Update xtts.py

xtts.py

@@ -8,13 +8,15 @@ import scipy
 from TTS.api import TTS
 
 import torch
-import torchaudio
 
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
+
+from qili import upload, upload_bytes
 # def upload_bytes(bytes, ext=".wav"):
 #     return bytes
+# def upload(file):
+#     return file
 
-from qili import upload, upload_bytes
 # if __name__ == "__main__":
 #     app = Flask(__name__)
 # else:
@@ -38,12 +40,16 @@ def predict(text, sample=None, language="zh"):
     global model
     try:
         text= re.sub("([^\x00-\x7F]|\w)(\.|\。|\?)",r"\1 \2\2",text)
-        wav = tts.tts(
+        output=tempfile.mktemp(suffix=".wav")
+        wav = tts.tts_to_file(
             text, 
             language=language if language is not None else "zh",
             speaker_wav=sample if sample is not None else default_sample[0],
+            file_path=output
         )
 
+        return upload(output)
+
         with io.BytesIO() as wav_buffer:
             if torch.is_tensor(wav):
                 wav = wav.cpu().numpy()
@@ -152,29 +158,6 @@ def tts_play():
     url=convert()
     return playInHTML(url)
 
-@app.route("/clone")
-def clone():
-    global tts
-    get_tts()
-    source=request.args.get('source')
-    sample=request.args.get('sample')
-    if not source:
-        return "No source", 400
-
-    source=download(source)
-    output=tempfile.mktemp(suffix=".wav")[1]
-    tts.voice_conversion_to_file(
-        source_wav=source, 
-        target_wav=sample, 
-        file_path=output
-    )
-    return upload(output)
-
-@app.route("/clone/play")
-def clone_play():
-    url=clone()
-    return playInHTML(url)
-
 
 @app.route("/setup")
 def get_tts():
@@ -188,10 +171,10 @@ def get_tts():
         model_name="tts_models/multilingual/multi-dataset/xtts_v2"
         logging.info(f"loading model {model_name} ...")
         tts = TTS(
-            # model_name,
-            model_path=model_path, 
-            config_path=config_path, 
-            vocoder_config_path=vocoder_config_path,  
+            model_name,
+            # model_path=model_path, 
+            # config_path=config_path, 
+            # vocoder_config_path=vocoder_config_path,  
             progress_bar=True
         )
         model=tts.synthesizer.tts_model
@@ -208,4 +191,4 @@ def get_tts():
 @app.route("/")
 def hello():
     return "welcome!"
-logging.info("xtts is ready")
+logging.info("xtts is ready")