Update app.py

app.py

@@ -34,14 +34,17 @@ f0method_mode = []
 f0method_info = ""
 if limitation is True:
     audio_mode = ["Upload audio", "TTS Audio"]
-    f0method_mode = ["pm", "harvest"]
-    f0method_info = "PM is fast, Harvest is good but extremely slow. (Default: PM)"
+    f0method_mode = ["pm", "crepe", "harvest"]
+    f0method_info = "PM is fast, rmvpe is middle, Crepe or harvest is good but it was extremely slow (Default: PM)"
 else:
-    audio_mode = ["Input path", "Upload audio", "Youtube", "TTS Audio"]
-    f0method_mode = ["pm", "harvest", "crepe"]
-    f0method_info = "PM is fast, Harvest is good but extremely slow, and Crepe effect is good but requires GPU (Default: PM)"
+    audio_mode = ["Upload audio", "Youtube", "TTS Audio"]
+    f0method_mode = ["pm", "crepe", "harvest"]
+    f0method_info = "PM is fast, rmvpe is middle. Crepe or harvest is good but it was extremely slow (Default: PM))"
 
-def create_vc_fn(model_name, tgt_sr, net_g, vc, if_f0, version, file_index):
+if os.path.isfile("rmvpe.pt"):
+    f0method_mode.insert(2, "rmvpe")
+
+def create_vc_fn(model_title, tgt_sr, net_g, vc, if_f0, version, file_index):
     def vc_fn(
         vc_audio_mode,
         vc_input, 
@@ -57,7 +60,6 @@ def create_vc_fn(model_name, tgt_sr, net_g, vc, if_f0, version, file_index):
         protect,
     ):
         try:
-            print(f"Converting using {model_name}...")
             if vc_audio_mode == "Input path" or "Youtube" and vc_input != "":
                 audio, sr = librosa.load(vc_input, sr=16000, mono=True)
             elif vc_audio_mode == "Upload audio":
@@ -65,15 +67,15 @@ def create_vc_fn(model_name, tgt_sr, net_g, vc, if_f0, version, file_index):
                     return "You need to upload an audio", None
                 sampling_rate, audio = vc_upload
                 duration = audio.shape[0] / sampling_rate
-                if duration > 20 and limitation:
-                    return "Please upload an audio file that is less than 20 seconds. If you need to generate a longer audio file, please use Colab.", None
+                if duration > 360 and limitation:
+                    return "Please upload an audio file that is less than 1 minute.", None
                 audio = (audio / np.iinfo(audio.dtype).max).astype(np.float32)
                 if len(audio.shape) > 1:
                     audio = librosa.to_mono(audio.transpose(1, 0))
                 if sampling_rate != 16000:
                     audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=16000)
             elif vc_audio_mode == "TTS Audio":
-                if len(tts_text) > 100 and limitation:
+                if len(tts_text) > 600 and limitation:
                     return "Text is too long", None
                 if tts_text is None or tts_voice is None:
                     return "You need to enter text and select a voice", None
@@ -104,12 +106,12 @@ def create_vc_fn(model_name, tgt_sr, net_g, vc, if_f0, version, file_index):
                 f0_file=None,
             )
             info = f"[{datetime.now().strftime('%Y-%m-%d %H:%M')}]: npy: {times[0]}, f0: {times[1]}s, infer: {times[2]}s"
-            print(f"{model_name} | {info}")
+            print(f"{model_title} | {info}")
             return info, (tgt_sr, audio_opt)
         except:
             info = traceback.format_exc()
             print(info)
-            return info, None
+            return info, (None, None)
     return vc_fn
 
 def load_model():
@@ -121,7 +123,6 @@ def load_model():
             continue
         category_title = category_info['title']
         category_folder = category_info['folder_path']
-        description = category_info['description']
         models = []
         with open(f"weights/{category_folder}/model_info.json", "r", encoding="utf-8") as f:
             models_info = json.load(f)
@@ -159,8 +160,8 @@ def load_model():
                 net_g = net_g.float()
             vc = VC(tgt_sr, config)
             print(f"Model loaded: {character_name} / {info['feature_retrieval_library']} | ({model_version})")
-            models.append((character_name, model_title, model_author, model_cover, model_version, create_vc_fn(model_name, tgt_sr, net_g, vc, if_f0, version, model_index)))
-        categories.append([category_title, category_folder, description, models])
+            models.append((character_name, model_title, model_author, model_cover, model_version, create_vc_fn(model_title, tgt_sr, net_g, vc, if_f0, version, model_index)))
+        categories.append([category_title, category_folder, models])
     return categories
 
