Spaces:
Running
Running
Delete app.py
Browse files
app.py
DELETED
@@ -1,56 +0,0 @@
|
|
1 |
-
import gradio as gr
|
2 |
-
import torch
|
3 |
-
import tempfile
|
4 |
-
import soundfile as sf
|
5 |
-
from tortoise.api import TextToSpeech
|
6 |
-
from tortoise.utils.audio import load_audio
|
7 |
-
|
8 |
-
# 1) Initialize the Tortoise TTS engine at startup
|
9 |
-
tts = TextToSpeech() # Downloads and caches models automatically
|
10 |
-
|
11 |
-
# 2) Define a helper to generate speech from a reference clip + text
|
12 |
-
def generate_speech(reference_audio_path, text):
|
13 |
-
"""
|
14 |
-
reference_audio_path: filepath to a WAV sampled at 22 050 Hz
|
15 |
-
text: the string to synthesize
|
16 |
-
returns: path to a 24 kHz WAV file with your cloned voice
|
17 |
-
"""
|
18 |
-
# β
FIXED: Provide sampling_rate as a required positional argument
|
19 |
-
ref_waveform = load_audio(reference_audio_path, 22050)
|
20 |
-
|
21 |
-
# Generate speech using 'fast' preset (alternatives: ultra_fast, standard, high_quality)
|
22 |
-
output_tensor = tts.tts_with_preset(
|
23 |
-
text,
|
24 |
-
voice_samples=[ref_waveform],
|
25 |
-
preset="fast"
|
26 |
-
)
|
27 |
-
|
28 |
-
# Save to temp WAV (float32, 24 kHz)
|
29 |
-
wav_np = output_tensor.squeeze().cpu().numpy()
|
30 |
-
tmp = tempfile.NamedTemporaryFile(suffix=".wav", delete=False)
|
31 |
-
sf.write(tmp.name, wav_np, samplerate=24000)
|
32 |
-
return tmp.name
|
33 |
-
|
34 |
-
# 3) Build the Gradio interface
|
35 |
-
with gr.Blocks(title="Tortoise Voice Cloning TTS") as app:
|
36 |
-
gr.Markdown("## π£οΈ Voice Cloning with Tortoise TTS")
|
37 |
-
gr.Markdown(
|
38 |
-
"Upload a ~10 sec WAV clip (22 050 Hz), enter English text, "
|
39 |
-
"and hear it spoken back in **your** voice!"
|
40 |
-
)
|
41 |
-
|
42 |
-
with gr.Row():
|
43 |
-
voice_sample = gr.Audio(type="filepath", label="ποΈ Upload Reference Voice (22 050 Hz WAV)")
|
44 |
-
text_input = gr.Textbox(label="π¬ Text to Synthesize", placeholder="e.g., Hello, world!")
|
45 |
-
|
46 |
-
generate_btn = gr.Button("π Generate Speech")
|
47 |
-
output_audio = gr.Audio(label="π’ Cloned Speech Output (24 kHz)", interactive=False)
|
48 |
-
|
49 |
-
generate_btn.click(
|
50 |
-
fn=generate_speech,
|
51 |
-
inputs=[voice_sample, text_input],
|
52 |
-
outputs=output_audio
|
53 |
-
)
|
54 |
-
|
55 |
-
if __name__ == "__main__":
|
56 |
-
app.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|