Update app.py
Browse files
app.py
CHANGED
|
@@ -46,39 +46,16 @@ for v in voicelist:
|
|
| 46 |
def synthesize(text, voice, lngsteps, password, progress=gr.Progress()):
|
| 47 |
if text.strip() == "":
|
| 48 |
raise gr.Error("You must enter some text")
|
| 49 |
-
|
| 50 |
-
raise gr.Error("Text must be <50k characters")
|
| 51 |
texts = split_and_recombine_text(text)
|
| 52 |
v = voice.lower()
|
| 53 |
audios = []
|
| 54 |
for t in progress.tqdm(texts):
|
| 55 |
audios.append(styletts2importable.inference(t, voices[v], alpha=0.3, beta=0.7, diffusion_steps=lngsteps, embedding_scale=1))
|
| 56 |
return (24000, np.concatenate(audios))
|
| 57 |
-
|
| 58 |
-
# if password == os.environ['ACCESS_CODE']:
|
| 59 |
-
# if text.strip() == "":
|
| 60 |
-
# raise gr.Error("You must enter some text")
|
| 61 |
-
# if lngsteps > 25:
|
| 62 |
-
# raise gr.Error("Max 25 steps")
|
| 63 |
-
# if lngsteps < 5:
|
| 64 |
-
# raise gr.Error("Min 5 steps")
|
| 65 |
-
# texts = split_and_recombine_text(text)
|
| 66 |
-
# v = voice.lower()
|
| 67 |
-
# audios = []
|
| 68 |
-
# for t in progress.tqdm(texts):
|
| 69 |
-
# audios.append(styletts2importable.inference(t, voices[v], alpha=0.3, beta=0.7, diffusion_steps=lngsteps, embedding_scale=1))
|
| 70 |
-
# return (24000, np.concatenate(audios))
|
| 71 |
-
# else:
|
| 72 |
-
# raise gr.Error('Wrong access code')
|
| 73 |
def clsynthesize(text, voice, vcsteps, progress=gr.Progress()):
|
| 74 |
-
|
| 75 |
-
# raise gr.Error("You must enter some text")
|
| 76 |
-
# # if global_phonemizer.phonemize([text]) > 300:
|
| 77 |
-
# if len(text) > 400:
|
| 78 |
-
# raise gr.Error("Text must be under 400 characters")
|
| 79 |
-
# # return (24000, styletts2importable.inference(text, styletts2importable.compute_style(voice), alpha=0.3, beta=0.7, diffusion_steps=20, embedding_scale=1))
|
| 80 |
-
# return (24000, styletts2importable.inference(text, styletts2importable.compute_style(voice), alpha=0.3, beta=0.7, diffusion_steps=vcsteps, embedding_scale=1))
|
| 81 |
-
if text.strip() == "":
|
| 82 |
raise gr.Error("You must enter some text")
|
| 83 |
if len(text) > 50000:
|
| 84 |
raise gr.Error("Text must be <50k characters")
|
|
@@ -88,11 +65,6 @@ def clsynthesize(text, voice, vcsteps, progress=gr.Progress()):
|
|
| 88 |
audios.append(styletts2importable.inference(t, styletts2importable.compute_style(voice), alpha=0.3, beta=0.7, diffusion_steps=vcsteps, embedding_scale=1))
|
| 89 |
return (24000, np.concatenate(audios))
|
| 90 |
def ljsynthesize(text, steps, progress=gr.Progress()):
|
| 91 |
-
# if text.strip() == "":
|
| 92 |
-
# raise gr.Error("You must enter some text")
|
| 93 |
-
# # if global_phonemizer.phonemize([text]) > 300:
|
| 94 |
-
# if len(text) > 400:
|
| 95 |
-
# raise gr.Error("Text must be under 400 characters")
|
| 96 |
noise = torch.randn(1,1,256).to('cuda' if torch.cuda.is_available() else 'cpu')
|
| 97 |
# return (24000, ljspeechimportable.inference(text, noise, diffusion_steps=7, embedding_scale=1))
|
| 98 |
if text.strip() == "":
|
|
|
|
| 46 |
def synthesize(text, voice, lngsteps, password, progress=gr.Progress()):
|
| 47 |
if text.strip() == "":
|
| 48 |
raise gr.Error("You must enter some text")
|
| 49 |
+
|
|
|
|
| 50 |
texts = split_and_recombine_text(text)
|
| 51 |
v = voice.lower()
|
| 52 |
audios = []
|
| 53 |
for t in progress.tqdm(texts):
|
| 54 |
audios.append(styletts2importable.inference(t, voices[v], alpha=0.3, beta=0.7, diffusion_steps=lngsteps, embedding_scale=1))
|
| 55 |
return (24000, np.concatenate(audios))
|
| 56 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
def clsynthesize(text, voice, vcsteps, progress=gr.Progress()):
|
| 58 |
+
if text.strip() == "":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 59 |
raise gr.Error("You must enter some text")
|
| 60 |
if len(text) > 50000:
|
| 61 |
raise gr.Error("Text must be <50k characters")
|
|
|
|
| 65 |
audios.append(styletts2importable.inference(t, styletts2importable.compute_style(voice), alpha=0.3, beta=0.7, diffusion_steps=vcsteps, embedding_scale=1))
|
| 66 |
return (24000, np.concatenate(audios))
|
| 67 |
def ljsynthesize(text, steps, progress=gr.Progress()):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 68 |
noise = torch.randn(1,1,256).to('cuda' if torch.cuda.is_available() else 'cpu')
|
| 69 |
# return (24000, ljspeechimportable.inference(text, noise, diffusion_steps=7, embedding_scale=1))
|
| 70 |
if text.strip() == "":
|