add TTS

app.py

@@ -1,12 +1,15 @@
 import os
+import io
 import gradio as gr
 import librosa
 import numpy as np
 import utils
 from inference.infer_tool import Svc
 import logging
-import webbrowser
+import soundfile
+import asyncio
 import argparse
+import edge_tts
 import gradio.processing_utils as gr_processing_utils
 logging.getLogger('numba').setLevel(logging.WARNING)
 logging.getLogger('markdown_it').setLevel(logging.WARNING)
@@ -26,32 +29,65 @@ def audio_postprocess(self, y):
 
 gr.Audio.postprocess = audio_postprocess
 def create_vc_fn(model, sid):
-    def vc_fn(input_audio, vc_transform, auto_f0):
+    def vc_fn(input_audio, vc_transform, auto_f0, tts_text, tts_voice, tts_mode):
+        if tts_mode:
+            if len(tts_text) > 100 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
+            asyncio.run(edge_tts.Communicate(tts_text, "-".join(tts_voice.split('-')[:-1])).save("tts.mp3"))
+            audio, sr = librosa.load("tts.mp3", sr=16000, mono=True)
+            raw_path = io.BytesIO()
+            soundfile.write(raw_path, audio, 16000, format="wav")
+            raw_path.seek(0)
+            out_audio, out_sr = model.infer(sid, vc_transform, raw_path,
+                                            auto_predict_f0=auto_f0,
+                                            )
+            return "Success", (44100, out_audio.cpu().numpy())
         if input_audio is None:
             return "You need to upload an audio", None
         sampling_rate, audio = input_audio
         duration = audio.shape[0] / sampling_rate
-        if duration > 30 and limitation:
-            return "Please upload an audio file that is less than 30 seconds. If you need to generate a longer audio file, please use Colab.", None
+        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
         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 != 44100:
-            audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=44100)
-        out_audio, out_sr = model.infer(sid, vc_transform, audio, auto_predict_f0=auto_f0)
-        model.clear_empty()
+        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)
+        out_audio, out_sr = model.infer(sid, vc_transform, raw_path,
+                                       auto_predict_f0=auto_f0,
+                                       )
         return "Success", (44100, out_audio.cpu().numpy())
     return vc_fn
 
+def change_to_tts_mode(tts_mode):
+    if tts_mode:
+        return gr.Audio.update(visible=False), gr.Textbox.update(visible=True), gr.Dropdown.update(visible=True), gr.Checkbox.update(value=True)
+    else:
+        return gr.Audio.update(visible=True), gr.Textbox.update(visible=False), gr.Dropdown.update(visible=False), gr.Checkbox.update(value=False)
+
 if __name__ == '__main__':
     parser = argparse.ArgumentParser()
     parser.add_argument('--device', type=str, default='cpu')
     parser.add_argument('--api', action="store_true", default=False)
     parser.add_argument("--share", action="store_true", default=False, help="share gradio app")
-    parser.add_argument("--colab", action="store_true", default=False, help="share gradio app")
     args = parser.parse_args()
     hubert_model = utils.get_hubert_model().to(args.device)
     models = []
+    others = {
+        "rudolf": "https://huggingface.co/spaces/sayashi/sovits-rudolf",
+        "alice": "https://huggingface.co/spaces/sayashi/sovits-models",
+        "goldship": "https://huggingface.co/spaces/sayashi/sovits-goldship",
+        "tannhauser": "https://huggingface.co/spaces/sayashi/sovits-tannhauser"
+    }
+    voices = []
+    tts_voice_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices())
+    for r in tts_voice_list:
+        voices.append(f"{r['ShortName']}-{r['Gender']}")
     for f in os.listdir("models"):
         name = f
         model = Svc(fr"models/{f}/{f}.pth", f"models/{f}/config.json", device=args.device, hubert_model=hubert_model)
@@ -62,9 +98,10 @@ if __name__ == '__main__':
             "# <center> Sovits Models\n"
             "## <center> The input audio should be clean and pure voice without background music.\n"
             "![visitor badge](https://visitor-badge.glitch.me/badge?page_id=sayashi.Sovits-Umamusume)\n\n"
-            "[Open In Colab](https://colab.research.google.com/drive/1wfsBbMzmtLflOJeqc5ZnJiLY7L239hJW?usp=share_link)"
-            " without queue and length limitation.\n\n"
-            "[Original Repo](https://github.com/innnky/so-vits-svc/tree/4.0)"
+            "[![image](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1wfsBbMzmtLflOJeqc5ZnJiLY7L239hJW?usp=share_link)\n\n"
+            "[![Duplicate this Space](https://huggingface.co/datasets/huggingface/badges/raw/main/duplicate-this-space-sm-dark.svg)](https://huggingface.co/spaces/sayashi/sovits-teio?duplicate=true)\n\n"
+            "[![Original Repo](https://badgen.net/badge/icon/github?icon=github&label=Original%20Repo)](https://github.com/svc-develop-team/so-vits-svc)"
+
         )
         with gr.Tabs():
             for (name, cover, vc_fn) in models:
@@ -77,14 +114,28 @@ if __name__ == '__main__':
                         )
                     with gr.Row():
                         with gr.Column():
-                            vc_input = gr.Audio(label="Input audio"+' (less than 45 seconds)' if limitation else '')
+                            vc_input = gr.Audio(label="Input audio"+' (less than 20 seconds)' if limitation else '')
                             vc_transform = gr.Number(label="vc_transform", value=0)
                             auto_f0 = gr.Checkbox(label="auto_f0", value=False)
+                            tts_mode = gr.Checkbox(label="tts (use edge-tts as input)", value=False)
+                            tts_text = gr.Textbox(visible=False, label="TTS text (100 words limitation)" if limitation else "TTS text")
+                            tts_voice = gr.Dropdown(choices=voices, visible=False)
                             vc_submit = gr.Button("Generate", variant="primary")
                         with gr.Column():
                             vc_output1 = gr.Textbox(label="Output Message")
                             vc_output2 = gr.Audio(label="Output Audio")
-                vc_submit.click(vc_fn, [vc_input, vc_transform, auto_f0], [vc_output1, vc_output2])
-        if args.colab:
-            webbrowser.open("http://127.0.0.1:7860")
-        app.queue(concurrency_count=1, api_open=args.api).launch(share=args.share)
+                vc_submit.click(vc_fn, [vc_input, vc_transform, auto_f0, tts_text, tts_voice, tts_mode], [vc_output1, vc_output2])
+                tts_mode.change(change_to_tts_mode, [tts_mode], [vc_input, tts_text, tts_voice, auto_f0])
+            for category, link in others.items():
+                with gr.TabItem(category):
+                    gr.Markdown(
+                        f'''
+                        <center>
+                          <h2>Click to Go</h2>
+                          <a href="{link}">
+                            <img src="https://huggingface.co/datasets/huggingface/badges/raw/main/open-in-hf-spaces-xl-dark.svg"
+                          </a>
+                        </center>
+                        '''
+                    )
+        app.queue(concurrency_count=1, api_open=args.api).launch(share=args.share)

