Spaces:
Runtime error
Runtime error
clementruhm
commited on
Commit
•
9ef73e2
1
Parent(s):
4263bcd
reload model on synthesis if needed
Browse files
app.py
CHANGED
@@ -15,6 +15,10 @@ from huggingface_hub import hf_hub_download, list_repo_files
|
|
15 |
|
16 |
# global tts module, initialized from a model selected
|
17 |
tts = None
|
|
|
|
|
|
|
|
|
18 |
model_repo_dir = "data"
|
19 |
for name in list_repo_files(repo_id="balacoon/tts"):
|
20 |
hf_hub_download(
|
@@ -58,14 +62,20 @@ def main():
|
|
58 |
|
59 |
def set_model(model_name_str: str):
|
60 |
"""
|
61 |
-
gets value from `model_name
|
62 |
-
|
63 |
-
|
64 |
"""
|
65 |
-
|
66 |
-
|
67 |
-
|
68 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
value = speakers[-1]
|
70 |
return gr.Dropdown.update(
|
71 |
choices=speakers, value=value, visible=True
|
@@ -78,23 +88,28 @@ def main():
|
|
78 |
with gr.Row(variant="panel"):
|
79 |
audio = gr.Audio()
|
80 |
|
81 |
-
def synthesize_audio(text_str: str,
|
82 |
"""
|
83 |
gets utterance to synthesize from `text` Textbox
|
84 |
and speaker name from `speaker` dropdown list.
|
85 |
speaker name might be empty for single-speaker models.
|
86 |
Synthesizes the waveform and updates `audio` with it.
|
87 |
"""
|
88 |
-
if not text_str:
|
89 |
-
logging.info("text or speaker are not provided")
|
90 |
return None
|
91 |
-
|
|
|
|
|
|
|
|
|
92 |
if len(text_str) > 1024:
|
|
|
93 |
text_str = text_str[:1024]
|
94 |
-
samples =
|
95 |
-
return gr.Audio.update(value=(
|
96 |
|
97 |
-
generate.click(synthesize_audio, inputs=[text, speaker], outputs=audio)
|
98 |
|
99 |
demo.queue(concurrency_count=1).launch()
|
100 |
|
|
|
15 |
|
16 |
# global tts module, initialized from a model selected
|
17 |
tts = None
|
18 |
+
# path to the model that is currently used in tts
|
19 |
+
cur_model_path = None
|
20 |
+
# cache of speakers, maps model name to speaker list
|
21 |
+
model_to_speakers = dict()
|
22 |
model_repo_dir = "data"
|
23 |
for name in list_repo_files(repo_id="balacoon/tts"):
|
24 |
hf_hub_download(
|
|
|
62 |
|
63 |
def set_model(model_name_str: str):
|
64 |
"""
|
65 |
+
gets value from `model_name`. either
|
66 |
+
uses cached list of speakers for the given model name
|
67 |
+
or loads the addon and checks what are the speakers.
|
68 |
"""
|
69 |
+
if model_name_str in model_to_speakers:
|
70 |
+
speakers = model_to_speakers[model_name_str]
|
71 |
+
else:
|
72 |
+
# need to load this model to learn the list of speakers
|
73 |
+
model_path = os.path.join(model_repo_dir, model_name_str)
|
74 |
+
tts = TTS(model_path)
|
75 |
+
cur_model_path = model_path
|
76 |
+
speakers = tts.get_speakers()
|
77 |
+
model_to_speakers[model_name_str] = speakers
|
78 |
+
|
79 |
value = speakers[-1]
|
80 |
return gr.Dropdown.update(
|
81 |
choices=speakers, value=value, visible=True
|
|
|
88 |
with gr.Row(variant="panel"):
|
89 |
audio = gr.Audio()
|
90 |
|
91 |
+
def synthesize_audio(text_str: str, model_name_str: str, speaker_str: str):
|
92 |
"""
|
93 |
gets utterance to synthesize from `text` Textbox
|
94 |
and speaker name from `speaker` dropdown list.
|
95 |
speaker name might be empty for single-speaker models.
|
96 |
Synthesizes the waveform and updates `audio` with it.
|
97 |
"""
|
98 |
+
if not text_str or not model_name_str or not speaker_str:
|
99 |
+
logging.info("text, model name or speaker are not provided")
|
100 |
return None
|
101 |
+
expected_model_path = os.path.join(model_repo_dir, model_name_str)
|
102 |
+
if expected_model_path != cur_model_path:
|
103 |
+
# reload model
|
104 |
+
tts = TTS(expected_model_path)
|
105 |
+
cur_model_path = expected_model_path
|
106 |
if len(text_str) > 1024:
|
107 |
+
# truncate the text
|
108 |
text_str = text_str[:1024]
|
109 |
+
samples = tts.synthesize(text_str, speaker_str)
|
110 |
+
return gr.Audio.update(value=(tts.get_sampling_rate(), samples))
|
111 |
|
112 |
+
generate.click(synthesize_audio, inputs=[text, model_name, speaker], outputs=audio)
|
113 |
|
114 |
demo.queue(concurrency_count=1).launch()
|
115 |
|