+                        label='Filter radius',
+                        minimum=0,
+                        maximum=7,
+                        step=1,
+                        value=3,
+                        visible=False
+                    )
+                    rms_mix_rate = gr.Slider(
+                        label='Volume envelope mix rate',
+                        minimum=0,
+                        maximum=1,
+                        step=0.1,
+                        value=1,
+                        visible=False                
+                    )
+                    resample_rate = gr.Dropdown(
+                        [
+                            'Disable resampling',
+                            '16000',
+                            '22050',
+                            '44100',
+                            '48000'
+                        ],
+                        label='Resample rate',
+                        value='Disable resampling'
+                    )
+
+            with gr.Column():
+                # Model select
+                model_index = gr.Dropdown(
+                    [
+                        '%s - %s' % (
+                            m['metadata'].get('source', 'Unknown'),
+                            m['metadata'].get('name')
+                        )
+                        for m in loaded_models
+                    ],
+                    label='请选择您的AI歌手(必选)',
+                    type='index'
+                )
+    
+                # Model info
+                with gr.Box():
+                    model_info = gr.Markdown(
+                        '### AI歌手信息\n'
+                        'Please select a model from dropdown above.',
+                        elem_id='model_info'
+                    )
+    
+                output_audio = gr.Audio(label='AI歌手(无伴奏)', type="filepath")
+                output_msg = gr.Textbox(label='Output message')
+
+        with gr.Tab('🎙️ - 文本转语音'):
+            with gr.Column():
                 tts_input = gr.Textbox(
                     label='请填写您想要转换的文本(中英皆可)',
                     lines=3
@@ -654,132 +758,95 @@ with app:
                 )
 
                 tts_convert_btn = gr.Button('进行AI变声吧', variant='primary')
-                            
-            pitch_adjust = gr.Slider(
-                label='Pitch',
-                minimum=-24,
-                maximum=24,
-                step=1,
-                value=0
-            )
-            f0_method = gr.Radio(
-                label='f0 methods',
-                choices=['pm', 'rmvpe'],
-                value='pm',
-                interactive=True
-            )
-
-            with gr.Accordion('更多设置', open=False):
-                feat_ratio = gr.Slider(
-                    label='Feature ratio',
-                    minimum=0,
-                    maximum=1,
-                    step=0.1,
-                    value=0.6,
-                    visible=False
-                )
-                filter_radius = gr.Slider(
-                    label='Filter radius',
-                    minimum=0,
-                    maximum=7,
+                pitch_adjust = gr.Slider(
+                    label='Pitch',
+                    minimum=-24,
+                    maximum=24,
                     step=1,
-                    value=3,
-                    visible=False
+                    value=0
                 )
-                rms_mix_rate = gr.Slider(
-                    label='Volume envelope mix rate',
-                    minimum=0,
-                    maximum=1,
-                    step=0.1,
-                    value=1,
-                    visible=False
-                    
+                f0_method = gr.Radio(
+                    label='f0 methods',
+                    choices=['pm', 'rmvpe'],
+                    value='pm',
+                    interactive=True
                 )
-                resample_rate = gr.Dropdown(
+
+                with gr.Accordion('更多设置', open=False):
+                    feat_ratio = gr.Slider(
+                        label='Feature ratio',
+                        minimum=0,
+                        maximum=1,
+                        step=0.1,
+                        value=0.6,
+                        visible=False
+                    )
+                    filter_radius = gr.Slider(
+                        label='Filter radius',
+                        minimum=0,
+                        maximum=7,
+                        step=1,
+                        value=3,
+                        visible=False
+                    )
+                    rms_mix_rate = gr.Slider(
+                        label='Volume envelope mix rate',
+                        minimum=0,
+                        maximum=1,
+                        step=0.1,
+                        value=1,
+                        visible=False                
+                    )
+                    resample_rate = gr.Dropdown(
+                        [
+                            'Disable resampling',
+                            '16000',
+                            '22050',
+                            '44100',
+                            '48000'
+                        ],
+                        label='Resample rate',
+                        value='Disable resampling'
+                    )
+
+            with gr.Column():
+                # Model select
+                model_index = gr.Dropdown(
                     [
-                        'Disable resampling',
-                        '16000',
-                        '22050',
-                        '44100',
-                        '48000'
+                        '%s - %s' % (
+                            m['metadata'].get('source', 'Unknown'),
+                            m['metadata'].get('name')
+                        )
+                        for m in loaded_models
                     ],
-                    label='Resample rate',
-                    value='Disable resampling'
+                    label='请选择您的AI歌手(必选)',
+                    type='index'
                 )
-
-        with gr.Column():
-            # Model select
-            model_index = gr.Dropdown(
-                [
-                    '%s - %s' % (
-                        m['metadata'].get('source', 'Unknown'),
-                        m['metadata'].get('name')
+    
+                # Model info
+                with gr.Box():
+                    model_info = gr.Markdown(
+                        '### AI歌手信息\n'
+                        'Please select a model from dropdown above.',
+                        elem_id='model_info'
                     )
-                    for m in loaded_models
-                ],
-                label='请选择您的AI歌手(必选)',
-                type='index'
-            )
-
-            # Model info
-            with gr.Box():
-                model_info = gr.Markdown(
-                    '### AI歌手信息\n'
-                    'Please select a model from dropdown above.',
-                    elem_id='model_info'
-                )
+    
+                output_audio = gr.Audio(label='AI歌手(无伴奏)', type="filepath")
+                output_msg = gr.Textbox(label='Output message')
 
