upload files

VC_inference.py

@@ -3,13 +3,47 @@ import numpy as np
 import torch
 from torch import no_grad, LongTensor
 import argparse
+import commons
 from mel_processing import spectrogram_torch
 import utils
 from models_infer import SynthesizerTrn
 import gradio as gr
 import librosa
 import webbrowser
+
+from text import text_to_sequence, _clean_text
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
+language_marks = {
+    "Japanese": "",
+    "日本語": "[JA]",
+    "简体中文": "[ZH]",
+    "English": "[EN]",
+    "Mix": "",
+}
+lang = ['日本語', '简体中文', 'English', 'Mix']
+def get_text(text, hps, is_symbol):
+    text_norm = text_to_sequence(text, hps.symbols, [] if is_symbol else hps.data.text_cleaners)
+    if hps.data.add_blank:
+        text_norm = commons.intersperse(text_norm, 0)
+    text_norm = LongTensor(text_norm)
+    return text_norm
+
+def create_tts_fn(model, hps, speaker_ids):
+    def tts_fn(text, speaker, language, speed):
+        if language is not None:
+            text = language_marks[language] + text + language_marks[language]
+        speaker_id = speaker_ids[speaker]
+        stn_tst = get_text(text, hps, False)
+        with no_grad():
+            x_tst = stn_tst.unsqueeze(0).to(device)
+            x_tst_lengths = LongTensor([stn_tst.size(0)]).to(device)
+            sid = LongTensor([speaker_id]).to(device)
+            audio = model.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8,
+                                length_scale=1.0 / speed)[0][0, 0].data.cpu().float().numpy()
+        del stn_tst, x_tst, x_tst_lengths, sid
+        return "Success", (hps.data.sampling_rate, audio)
+
+    return tts_fn
 
 def create_vc_fn(model, hps, speaker_ids):
     def vc_fn(original_speaker, target_speaker, record_audio, upload_audio):
@@ -63,23 +97,43 @@ if __name__ == "__main__":
     _ = utils.load_checkpoint(args.model_dir, net_g, None)
     speaker_ids = hps.speakers
     speakers = list(hps.speakers.keys())
+    tts_fn = create_tts_fn(net_g, hps, speaker_ids)
     vc_fn = create_vc_fn(net_g, hps, speaker_ids)
     app = gr.Blocks()
     with app:
-        gr.Markdown("""
+        with gr.Tab("Text-to-Speech"):
-                        录制或上传声音，并选择要转换的音色。User代表的音色是你自己。
+            with gr.Row():
-        """)
+                with gr.Column():
-        with gr.Column():
+                    textbox = gr.TextArea(label="Text",
-            record_audio = gr.Audio(label="record your voice", source="microphone")
+                                          placeholder="Type your sentence here",
-            upload_audio = gr.Audio(label="or upload audio here", source="upload")
+                                          value="こんにちわ。", elem_id=f"tts-input")
-            source_speaker = gr.Dropdown(choices=speakers, value="User", label="source speaker")
+                    # select character
-            target_speaker = gr.Dropdown(choices=speakers, value=speakers[0], label="target speaker")
+                    char_dropdown = gr.Dropdown(choices=speakers, value=speakers[0], label='character')
-        with gr.Column():
+                    language_dropdown = gr.Dropdown(choices=lang, value=lang[0], label='language')
-            message_box = gr.Textbox(label="Message")
+                    duration_slider = gr.Slider(minimum=0.1, maximum=5, value=1, step=0.1,
-            converted_audio = gr.Audio(label='converted audio')
+                                                label='速度 Speed')
-        btn = gr.Button("Convert!")
+                with gr.Column():
-        btn.click(vc_fn, inputs=[source_speaker, target_speaker, record_audio, upload_audio],
+                    text_output = gr.Textbox(label="Message")
-                  outputs=[message_box, converted_audio])
+                    audio_output = gr.Audio(label="Output Audio", elem_id="tts-audio")
+                    btn = gr.Button("Generate!")
+                    btn.click(tts_fn,
+                              inputs=[textbox, char_dropdown, language_dropdown, duration_slider,],
+                              outputs=[text_output, audio_output])
+        with gr.Tab("Voice Conversion"):
+            gr.Markdown("""
+                            录制或上传声音，并选择要转换的音色。User代表的音色是你自己。
+            """)
+            with gr.Column():
+                record_audio = gr.Audio(label="record your voice", source="microphone")
+                upload_audio = gr.Audio(label="or upload audio here", source="upload")
+                source_speaker = gr.Dropdown(choices=speakers, value="User", label="source speaker")
+                target_speaker = gr.Dropdown(choices=speakers, value=speakers[0], label="target speaker")
+            with gr.Column():
+                message_box = gr.Textbox(label="Message")
+                converted_audio = gr.Audio(label='converted audio')
+            btn = gr.Button("Convert!")
+            btn.click(vc_fn, inputs=[source_speaker, target_speaker, record_audio, upload_audio],
+                      outputs=[message_box, converted_audio])
     webbrowser.open("http://127.0.0.1:7860")
     app.launch(share=args.share)