Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
os.system("pip install speechbrain torchaudio gradio")
|
3 |
+
|
4 |
+
import torchaudio
|
5 |
+
from speechbrain.pretrained import Tacotron2, HIFIGAN
|
6 |
+
import gradio as gr
|
7 |
+
|
8 |
+
# Load SpeechBrain models
|
9 |
+
print("Loading SpeechBrain models...")
|
10 |
+
tacotron2 = Tacotron2.from_hparams(source="speechbrain/tts-tacotron2-ljspeech", savedir="models/tacotron2")
|
11 |
+
hifi_gan = HIFIGAN.from_hparams(source="speechbrain/tts-hifigan-ljspeech", savedir="models/hifigan")
|
12 |
+
print("Models loaded! Ready to generate speech.")
|
13 |
+
|
14 |
+
# Generate speech from text
|
15 |
+
def text_to_speech(text):
|
16 |
+
mel_output, _, _ = tacotron2.encode_text(text)
|
17 |
+
waveforms = hifi_gan.decode_batch(mel_output)
|
18 |
+
output_file = "output_speech.wav"
|
19 |
+
torchaudio.save(output_file, waveforms.squeeze(1), 22050)
|
20 |
+
return output_file # Return the audio file
|
21 |
+
|
22 |
+
# Create Gradio UI
|
23 |
+
iface = gr.Interface(
|
24 |
+
fn=text_to_speech,
|
25 |
+
inputs=gr.Textbox(label="Text to speak"),
|
26 |
+
outputs=gr.Audio(type="file", label="Generated Speech"),
|
27 |
+
title="SpeechBrain TTS Demo",
|
28 |
+
description="Enter text and get an AI-generated voice output!"
|
29 |
+
)
|
30 |
+
|
31 |
+
# Launch the app
|
32 |
+
iface.launch()
|