inference/infer_tool.py

@@ -142,7 +142,8 @@ class Svc(object):
 
 
 
-    def get_unit_f0(self, wav, tran, cluster_infer_ratio, speaker):
+    def get_unit_f0(self, in_path, tran, cluster_infer_ratio, speaker):
+        wav, sr = librosa.load(in_path, sr=self.target_sample)
         f0 = utils.compute_f0_parselmouth(wav, sampling_rate=self.target_sample, hop_length=self.hop_size)
         f0, uv = utils.interpolate_f0(f0)
         f0 = torch.FloatTensor(f0)
@@ -151,12 +152,12 @@ class Svc(object):
         f0 = f0.unsqueeze(0).to(self.dev)
         uv = uv.unsqueeze(0).to(self.dev)
 
-        wav16k = librosa.resample(wav, orig_sr=44100, target_sr=16000)
+        wav16k = librosa.resample(wav, orig_sr=self.target_sample, target_sr=16000)
         wav16k = torch.from_numpy(wav16k).to(self.dev)
         c = utils.get_hubert_content(self.hubert_model, wav_16k_tensor=wav16k)
         c = utils.repeat_expand_2d(c.squeeze(0), f0.shape[1])
 
-        if cluster_infer_ratio !=0:
+        if cluster_infer_ratio != 0:
             cluster_c = cluster.get_cluster_center_result(self.cluster_model, c.cpu().numpy().T, speaker).T
             cluster_c = torch.FloatTensor(cluster_c).to(self.dev)
             c = cluster_infer_ratio * cluster_c + (1 - cluster_infer_ratio) * c
@@ -164,7 +165,7 @@ class Svc(object):
         c = c.unsqueeze(0)
         return c, f0, uv
 
-    def infer(self, speaker, tran, raw_wav,
+    def infer(self, speaker, tran, raw_path,
               cluster_infer_ratio=0,
               auto_predict_f0=False,
               noice_scale=0.4):
@@ -173,7 +174,7 @@ class Svc(object):
             if len(self.spk2id.__dict__) >= speaker:
                 speaker_id = speaker
         sid = torch.LongTensor([int(speaker_id)]).to(self.dev).unsqueeze(0)
-        c, f0, uv = self.get_unit_f0(raw_wav, tran, cluster_infer_ratio, speaker)
+        c, f0, uv = self.get_unit_f0(raw_path, tran, cluster_infer_ratio, speaker)
         if "half" in self.net_g_path and torch.cuda.is_available():
             c = c.half()
         with torch.no_grad():
@@ -187,17 +188,18 @@ class Svc(object):
         # 清理显存
         torch.cuda.empty_cache()
 
