Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -733,36 +733,21 @@ if __name__ == "__main__":
|
|
733 |
).launch()
|
734 |
|
735 |
|
736 |
-
|
737 |
-
|
738 |
import gradio as gr
|
739 |
|
740 |
-
|
741 |
-
gallery_data = [
|
742 |
-
{
|
743 |
-
"image": "default/sikokumetan.webp",
|
744 |
-
"mp3": "default/sikokumetan.mp3"
|
745 |
-
}
|
746 |
-
]
|
747 |
|
748 |
-
|
749 |
-
|
750 |
-
|
751 |
-
|
752 |
-
|
753 |
-
return None
|
754 |
-
|
755 |
-
# 出力の再定義 (重要)
|
756 |
-
outputs = [
|
757 |
-
gr.Audio(label="ストリーム出力音声", streaming=True, format='mp3'),
|
758 |
-
gr.Audio(label="完全出力音声", streaming=False, format='wav')
|
759 |
-
]
|
760 |
|
761 |
if __name__ == "__main__":
|
762 |
description = ("Zero-shot音声変換モデル(学習不要)。ローカルでの利用方法は[GitHubリポジトリ](https://github.com/Plachtaa/seed-vc)をご覧ください。"
|
763 |
"参考音声が25秒を超える場合、自動的に25秒にクリップされます。"
|
764 |
"また、元音声と参考音声の合計時間が30秒を超える場合、元音声は分割処理されます。")
|
765 |
-
|
766 |
inputs = [
|
767 |
gr.Audio(type="filepath", label="元音声"),
|
768 |
gr.Audio(type="filepath", label="参考音声"),
|
@@ -774,25 +759,24 @@ if __name__ == "__main__":
|
|
774 |
gr.Slider(label='音程変換', minimum=-24, maximum=24, step=1, value=0, info="半音単位の音程変換。F0条件付きモデル使用時にのみ有効です"),
|
775 |
]
|
776 |
|
777 |
-
|
778 |
-
|
779 |
-
|
780 |
-
|
781 |
-
|
782 |
-
|
783 |
-
|
784 |
|
785 |
-
|
786 |
-
|
787 |
-
return on_image_select(selected_image)
|
788 |
|
789 |
-
|
790 |
-
|
791 |
-
|
792 |
-
|
793 |
-
|
794 |
-
|
795 |
-
|
796 |
-
|
797 |
-
|
798 |
-
|
|
|
733 |
).launch()
|
734 |
|
735 |
|
|
|
|
|
736 |
import gradio as gr
|
737 |
|
738 |
+
gallery_data = {"sikokumetan": {"webp": "default/sikokumetan.webp", "mp3": "default/sikokumetan.mp3"}}
|
|
|
|
|
|
|
|
|
|
|
|
|
739 |
|
740 |
+
def update_reference(selected_image):
|
741 |
+
for key, value in gallery_data.items():
|
742 |
+
if value["webp"] == selected_image:
|
743 |
+
return value["mp3"]
|
744 |
+
return ""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
745 |
|
746 |
if __name__ == "__main__":
|
747 |
description = ("Zero-shot音声変換モデル(学習不要)。ローカルでの利用方法は[GitHubリポジトリ](https://github.com/Plachtaa/seed-vc)をご覧ください。"
|
748 |
"参考音声が25秒を超える場合、自動的に25秒にクリップされます。"
|
749 |
"また、元音声と参考音声の合計時間が30秒を超える場合、元音声は分割処理されます。")
|
750 |
+
|
751 |
inputs = [
|
752 |
gr.Audio(type="filepath", label="元音声"),
|
753 |
gr.Audio(type="filepath", label="参考音声"),
|
|
|
759 |
gr.Slider(label='音程変換', minimum=-24, maximum=24, step=1, value=0, info="半音単位の音程変換。F0条件付きモデル使用時にのみ有効です"),
|
760 |
]
|
761 |
|
762 |
+
examples = [["examples/source/yae_0.wav", "examples/reference/dingzhen_0.wav", 25, 1.0, 0.7, False, True, 0],
|
763 |
+
["examples/source/jay_0.wav", "examples/reference/azuma_0.wav", 25, 1.0, 0.7, True, True, 0],
|
764 |
+
["examples/source/Wiz Khalifa,Charlie Puth - See You Again [vocals]_[cut_28sec].wav",
|
765 |
+
"examples/reference/teio_0.wav", 100, 1.0, 0.7, True, False, 0],
|
766 |
+
["examples/source/TECHNOPOLIS - 2085 [vocals]_[cut_14sec].wav",
|
767 |
+
"examples/reference/trump_0.wav", 50, 1.0, 0.7, True, False, -12],
|
768 |
+
]
|
769 |
|
770 |
+
outputs = [gr.Audio(label="ストリーム出力音声", streaming=True, format='mp3'),
|
771 |
+
gr.Audio(label="完全出力音声", streaming=False, format='wav')]
|
|
|
772 |
|
773 |
+
with gr.Blocks() as interface:
|
774 |
+
gr.Interface(fn=voice_conversion, description=description, inputs=inputs, outputs=outputs, title="Seed Voice Conversion", examples=examples, cache_examples=False)
|
775 |
+
|
776 |
+
with gr.Row():
|
777 |
+
gallery = gr.Gallery(label="ギャラリー", show_label=True)
|
778 |
+
gallery.add([gallery_data["sikokumetan"]["webp"]])
|
779 |
+
|
780 |
+
gallery.select(update_reference, outputs=inputs[1])
|
781 |
+
|
782 |
+
interface.launch()
|