Update app.py
Browse files
app.py
CHANGED
@@ -41,20 +41,38 @@ def synthesize_speech(text):
|
|
41 |
return (16000, speech.cpu().numpy())
|
42 |
|
43 |
# Title and description for the Gradio interface
|
44 |
-
title = "Fine-tuning TTS for a
|
45 |
-
description = """
|
46 |
-
This Space generates speech in Italian using
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
"""
|
49 |
|
50 |
-
# Create Gradio interface
|
51 |
interface = gr.Interface(
|
52 |
fn=synthesize_speech,
|
53 |
inputs=gr.Textbox(label="Input Text", placeholder="Enter Italian text here..."),
|
54 |
outputs=gr.Audio(label="Generated Speech"),
|
55 |
title=title,
|
56 |
description=description,
|
57 |
-
examples=[
|
|
|
|
|
|
|
|
|
|
|
|
|
58 |
)
|
59 |
|
60 |
# Launch the interface
|
|
|
41 |
return (16000, speech.cpu().numpy())
|
42 |
|
43 |
# Title and description for the Gradio interface
|
44 |
+
title = "Fine-tuning TTS for Italian as a Regional Language Using SpeechT5"
|
45 |
+
description = f"""
|
46 |
+
This Space generates speech in Italian, a regional language, using a fine-tuned SpeechT5 model from Hugging Face.
|
47 |
+
Italian is considered a regional language because it is primarily spoken within Italy and a few Italian-speaking regions in
|
48 |
+
other countries, such as Switzerland, San Marino, Vatican City, and areas in Croatia and Slovenia.
|
49 |
+
With about 85 million speakers worldwide, Italian's regional usage contrasts with the global reach of languages like English or Spanish.
|
50 |
+
|
51 |
+
**Fine-Tuned Model Preparation:** This model has been fine-tuned using the VoxPopuli Italian dataset to optimize SpeechT5 for
|
52 |
+
Italian pronunciation, intonation, and fluency. The fine-tuning process involved preprocessing the text data to ensure accurate
|
53 |
+
Italian accents and phonetics, resulting in high-quality Italian speech synthesis.
|
54 |
+
|
55 |
+
The fine-tuned model is available [here](https://huggingface.co/Vinay15/speecht5_finetuned_voxpopuli_it).
|
56 |
+
|
57 |
+
**Note:** Processing time may vary based on sentence length. Longer sentences may take more time to process and generate audio.
|
58 |
+
|
59 |
+
For more details, visit the [GitHub repository](https://github.com/Vinay152003/Fine-tuning-TTS-for-a-Italian-it-Language) and review the project [report](https://drive.google.com/file/d/1cvNPkuFlTZAu1iDaagCwVRGXFd6r6vqi/view?usp=sharing).
|
60 |
"""
|
61 |
|
62 |
+
# Create Gradio interface with multiple examples
|
63 |
interface = gr.Interface(
|
64 |
fn=synthesize_speech,
|
65 |
inputs=gr.Textbox(label="Input Text", placeholder="Enter Italian text here..."),
|
66 |
outputs=gr.Audio(label="Generated Speech"),
|
67 |
title=title,
|
68 |
description=description,
|
69 |
+
examples=[
|
70 |
+
["Questa è una dimostrazione di sintesi vocale in italiano."],
|
71 |
+
["Benvenuti alla nostra piattaforma di sintesi vocale!"],
|
72 |
+
["Il modello è stato addestrato per parlare l'italiano in modo naturale e fluido."],
|
73 |
+
["Oggi il tempo è bello e il sole splende."],
|
74 |
+
["La città di Roma è una delle destinazioni turistiche più popolari al mondo."]
|
75 |
+
]
|
76 |
)
|
77 |
|
78 |
# Launch the interface
|