app.py

import ssl
ssl._create_default_https_context = ssl._create_unverified_context

import gradio as gr
import wespeaker

STYLE = """
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/bootstrap@5.1.3/dist/css/bootstrap.min.css" integrity="sha256-YvdLHPgkqJ8DVUxjjnGVlMMJtNimJ6dYkowFFvp4kKs=" crossorigin="anonymous">
"""
OUTPUT_OK = (STYLE + """
    <div class="container">
        <div class="row"><h1 style="text-align: center">The speakers are</h1></div>
        <div class="row"><h1 class="display-1 text-success" style="text-align: center">{:.1f}%</h1></div>
        <div class="row"><h1 style="text-align: center">similar</h1></div>
        <div class="row"><h1 class="text-success" style="text-align: center">Welcome, brother!</h1></div>
    </div>
""")
OUTPUT_FAIL = (STYLE + """
    <div class="container">
        <div class="row"><h1 style="text-align: center">The speakers are</h1></div>
        <div class="row"><h1 class="display-1 text-danger" style="text-align: center">{:.1f}%</h1></div>
        <div class="row"><h1 style="text-align: center">similar</h1></div>
        <div class="row"><h1 class="text-danger" style="text-align: center">Warning! stranger!</h1></div>
       
    </div>
""")

OUTPUT_ERROR = (STYLE + """
    <div class="container">
        <div class="row"><h1 style="text-align: center">Input Error</h1></div>
        <div class="row"><h1 class="text-danger" style="text-align: center">{}!</h1></div>
    </div>
""")

cn_model = wespeaker.load_model("chinese")
en_model = wespeaker.load_model("english")


def speaker_verification(audio_path1, audio_path2, lang='CN'):
    if audio_path1 == None or audio_path2 == None:
        output = OUTPUT_ERROR.format('Please enter two audios')
        return output
    if lang == 'EN':
        model = en_model
    elif lang == 'CN':
        model = cn_model
    else:
        output = OUTPUT_ERROR.format('Please select a language')
        return output
    cos_score = model.compute_similarity(audio_path1, audio_path2)

    if cos_score >= 0.83:
        output = OUTPUT_OK.format(cos_score * 100)
    else:
        output = OUTPUT_FAIL.format(cos_score * 100)

    return output


# input
inputs = [
    gr.inputs.Audio(source="microphone",
                    type="filepath",
                    optional=True,
                    label='Speaker#1'),
    gr.inputs.Audio(source="microphone",
                    type="filepath",
                    optional=True,
                    label='Speaker#2'),
    gr.Radio(['EN', 'CN'], label='Language'),
]

output = gr.outputs.HTML(label="")

# description
description = ("<p>VoicePrint Demo ! Note: We recommend that the audio length be greater than 2s !</p>")


examples = [
    ['examples/BAC009S0764W0228.wav', 'examples/BAC009S0764W0328.wav', 'CN'],
    ['examples/BAC009S0913W0133.wav', 'examples/BAC009S0764W0228.wav', 'CN'],
    ['examples/00001_spk1.wav', 'examples/00003_spk2.wav', 'EN'],
    ['examples/00010_spk2.wav', 'examples/00024_spk1.wav', 'EN'],
    ['examples/00001_spk1.wav', 'examples/00024_spk1.wav', 'EN'],
    ['examples/00010_spk2.wav', 'examples/00003_spk2.wav', 'EN'],
]

interface = gr.Interface(
    fn=speaker_verification,
    inputs=inputs,
    outputs=output,
    title="VoicePrint Recognition-声纹识别",
    description=description,
    examples=examples,
    theme="huggingface",
)

interface.launch(enable_queue=True,server_name="0.0.0.0",server_port=10001)