-            output_audio = gr.Audio(label='AI歌手(无伴奏)', type="filepath")
-            output_msg = gr.Textbox(label='Output message')
         with gr.Tab("📺 - 音乐视频"):
             with gr.Row():
                 with gr.Column():
-                    inp1 = gr.Textbox(label="为视频配上精���的文案吧(选填)")
+                    inp1 = gr.Textbox(label="为视频配上精彩的文案吧(选填;英文)")
                     inp2 = new_song
                     inp3 = gr.Image(source='upload', type='filepath', label="上传一张背景图片吧")
                     btn = gr.Button("生成您的专属音乐视频吧", variant="primary")
               
                 with gr.Column():
                     out1 = gr.Video(label='您的专属音乐视频')
-        btn.click(fn=infer, inputs=[inp1, inp2, inp3], outputs=[out1])
-
-    multi_examples = multi_cfg.get('examples')
-    if (
-        multi_examples and
-        multi_examples.get('vc') and multi_examples.get('tts_vc')
-    ):
-        with gr.Accordion('Sweet sweet examples', open=False):
-            with gr.Row():
-                # VC Example
-                if multi_examples.get('vc'):
-                    gr.Examples(
-                        label='Audio conversion examples',
-                        examples=multi_examples.get('vc'),
-                        inputs=[
-                            input_audio, model_index, pitch_adjust, f0_method,
-                            feat_ratio
-                        ],
-                        outputs=[output_audio, output_msg, model_info],
-                        fn=_example_vc,
-                        cache_examples=args.cache_examples,
-                        run_on_click=args.cache_examples
-                    )
-
-                # Edge TTS Example
-                if multi_examples.get('tts_vc'):
-                    gr.Examples(
-                        label='TTS conversion examples',
-                        examples=multi_examples.get('tts_vc'),
-                        inputs=[
-                            tts_input, model_index, tts_speaker, pitch_adjust,
-                            f0_method, feat_ratio
-                        ],
-                        outputs=[output_audio, output_msg, model_info],
-                        fn=_example_edge_tts,
-                        cache_examples=args.cache_examples,
-                        run_on_click=args.cache_examples
-                    )
-
+                    
+    btn.click(fn=infer, inputs=[inp1, inp2, inp3], outputs=[out1])
+                                        
     vc_convert_btn.click(
         vc_func,
         [