-    def slice_inference(self,raw_audio_path, spk, tran, slice_db,cluster_infer_ratio, auto_predict_f0,noice_scale, pad_seconds=0.5, clip_seconds=0,lg_num=0,lgr_num =0.75):
+    def slice_inference(self, raw_audio_path, spk, tran, slice_db, cluster_infer_ratio, auto_predict_f0, noice_scale,
+                        pad_seconds=0.5, clip_seconds=0, lg_num=0, lgr_num=0.75):
         wav_path = raw_audio_path
         chunks = slicer.cut(wav_path, db_thresh=slice_db)
         audio_data, audio_sr = slicer.chunks2audio(wav_path, chunks)
-        per_size = int(clip_seconds*audio_sr)
-        lg_size = int(lg_num*audio_sr)
-        lg_size_r = int(lg_size*lgr_num)
-        lg_size_c_l = (lg_size-lg_size_r)//2
-        lg_size_c_r = lg_size-lg_size_r-lg_size_c_l
-        lg = np.linspace(0,1,lg_size_r) if lg_size!=0 else 0
-        
+        per_size = int(clip_seconds * audio_sr)
+        lg_size = int(lg_num * audio_sr)
+        lg_size_r = int(lg_size * lgr_num)
+        lg_size_c_l = (lg_size - lg_size_r) // 2
+        lg_size_c_r = lg_size - lg_size_r - lg_size_c_l
+        lg = np.linspace(0, 1, lg_size_r) if lg_size != 0 else 0
+
         audio = []
         for (slice_tag, data) in audio_data:
             print(f'#=====segment start, {round(len(data) / audio_sr, 3)}s======')
@@ -209,12 +211,12 @@ class Svc(object):
                 audio.extend(list(pad_array(_audio, length)))
                 continue
             if per_size != 0:
-                datas = split_list_by_n(data, per_size,lg_size)
+                datas = split_list_by_n(data, per_size, lg_size)
             else:
                 datas = [data]
-            for k,dat in enumerate(datas):
-                per_length = int(np.ceil(len(dat) / audio_sr * self.target_sample)) if clip_seconds!=0 else length
-                if clip_seconds!=0: print(f'###=====segment clip start, {round(len(dat) / audio_sr, 3)}s======')
+            for k, dat in enumerate(datas):
+                per_length = int(np.ceil(len(dat) / audio_sr * self.target_sample)) if clip_seconds != 0 else length
+                if clip_seconds != 0: print(f'###=====segment clip start, {round(len(dat) / audio_sr, 3)}s======')
                 # padd
                 pad_len = int(audio_sr * pad_seconds)
                 dat = np.concatenate([np.zeros([pad_len]), dat, np.zeros([pad_len])])
@@ -222,24 +224,25 @@ class Svc(object):
                 soundfile.write(raw_path, dat, audio_sr, format="wav")
                 raw_path.seek(0)
                 out_audio, out_sr = self.infer(spk, tran, raw_path,
-                                                    cluster_infer_ratio=cluster_infer_ratio,
-                                                    auto_predict_f0=auto_predict_f0,
-                                                    noice_scale=noice_scale
-                                                    )
+                                               cluster_infer_ratio=cluster_infer_ratio,
+                                               auto_predict_f0=auto_predict_f0,
+                                               noice_scale=noice_scale
+                                               )
                 _audio = out_audio.cpu().numpy()
                 pad_len = int(self.target_sample * pad_seconds)
                 _audio = _audio[pad_len:-pad_len]
                 _audio = pad_array(_audio, per_length)
-                if lg_size!=0 and k!=0:
-                    lg1 = audio[-(lg_size_r+lg_size_c_r):-lg_size_c_r] if lgr_num != 1 else audio[-lg_size:]
-                    lg2 = _audio[lg_size_c_l:lg_size_c_l+lg_size_r]  if lgr_num != 1 else _audio[0:lg_size]
-                    lg_pre = lg1*(1-lg)+lg2*lg
-                    audio = audio[0:-(lg_size_r+lg_size_c_r)] if lgr_num != 1 else audio[0:-lg_size]
+                if lg_size != 0 and k != 0:
+                    lg1 = audio[-(lg_size_r + lg_size_c_r):-lg_size_c_r] if lgr_num != 1 else audio[-lg_size:]
+                    lg2 = _audio[lg_size_c_l:lg_size_c_l + lg_size_r] if lgr_num != 1 else _audio[0:lg_size]
+                    lg_pre = lg1 * (1 - lg) + lg2 * lg
+                    audio = audio[0:-(lg_size_r + lg_size_c_r)] if lgr_num != 1 else audio[0:-lg_size]
                     audio.extend(lg_pre)
-                    _audio = _audio[lg_size_c_l+lg_size_r:] if lgr_num != 1 else _audio[lg_size:]
+                    _audio = _audio[lg_size_c_l + lg_size_r:] if lgr_num != 1 else _audio[lg_size:]
                 audio.extend(list(_audio))
         return np.array(audio)
 
+
 class RealTimeVC:
     def __init__(self):
         self.last_chunk = None

requirements.txt

@@ -19,3 +19,4 @@ onnxsim
 onnxoptimizer
 fairseq
 librosa
+edge-tts