 def cut_vocal_and_inst(url, audio_provider, split_model):
@@ -169,17 +170,25 @@ def cut_vocal_and_inst(url, audio_provider, split_model):
             os.mkdir("dl_audio")
         if audio_provider == "Youtube":
             ydl_opts = {
-                'noplaylist': True,
-                'format': 'bestaudio/best',
-                'postprocessors': [{
-                    'key': 'FFmpegExtractAudio',
-                    'preferredcodec': 'wav',
-                }],
-                "outtmpl": 'dl_audio/youtube_audio',
+            'format': 'bestaudio/best',
+            'postprocessors': [{
+                'key': 'FFmpegExtractAudio',
+                'preferredcodec': 'wav',
+            }],
+            "outtmpl": 'dl_audio/youtube_audio',
             }
             with yt_dlp.YoutubeDL(ydl_opts) as ydl:
                 ydl.download([url])
             audio_path = "dl_audio/youtube_audio.wav"
+        else:
+            # Spotify doesnt work.
+            # Need to find other solution soon.
+            ''' 
+            command = f"spotdl download {url} --output dl_audio/.wav"
+            result = subprocess.run(command.split(), stdout=subprocess.PIPE)
+            print(result.stdout.decode())
+            audio_path = "dl_audio/spotify_audio.wav"
+            '''
         if split_model == "htdemucs":
             command = f"demucs --two-stems=vocals {audio_path} -o output"
             result = subprocess.run(command.split(), stdout=subprocess.PIPE)
@@ -232,7 +241,6 @@ def change_audio_mode(vc_audio_mode):
         return (
             # Input & Upload
             gr.Textbox.update(visible=True),
-            gr.Checkbox.update(visible=False),
             gr.Audio.update(visible=False),
             # Youtube
             gr.Dropdown.update(visible=False),
@@ -253,7 +261,6 @@ def change_audio_mode(vc_audio_mode):
         return (
             # Input & Upload
             gr.Textbox.update(visible=False),
-            gr.Checkbox.update(visible=True),
             gr.Audio.update(visible=True),
             # Youtube
             gr.Dropdown.update(visible=False),
@@ -274,7 +281,6 @@ def change_audio_mode(vc_audio_mode):
         return (
             # Input & Upload
             gr.Textbox.update(visible=False),
-            gr.Checkbox.update(visible=False),
             gr.Audio.update(visible=False),
             # Youtube
             gr.Dropdown.update(visible=True),
@@ -295,7 +301,6 @@ def change_audio_mode(vc_audio_mode):
         return (
             # Input & Upload
             gr.Textbox.update(visible=False),
-            gr.Checkbox.update(visible=False),
             gr.Audio.update(visible=False),
             # Youtube
             gr.Dropdown.update(visible=False),
@@ -316,7 +321,6 @@ def change_audio_mode(vc_audio_mode):
         return (
             # Input & Upload
             gr.Textbox.update(visible=False),
-            gr.Checkbox.update(visible=True),
             gr.Audio.update(visible=True),
             # Youtube
             gr.Dropdown.update(visible=False),
@@ -334,28 +338,20 @@ def change_audio_mode(vc_audio_mode):
             gr.Dropdown.update(visible=False)
         )
 
