Update app.py
Browse files
app.py
CHANGED
@@ -3,8 +3,8 @@ import torch
|
|
3 |
from datasets import load_dataset
|
4 |
from transformers import SpeechT5Processor, SpeechT5HifiGan, SpeechT5ForTextToSpeech
|
5 |
|
6 |
-
# Load the fine-tuned model and vocoder for Italian
|
7 |
-
model_id = "
|
8 |
model = SpeechT5ForTextToSpeech.from_pretrained(model_id)
|
9 |
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
|
10 |
|
@@ -12,7 +12,7 @@ vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
|
|
12 |
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
|
13 |
speaker_embeddings = torch.tensor(embeddings_dataset[7440]["xvector"]).unsqueeze(0)
|
14 |
|
15 |
-
# Load processor for the Italian model
|
16 |
processor = SpeechT5Processor.from_pretrained(model_id)
|
17 |
|
18 |
# Optional: Text cleanup for Italian-specific characters
|
@@ -27,7 +27,7 @@ replacements = [
|
|
27 |
|
28 |
# Text-to-speech synthesis function
|
29 |
def synthesize_speech(text):
|
30 |
-
# Clean up text
|
31 |
for src, dst in replacements:
|
32 |
text = text.replace(src, dst)
|
33 |
|
@@ -58,4 +58,4 @@ interface = gr.Interface(
|
|
58 |
)
|
59 |
|
60 |
# Launch the interface
|
61 |
-
interface.launch()
|
|
|
3 |
from datasets import load_dataset
|
4 |
from transformers import SpeechT5Processor, SpeechT5HifiGan, SpeechT5ForTextToSpeech
|
5 |
|
6 |
+
# Load the fine-tuned model and vocoder for Italian from the new model ID
|
7 |
+
model_id = "Vinay15/speecht5_finetuned_voxpopuli_it"
|
8 |
model = SpeechT5ForTextToSpeech.from_pretrained(model_id)
|
9 |
vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
|
10 |
|
|
|
12 |
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
|
13 |
speaker_embeddings = torch.tensor(embeddings_dataset[7440]["xvector"]).unsqueeze(0)
|
14 |
|
15 |
+
# Load processor for the new Italian model
|
16 |
processor = SpeechT5Processor.from_pretrained(model_id)
|
17 |
|
18 |
# Optional: Text cleanup for Italian-specific characters
|
|
|
27 |
|
28 |
# Text-to-speech synthesis function
|
29 |
def synthesize_speech(text):
|
30 |
+
# Clean up text for Italian-specific accents
|
31 |
for src, dst in replacements:
|
32 |
text = text.replace(src, dst)
|
33 |
|
|
|
58 |
)
|
59 |
|
60 |
# Launch the interface
|
61 |
+
interface.launch()
|