Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import torch
|
3 |
+
import gradio as gr
|
4 |
+
from huggingface_hub import hf_hub_download
|
5 |
+
import langid
|
6 |
+
from openvoice.api import BaseSpeakerTTS, ToneColorConverter
|
7 |
+
import openvoice.se_extractor as se_extractor
|
8 |
+
|
9 |
+
# Constants
|
10 |
+
CKPT_BASE_PATH = "checkpoints"
|
11 |
+
EN_SUFFIX = f"{CKPT_BASE_PATH}/base_speakers/EN"
|
12 |
+
CONVERTER_SUFFIX = f"{CKPT_BASE_PATH}/converter"
|
13 |
+
OUTPUT_DIR = "outputs/"
|
14 |
+
os.makedirs(OUTPUT_DIR, exist_ok=True)
|
15 |
+
|
16 |
+
# Download necessary files
|
17 |
+
def download_from_hf_hub(filename, local_dir="./"):
|
18 |
+
os.makedirs(local_dir, exist_ok=True)
|
19 |
+
hf_hub_download(repo_id="myshell-ai/OpenVoice", filename=filename, local_dir=local_dir)
|
20 |
+
|
21 |
+
for file in [f"{CONVERTER_SUFFIX}/checkpoint.pth", f"{CONVERTER_SUFFIX}/config.json",
|
22 |
+
f"{EN_SUFFIX}/checkpoint.pth", f"{EN_SUFFIX}/config.json",
|
23 |
+
f"{EN_SUFFIX}/en_default_se.pth", f"{EN_SUFFIX}/en_style_se.pth"]:
|
24 |
+
download_from_hf_hub(file)
|
25 |
+
|
26 |
+
# Initialize models
|
27 |
+
pt_device = "cpu"
|
28 |
+
en_base_speaker_tts = BaseSpeakerTTS(f"{EN_SUFFIX}/config.json", device=pt_device)
|
29 |
+
en_base_speaker_tts.load_ckpt(f"{EN_SUFFIX}/checkpoint.pth")
|
30 |
+
|
31 |
+
tone_color_converter = ToneColorConverter(f"{CONVERTER_SUFFIX}/config.json", device=pt_device)
|
32 |
+
tone_color_converter.load_ckpt(f"{CONVERTER_SUFFIX}/checkpoint.pth")
|
33 |
+
|
34 |
+
en_source_default_se = torch.load(f"{EN_SUFFIX}/en_default_se.pth")
|
35 |
+
en_source_style_se = torch.load(f"{EN_SUFFIX}/en_style_se.pth")
|
36 |
+
|
37 |
+
# Main prediction function
|
38 |
+
def predict(prompt, style, audio_file_pth, tau):
|
39 |
+
if len(prompt) < 2 or len(prompt) > 200:
|
40 |
+
return "Text should be between 2 and 200 characters.", None
|
41 |
+
|
42 |
+
try:
|
43 |
+
target_se, _ = se_extractor.get_se(audio_file_pth, tone_color_converter, target_dir=OUTPUT_DIR, vad=True)
|
44 |
+
except Exception as e:
|
45 |
+
return f"Error getting target tone color: {str(e)}", None
|
46 |
+
|
47 |
+
src_path = f"{OUTPUT_DIR}/tmp.wav"
|
48 |
+
en_base_speaker_tts.tts(prompt, src_path, speaker=style, language="English")
|
49 |
+
|
50 |
+
save_path = f"{OUTPUT_DIR}/output.wav"
|
51 |
+
tone_color_converter.convert(
|
52 |
+
audio_src_path=src_path,
|
53 |
+
src_se=en_source_style_se if style != "default" else en_source_default_se,
|
54 |
+
tgt_se=target_se,
|
55 |
+
output_path=save_path,
|
56 |
+
tau=tau
|
57 |
+
)
|
58 |
+
|
59 |
+
return "Voice cloning completed successfully.", save_path
|
60 |
+
|
61 |
+
# Gradio interface
|
62 |
+
def create_demo():
|
63 |
+
with gr.Blocks() as demo:
|
64 |
+
gr.Markdown("# OpenVoice: Instant Voice Cloning with fine-tuning")
|
65 |
+
|
66 |
+
with gr.Row():
|
67 |
+
input_text = gr.Textbox(label="Text to speak", placeholder="Enter text here (2-200 characters)")
|
68 |
+
style = gr.Dropdown(
|
69 |
+
label="Style",
|
70 |
+
choices=["default", "whispering", "cheerful", "terrified", "angry", "sad", "friendly"],
|
71 |
+
value="default"
|
72 |
+
)
|
73 |
+
|
74 |
+
with gr.Row():
|
75 |
+
reference_audio = gr.Audio(label="Reference Audio", type="filepath")
|
76 |
+
tau_slider = gr.Slider(minimum=0.1, maximum=1.0, value=0.7, label="Tau (Voice similarity)", info="Higher values make the output more similar to the reference voice")
|
77 |
+
|
78 |
+
submit_button = gr.Button("Generate Voice")
|
79 |
+
|
80 |
+
output_text = gr.Textbox(label="Status")
|
81 |
+
output_audio = gr.Audio(label="Generated Audio")
|
82 |
+
|
83 |
+
submit_button.click(
|
84 |
+
predict,
|
85 |
+
inputs=[input_text, style, reference_audio, tau_slider],
|
86 |
+
outputs=[output_text, output_audio]
|
87 |
+
)
|
88 |
+
|
89 |
+
return demo
|
90 |
+
|
91 |
+
# Launch the demo
|
92 |
+
if __name__ == "__main__":
|
93 |
+
demo = create_demo()
|
94 |
+
demo.launch()
|