Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -11,16 +11,10 @@ from datasets import load_dataset
|
|
11 |
|
12 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
13 |
|
14 |
-
def load_models_and_data(
|
15 |
model_name = "microsoft/speecht5_tts"
|
16 |
processor = SpeechT5Processor.from_pretrained(model_name)
|
17 |
-
|
18 |
-
# Replace with English technical TTS model or regional language-specific model
|
19 |
-
if language == "en":
|
20 |
-
model = SpeechT5ForTextToSpeech.from_pretrained("my_finetuned_english_tech_tts").to(device)
|
21 |
-
else:
|
22 |
-
model = SpeechT5ForTextToSpeech.from_pretrained("my_finetuned_regional_language_tts").to(device)
|
23 |
-
|
24 |
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
|
25 |
|
26 |
spk_model_name = "speechbrain/spkrec-xvect-voxceleb"
|
@@ -31,15 +25,13 @@ def load_models_and_data(language="en"):
|
|
31 |
)
|
32 |
|
33 |
# Load a sample from a dataset for default embedding
|
34 |
-
|
35 |
-
|
36 |
-
else:
|
37 |
-
dataset = load_dataset("regional_language_dataset", split="train")
|
38 |
-
|
39 |
-
example = dataset[0]
|
40 |
|
41 |
return model, processor, vocoder, speaker_model, example
|
42 |
|
|
|
|
|
43 |
# Choose the language dynamically (English or Regional Language)
|
44 |
model, processor, vocoder, speaker_model, default_example = load_models_and_data(language="en")
|
45 |
|
|
|
11 |
|
12 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
13 |
|
14 |
+
def load_models_and_data():
|
15 |
model_name = "microsoft/speecht5_tts"
|
16 |
processor = SpeechT5Processor.from_pretrained(model_name)
|
17 |
+
model = SpeechT5ForTextToSpeech.from_pretrained("emirhanbilgic/speecht5_finetuned_emirhan_tr").to(device)
|
|
|
|
|
|
|
|
|
|
|
|
|
18 |
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
|
19 |
|
20 |
spk_model_name = "speechbrain/spkrec-xvect-voxceleb"
|
|
|
25 |
)
|
26 |
|
27 |
# Load a sample from a dataset for default embedding
|
28 |
+
dataset = load_dataset("ylacombe/english_dialects ", split="train")
|
29 |
+
example = dataset[304]
|
|
|
|
|
|
|
|
|
30 |
|
31 |
return model, processor, vocoder, speaker_model, example
|
32 |
|
33 |
+
model, processor, vocoder, speaker_model, default_example = load_models_and_data()
|
34 |
+
|
35 |
# Choose the language dynamically (English or Regional Language)
|
36 |
model, processor, vocoder, speaker_model, default_example = load_models_and_data(language="en")
|
37 |
|