-def use_microphone(microphone):
-    if microphone == True:
-        return gr.Audio.update(source="microphone")
-    else:
-        return gr.Audio.update(source="upload")
-
 if __name__ == '__main__':
     load_hubert()
     categories = load_model()
     tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
     voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]
-    with gr.Blocks() as app:
+    with gr.Blocks(theme=gr.themes.Base()) as app:
         gr.Markdown(
-            "<div align='center'>\n\n"+
-            "# Multi Model RVC Inference\n\n"+
-            "[![Repository](https://img.shields.io/badge/Github-Multi%20Model%20RVC%20Inference-blue?style=for-the-badge&logo=github)](https://github.com/ArkanDash/Multi-Model-RVC-Inference)\n\n"+
-            "</div>"
+            "# <center> RVC Models\n"
+            "### <center> will update every hololive ai model that i can find or make.\n"
+            "[![image](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/aziib/hololive-rvc-models-v2/blob/main/hololive_rvc_models_v2.ipynb)\n\n"
+            "[![ko-fi](https://ko-fi.com/img/githubbutton_sm.svg)](https://ko-fi.com/megaaziib)\n\n"
         )
-        for (folder_title, folder, description, models) in categories:
+        for (folder_title, folder, models) in categories:
             with gr.TabItem(folder_title):
-                if description:
-                    gr.Markdown(f"### <center> {description}")
                 with gr.Tabs():
                     if not models:
                         gr.Markdown("# <center> No Model Loaded.")
@@ -375,11 +371,9 @@ if __name__ == '__main__':
                             with gr.Row():
                                 with gr.Column():
                                     vc_audio_mode = gr.Dropdown(label="Input voice", choices=audio_mode, allow_custom_value=False, value="Upload audio")
-                                    # Input
+                                    # Input and Upload
                                     vc_input = gr.Textbox(label="Input audio path", visible=False)
-                                    # Upload
-                                    vc_microphone_mode = gr.Checkbox(label="Use Microphone", value=False, visible=True, interactive=True)
-                                    vc_upload = gr.Audio(label="Upload audio file", source="upload", visible=True, interactive=True)
+                                    vc_upload = gr.Audio(label="Upload audio file", visible=True, interactive=True)
                                     # Youtube
                                     vc_download_audio = gr.Dropdown(label="Provider", choices=["Youtube"], allow_custom_value=False, visible=False, value="Youtube", info="Select provider (Default: Youtube)")
                                     vc_link = gr.Textbox(label="Youtube URL", visible=False, info="Example: https://www.youtube.com/watch?v=Nc0sB1Bmf-A", placeholder="https://www.youtube.com/watch?v=...")
@@ -404,8 +398,8 @@ if __name__ == '__main__':
                                         minimum=0,
                                         maximum=1,
                                         label="Retrieval feature ratio",
-                                        info="(Default: 0.7)",
-                                        value=0.7,
+                                        info="Accents controling. Too high prob gonna sounds too robotic (Default: 0.4)",
+                                        value=0.4,
                                         interactive=True,
                                     )
                                     filter_radius0 = gr.Slider(
@@ -413,7 +407,7 @@ if __name__ == '__main__':
                                         maximum=7,
                                         label="Apply Median Filtering",
                                         info="The value represents the filter radius and can reduce breathiness.",
-                                        value=3,
+                                        value=1,
                                         step=1,
                                         interactive=True,
                                     )
@@ -439,7 +433,7 @@ if __name__ == '__main__':
                                         maximum=0.5,
                                         label="Voice Protection",
                                         info="Protect voiceless consonants and breath sounds to prevent artifacts such as tearing in electronic music. Set to 0.5 to disable. Decrease the value to increase protection, but it may reduce indexing accuracy",
-                                        value=0.5,
+                                        value=0.23,
                                         step=0.01,
                                         interactive=True,
                                     )
@@ -463,7 +457,7 @@ if __name__ == '__main__':
                             fn=vc_fn, 
                             inputs=[
                                 vc_audio_mode,
-                                vc_input,
+                                vc_input, 
                                 vc_upload,
                                 tts_text,
                                 tts_voice,
@@ -487,17 +481,11 @@ if __name__ == '__main__':
                             inputs=[vc_output, vc_volume, vc_split_model],
                             outputs=[vc_combined_output]
                         )
-                        vc_microphone_mode.change(
-                            fn=use_microphone,
-                            inputs=vc_microphone_mode,
-                            outputs=vc_upload
-                        )
                         vc_audio_mode.change(
                             fn=change_audio_mode,
                             inputs=[vc_audio_mode],
                             outputs=[
-                                vc_input,
-                                vc_microphone_mode,
+                                vc_input, 
                                 vc_upload,
                                 vc_download_audio,
                                 vc_link,
@@ -513,4 +501,7 @@ if __name__ == '__main__':
                                 tts_voice
                             ]
                         )
-        app.queue(concurrency_count=1, max_size=20, api_open=config.api).launch(share=config.colab)
+if limitation is True:
+    app.queue(concurrency_count=1, max_size=20, api_open=config.api).launch(share=config.colab)
+else:
+    app.queue(concurrency_count=1, max_size=20, api_open=config.api).launch(share=True)