Spaces:
Running
Running
tu
commited on
Commit
•
c17bcc7
1
Parent(s):
9a436eb
update model + change preprocess
Browse files- app.py +22 -4
- model.onnx +2 -2
app.py
CHANGED
@@ -10,11 +10,29 @@ from text import text_to_sequence
|
|
10 |
from scipy.io.wavfile import write
|
11 |
from preprocess import preprocess
|
12 |
import onnxruntime
|
|
|
13 |
|
14 |
def get_text(texts, hps):
|
15 |
text_norm_list = []
|
16 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
text = preprocess(text)
|
|
|
18 |
chunk_strings = []
|
19 |
chunk_len = 30
|
20 |
for i in range(0, len(text.split()), chunk_len):
|
@@ -29,8 +47,8 @@ def get_text(texts, hps):
|
|
29 |
|
30 |
def tts(text):
|
31 |
model_path = "model.onnx"
|
32 |
-
config_path = "
|
33 |
-
sid =
|
34 |
output_wav_path = "output.wav"
|
35 |
sess_options = onnxruntime.SessionOptions()
|
36 |
model = onnxruntime.InferenceSession(str(model_path), sess_options=sess_options, providers=["CPUExecutionProvider"])
|
@@ -43,7 +61,7 @@ def tts(text):
|
|
43 |
for stn_tst in stn_tst_list:
|
44 |
text = np.expand_dims(np.array(stn_tst, dtype=np.int64), 0)
|
45 |
text_lengths = np.array([text.shape[1]], dtype=np.int64)
|
46 |
-
scales = np.array([0.667, 1.
|
47 |
sid = np.array([int(sid)]) if sid is not None else None
|
48 |
|
49 |
audio = model.run(
|
|
|
10 |
from scipy.io.wavfile import write
|
11 |
from preprocess import preprocess
|
12 |
import onnxruntime
|
13 |
+
import re
|
14 |
|
15 |
def get_text(texts, hps):
|
16 |
text_norm_list = []
|
17 |
+
texts = texts.replace("“", "")
|
18 |
+
texts = texts.replace("”", "")
|
19 |
+
texts = texts.replace("…", ".")
|
20 |
+
texts = re.split('[!:;\.\n]', texts)
|
21 |
+
tmp = []
|
22 |
+
for t in texts:
|
23 |
+
t = t.rstrip().strip()
|
24 |
+
if len(t) > 0:
|
25 |
+
tmp.append(t)
|
26 |
+
texts = []
|
27 |
+
for t in tmp:
|
28 |
+
texts.append(t)
|
29 |
+
texts.append(".")
|
30 |
+
#texts.append(".")
|
31 |
+
|
32 |
+
#print(texts)
|
33 |
+
for text in texts:
|
34 |
text = preprocess(text)
|
35 |
+
print(text)
|
36 |
chunk_strings = []
|
37 |
chunk_len = 30
|
38 |
for i in range(0, len(text.split()), chunk_len):
|
|
|
47 |
|
48 |
def tts(text):
|
49 |
model_path = "model.onnx"
|
50 |
+
config_path = "configs/thu_hue.json"
|
51 |
+
sid = 9
|
52 |
output_wav_path = "output.wav"
|
53 |
sess_options = onnxruntime.SessionOptions()
|
54 |
model = onnxruntime.InferenceSession(str(model_path), sess_options=sess_options, providers=["CPUExecutionProvider"])
|
|
|
61 |
for stn_tst in stn_tst_list:
|
62 |
text = np.expand_dims(np.array(stn_tst, dtype=np.int64), 0)
|
63 |
text_lengths = np.array([text.shape[1]], dtype=np.int64)
|
64 |
+
scales = np.array([0.667, 1.08, 0.8], dtype=np.float32)
|
65 |
sid = np.array([int(sid)]) if sid is not None else None
|
66 |
|
67 |
audio = model.run(
|
model.onnx
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c781e3fdb725651e0074d4549e272fe22d4cc71872b098bbc62d6bb995b567c4
|
3 |
+
size 130096954
|