Update xtts.py
Browse files
xtts.py
CHANGED
@@ -30,24 +30,10 @@ if not os.path.exists(sample_root):
|
|
30 |
os.makedirs(sample_root)
|
31 |
|
32 |
default_sample=f'{os.path.dirname(os.path.abspath(__file__))}/sample.wav', f'{sample_root}/sample.pt'
|
33 |
-
|
34 |
ffmpeg=f'{os.path.dirname(os.path.abspath(__file__))}/ffmpeg'
|
35 |
-
try:
|
36 |
-
st = os.stat(ffmpeg)
|
37 |
-
os.chmod(ffmpeg, st.st_mode | stat.S_IEXEC)
|
38 |
-
except:
|
39 |
-
traceback.print_exc()
|
40 |
|
41 |
tts=None
|
42 |
model=None
|
43 |
-
|
44 |
-
model_name="tts_models/multilingual/multi-dataset/xtts_v2"
|
45 |
-
try:
|
46 |
-
tts = TTS(model_name=model_name)
|
47 |
-
model=tts.synthesizer.tts_model
|
48 |
-
except:
|
49 |
-
traceback.print_exc()
|
50 |
-
|
51 |
@app.route("/convert")
|
52 |
def predict():
|
53 |
global tts
|
@@ -66,12 +52,13 @@ def predict():
|
|
66 |
if tts is None:
|
67 |
TTS=import_module("TTS.api").TTS
|
68 |
model_name="tts_models/multilingual/multi-dataset/xtts_v2"
|
69 |
-
|
|
|
70 |
model=tts.synthesizer.tts_model
|
71 |
-
if not hasattr(model, '__get_conditioning_latents'):
|
72 |
#hack to use cache
|
73 |
model.__get_conditioning_latents=model.get_conditioning_latents
|
74 |
model.get_conditioning_latents=get_conditioning_latents
|
|
|
75 |
|
76 |
wav = tts.tts(
|
77 |
text,
|
|
|
30 |
os.makedirs(sample_root)
|
31 |
|
32 |
default_sample=f'{os.path.dirname(os.path.abspath(__file__))}/sample.wav', f'{sample_root}/sample.pt'
|
|
|
33 |
ffmpeg=f'{os.path.dirname(os.path.abspath(__file__))}/ffmpeg'
|
|
|
|
|
|
|
|
|
|
|
34 |
|
35 |
tts=None
|
36 |
model=None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
37 |
@app.route("/convert")
|
38 |
def predict():
|
39 |
global tts
|
|
|
52 |
if tts is None:
|
53 |
TTS=import_module("TTS.api").TTS
|
54 |
model_name="tts_models/multilingual/multi-dataset/xtts_v2"
|
55 |
+
print(f"loading model {model_name} ...")
|
56 |
+
tts = TTS(model_name=model_name, progress_bar=False)
|
57 |
model=tts.synthesizer.tts_model
|
|
|
58 |
#hack to use cache
|
59 |
model.__get_conditioning_latents=model.get_conditioning_latents
|
60 |
model.get_conditioning_latents=get_conditioning_latents
|
61 |
+
print("model is ready")
|
62 |
|
63 |
wav = tts.tts(
|
64 |
text,
|