Update app.py

app.py

@@ -5,6 +5,7 @@ import librosa
 import soundfile
 import io
 import argparse
+import numpy as np
 from inference.infer_tool import Svc
 
 def get_or_create_eventloop():
@@ -25,8 +26,8 @@ def tts_get_voices_list():
     return voices
 
 def tts_mode(txt, voice):
-    tts = asyncio.run(edge_tts.Communicate(txt, voice).save('test.mp3'))
-    audio, sr = librosa.load('test.mp3', sr=16000, mono=True)
+    tts = asyncio.run(edge_tts.Communicate(txt, voice).save('audio.mp3'))
+    audio, sr = librosa.load('audio.mp3', sr=16000, mono=True)
     raw_path = io.BytesIO()
     soundfile.write(raw_path, audio, 16000, format="wav")
     raw_path.seek(0)
@@ -35,6 +36,23 @@ def tts_mode(txt, voice):
    
     return (44100, out_audio.cpu().numpy())
 
+def audio_infer_mode(input_audio):
+    sampling_rate, audio = input_audio
+    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, org_sr=sampling_rate, target_sr=16000)
+
+    raw_path = io.BytesIO()
+    soundfile.write(raw_path, audio, 16000, format="wav")
+    raw_path.seek(0)
+    model = Svc(fr"Herta-Svc/G_10000.pth", f"Herta-Svc/config.json", device = 'cpu')
+    out_audio, out_sr = model.infer('speaker0', 0, raw_path, auto_predict_f0 = True,)
+
+    return (44100, out_audio.cpu().numpy())
+
 if __name__ == '__main__':
     parser = argparse.ArgumentParser()
     parser.add_argument('--device', type=str, default='cpu')
@@ -52,9 +70,20 @@ if __name__ == '__main__':
                             '</div>')
                 tts_text = gr.Textbox(label="TTS text (100 words limitation)", visible = True)
                 tts_voice = gr.Dropdown(choices= tts_get_voices_list(), visible = True)
+                audio_mode = gr.Checkbox(label = 'Upload audio instead')
+                audio_input = gr.Audio(label = 'Input Audio')
                 audio_output = gr.Audio(label="Output Audio")
                 btn_submit = gr.Button("Generate")
-    
-        btn_submit.click(tts_mode, [tts_text, tts_voice], [audio_output])
+
+            if audio_mode.update == True:
+                tts_text = gr.Textbox.update(label="TTS text (100 words limitation)", visible = False, show_label = False)
+                tts_voice = gr.Dropdown.update(choices= tts_get_voices_list(), visible = False, show_label = False)
+                audio_input = gr.Audio.update(label = 'Input Audio', visible = True, show_label = True)
+            else:
+                tts_text = gr.Textbox.update(label="TTS text (100 words limitation)", visible = True, show_label = True)
+                tts_voice = gr.Dropdown.update(choices= tts_get_voices_list(), visible = True, show_label = True)
+                audio_input = gr.Audio.update(label = 'Input Audio', visible = False, show_label = False)
+
+        btn_submit.click([tts_mode, audio_infer_mode], [tts_text, tts_voice, audio_input], [audio_output])
 
     app.queue(concurrency_count=1, api_open=args.api).launch(share=args.share)