Update app.py

app.py

@@ -25,8 +25,31 @@ def tts_get_voices_list():
     
     return voices
 
-def infer(txt, voice):
-    tts = asyncio.run(edge_tts.Communicate(txt, voice).save('audio.mp3'))
+def infer(txt, tts_voice, input_audio, predict_f0, audio_mode):
+    if audio_mode:
+        if input_audio is  None:
+            return 'Please upload your audio file'
+        
+        sampling_rate, audio = input_audio
+        duration = audio.shape[0] / sampling_rate
+
+        if duration > 30:
+            return 'The audio file is too long, Please upload audio file that less than 30 seconds'
+        
+        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)
+
+        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 = predict_f0,)
+        return (44100, out_audio.cpu().numpy())
+
+    tts = asyncio.run(edge_tts.Communicate(txt, tts_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")
@@ -35,6 +58,12 @@ def infer(txt, voice):
     out_audio, out_sr = model.infer('speaker0', 0, raw_path, auto_predict_f0 = True,)
     return (44100, out_audio.cpu().numpy())
 
+def change_to_audio_mode(audio_mode):
+    if audio_mode:    
+        return gr.Audio.update(visible = True), gr.Textbox.update(visible= False), gr.Dropdown.update(visible = False), gr.Checkbox.update(value = True)
+    else:
+        return gr.Audio.update(visible = False), gr.Textbox.update(visible= True), gr.Dropdown.update(visible = True), gr.Checkbox.update(value = False)
+
 if __name__ == '__main__':
     parser = argparse.ArgumentParser()
     parser.add_argument('--device', type=str, default='cpu')
@@ -50,11 +79,15 @@ if __name__ == '__main__':
                 cover = gr.Markdown('<div align="center">'
                             f'<img style="width:auto;height:300px;" src="file/Herta-Svc/herta.png">'
                             '</div>')
-                tts_text = gr.Textbox(label="TTS text (100 words limitation)", visible = True)
-                tts_voice = gr.Dropdown(choices= tts_get_voices_list(), visible = True)
+                tts_text = gr.Textbox(label="TTS text (100 words limitation)")
+                audio_input = gr.Audio(label = 'Please upload audio file that less than 30 seconds', visible = False)
+                tts_voice = gr.Dropdown(choices= tts_get_voices_list())
+                predict_f0 = gr.Checkbox(label = 'Auto predict F0', value = False)
+                audio_mode = gr.Checkbox(label = 'Upload audio instead', value = False)
                 audio_output = gr.Audio(label="Output Audio")
                 btn_submit = gr.Button("Generate")
 
-
-        btn_submit.click(infer, [tts_text, tts_voice], [audio_output])
-    app.queue(concurrency_count=1, api_open=args.api).launch(share=args.share)
+        btn_submit.click(infer, [tts_text, tts_voice, audio_input, predict_f0, audio_mode], [audio_output])
+        audio_mode.change(change_to_audio_mode, audio_mode, [audio_input, tts_text, tts_voice])
+    
+    app.queue(concurrency_count=1, api_open=args.api).launch(share=args.share)